China's ability to produce lithography tools is comparable to that of market leader ASML sometime in 2004, an analyst with UBS wrote in a note to clients. The situation may change in the next two or five years when Chinese companies start producing immersion DUV lithography systems in mass quantities, and Chinese chipmakers begin to deploy them for production of actual chips. However, China's semiconductor industry will remain well behind Western industry. "They seem to be at a similar stage to ASML in 2004," wrote Francois-Xavier Bouvignies, an analyst with UBS, in a note for clients, reports Bloomberg.

For years, China's pursuit of semiconductor self-sufficiency stemmed from its ability to produce mainstream chips on trailing nodes using fairly advanced, though not the latest, tools from leading producers such as ASML, KLA, and Lam Research. In recent years, China found itself in a new reality in which it could no longer obtain the latest chipmaking tools and had to build them domestically. Although companies like ACM Research, AMEC, and Naura have developed world-class chemical wafer deposition, cleaning, etching, and oxidation/diffusion tools that are now mass-produced and used by Chinese chipmakers, none of the Chinese companies have managed to develop a competitive lithography machine that can be used to make chips on more or less modern nodes and initiate its mass production.

Photolithography is generally considered the most technologically complex and demanding individual process in advanced semiconductor manufacturing. Firstly, lithography systems themselves are extraordinarily complex and contain tens of thousands of individual components. Secondly, the required positional accuracy is extraordinary, as a modern lithography scanner needs excellent resolution, overlay, focus control, scanner-to-scanner matching, CD uniformity, and line edge roughness, just to name some of the requirements. Finally, a competitive litho system must guarantee predictable uptime, defect density, and performance. Perhaps the key thing here is that all of the required features must be achieved without compromises, as, for example, a machine with high resolution and ideal uniformity that can process one wafer per hour cannot be used for mass production.

Historically, over a dozen companies produced lithography tools. However, as they became more complex, only ASML, Canon, and Nikon survived, with ASML being the undisputed market leader and the only maker of EUV lithography scanners.

As China is essentially developing a parallel semiconductor ecosystem, there have been reports of multiple entities working on lithography systems, including Shanghai Micro Electronics Equipment (SMEE), AMIES (which seems to be a SMEE spin-off that includes Aishengna and Yuliangsheng units), SiCarrier (reportedly controlled by Huawei), and even Naura (which denies that it is developing litho tools). SMEE, which was established in 2002, is by far the most important established Chinese manufacturer of litho tools. Meanwhile, so far none of China-based makers of lithography have established mass production of immersion DUV scanners capable of producing chips at 45nm and below.

SMEE reportedly formally introduced its first immersion DUV lithography system called SSA/800-10W and capable of making chips on nodes down to 28nm back in 2023. However, there have been no evidence that SMEE has indeed started mass production of the SSA/800-10W and that it has been adopted by a single manufacturer for mass production of chips. While it is conceivable that not all chipmakers announce deployment of breakthrough tools, especially keeping in mind that Chinese vendors like SMEE have plenty of foreign suppliers, we would have seen at least some indirect evidence (starting from procurement/acceptance records as well as component orders all the way to job postings and scientific papers) that SMEE started shipments of the SSA/800-10W in 2023 – 2024 by now, assuming of course that there were any shipments.

Interestingly, but the reports about China-made immersion DUV scanners now mass-produced by Shanghai Aishengna Electronic Technology Group (a unit, or an affiliate of SMEE) re-emerged this July, again, without any evidence. This time around, the reports did not even mention targeted nodes or throughput capabilities. To make matters even more suspicious is the lack of reports about shipments of evaluation tools to chipmakers (like ASML does this with its High-NA EUV machines) as well as preliminary results of their process qualifications (like Intel does with ASML's High-NA EUV machines).

For a first-ever Chinese immersion scanner, it is reasonable to expect Chinese chipmakers to use it on engineering wafers, characterize it against ASML machines, develop recipes, identify drawbacks, and pass that information back to SMEE, something that should take about a year. Only once that first machine performs adequately does it make sense to order and qualify multiple SSA/800-10W units for mass production. In fact, such qualification will likely take another year for a single layer and more time for additional layers. To that end, an insertion of an all-new lithography scanner into an existing flow will take at least two years, but likely more. To that end, once SMEE (or its business units) and its customers figure out how the first Chinese immersion scanner should work, these scanners will still be far from mass deployment.

In any case, without any real indicators that Chinese makers of wafer fab tools can produce and ship immersion lithography scanners to customers, we can only state what the UBS analyst did: China's lithography industry is in a position where ASML was in the mid-2000s.

Samsung has unveiled a three-phase roadmap to progressively transform high-bandwidth memory (HBM) into an integrated memory-and-compute system, culminating in the company's zHBM architecture, which places the processor directly beneath the DRAM stack and eliminates the conventional 2.5D interposer link between the two. Detailing the roadmap at Hot Chips 2026, Samsung's Sangwook Han, of the company's DRAM design team, identified the base die as the key enabler of the evolution, which began with the company’s decision to manufacture the HBM base die on an advanced logic process.

In conventional HBM, the base die (B-die) was fabricated on the same DRAM process node as the core dies (C-dies) in the stack above. Starting with HBM4, Samsung moved the base die to a 4nm logic process, primarily to reduce power draw and minimize die area. Additionally, it gave Samsung a much more capable piece of silicon.

The company contends that a die built on the same class of logic process as XPUs could do much more than serve as a data interface. Samsung now plans to progressively offload more functions into the base die, eventually removing the physical gap between memory and the XPU entirely.

The current state of HBM and its growing constraints

The current HBM architecture comprises multiple DRAM core dies stacked vertically on a base die and connected through thousands of TSVs. The stack sits beside an XPU on an interposer, with the base die bridging the memory and compute silicon.

Bandwidth has been the main driver of HBM’s evolution. The current HBM4 stack has roughly 1 to 5 TB/s of bandwidth obtained through 1,000 to 2,000 I/Os running at about 8 to 16 Gbps each. These figures are expected to rise with upcoming HBM generations. The problem is that conventional ways of scaling bandwidth present significant challenges.

TSV signaling speed is difficult to increase, so HBM generations have added more TSVs instead. However, this consumes area and forces tighter TSV pitches. The PHY has also grown more demanding. HBM4 doubled the data I/O count from 1,024 to 2,048 DQs, and signaling speed keeps rising. Power is an even bigger issue. While energy per bit is improving, total HBM power continues to rise as bandwidth is scaling faster. Samsung says this is why HBM4 moves the base die to an advanced logic process, as the denser, more efficient logic reduces power draw.

This move underpins and enables the three-phase plan. An advanced logic node shrinks the interface circuitry while enabling the HBM base die to perform functions previously handled by the processor. Samsung calls this direction custom HBM, or cHBM, which keeps the conventional DRAM stack but customizes the logic underneath it for a specific accelerator.

Phase 1: Reclaim XPU area

The first phase is about handing processor area back to compute in what Samsung calls “XPU area reclamation.” AI accelerators are hitting familiar scaling walls, such as slowing process scaling and dies pressing against reticle and interposer limits. To expand compute, Samsung plans to evict non-compute blocks, moving their functions to the base die’s underutilized silicon.

The first target is the HBM Physical Interface (PHY), one of the largest blocks on the base die. Samsung proposes replacing the traditional interface with a much smaller die-to-die (D2D) link. On an 11 × 12.8mm HBM4 base die, the conventional PHY occupies more than 8 × 4mm, while the custom HBM D2D block is about 8.5 × 1.5mm, with channel depth cut from 5.5mm to 2mm. Because the matching interface on the XPU shrinks too, Samsung also reclaims processor silicon.

Conversely, shrinking the same power into less silicon increases power density and creates hotspots. Samsung’s answer is a Heat Path Block (HPB) that provides an alternative route for heat to exit the concentrated interface region. The company says an HPB covering more than half of the PHY can slash peak temperature by more than 35%.

The bigger Phase 1 change is moving the memory controller from the XPU to the custom HBM base die. Han estimated controllers account for 5 to 10% of an XPU's area — space that, refilled with compute, could yield a 10–20% performance gain. Moving the controller next to memory also enables a new SRAM-based repair scheme in which failed C-die addresses can be redirected to SRAM on the base die, avoiding the need to sacrifice an entire spare row or column for a single defective cell.

Phase 2: Making the die a more useful smart memory subsystem

Even with the controller moved in, Samsung says a substantial portion of the base-die area remains unused. Phase 2 fills that space with more functions, first with some relatively straightforward additions. The company proposes SoC-like telemetry and reliability features, including thermal, voltage, process, and aging sensors, as well as more advanced self-test hardware.

It also wants to use the edge of the base die for direct memory expansion, arguing that capacity is becoming as important as bandwidth. Dedicated controllers and PHYs could connect a secondary tier of external memory directly to custom HBM, rather than going through conventional PCIe expansion. Han said that extra memory could be LPDDR or even HBM, offering higher bandwidth and lower latency than PCIe-based memory extension.

Last in Phase 2 is compute — right on the base die. Samsung wants to place selected processing elements (PEs) under the DRAM, offloading memory-bound work while compute-heavy operations remain on the GPU. It calls this broader 2.5D architecture advanced HBM (aHBM), citing benefits such as less traffic across the interposer and reduced latency and I/O power draw.

Phase 3: zHBM goes fully 3D, placing the processor underneath the memory

Phase 3 appears to be Samsung's most radical step, with the company halting HBM architecture optimization and rebuilding it instead. Introducing zHBM, Samsung's “ultimate solution” for maximizing bandwidth under future AI's brutal power limits.

The zHBM concept eliminates the conventional side-by-side arrangement of XPU and HBM across an interposer. Instead, the processor sits directly beneath the DRAM stack in a true 3D structure. This architecture allows Samsung to replace the large edge PHY with distributed I/Os spread across the die. Data no longer has to travel laterally across an interposer, thereby shortening the physical path and eliminating the need for conventional HBM PHY and D2D link interfaces.

Samsung says the biggest payoff is power. Its projections show zHBM cutting I/O power by around 70% compared with HBM5. In another example, Samsung models roughly 2.3X more DRAM bandwidth while reducing memory power by about 100W compared to a four-stack HBM4E system.

On the flip side, thermals are the obvious complication. Han said Samsung is targeting roughly four-high zHBM stacks, compared with the much taller 12-high or 16-high configurations possible with conventional HBM, specifically because of heat. Distributed I/O helps by spreading the circuitry rather than concentrating it into hotspots, but zHBM is a balancing act involving capacity, bandwidth, heat, and physical integration.

Manufacturing zHBM will also require advanced wafer-on-wafer bonding and hybrid copper bonding to meet the required I/O density, with a much tighter co-design process between the DRAM and SoC teams. Samsung did not provide a firm launch date or timeline for the phases. However, HBM4’s 4nm logic base die is the concrete starting point, while cHBM and aHBM are nearer-term extensions, with zHBM as the long-term endpoint.

A team of scientists in Germany and Japan has demonstrated a solid-state cooling system that uses heat to generate the mechanical work required for refrigeration, potentially opening a route to processors and data centers that recycle some of their own waste heat for cooling. Developed by researchers at the Karlsruhe Institute of Technology (KIT) and the University of Tsukuba, the system — detailed in Nature Energy on August 28 — replaces the electrically powered actuator normally required for elastocaloric cooling with a heat-responsive shape-memory alloy, allowing the cooling cycle to run from an external heat source rather than a motor.

The prototype combines two ultra-thin metal films that serve as an actuator and a refrigerant. A 22-micrometer titanium-nickel (TiNi) shape-memory film contracts when heated, converting thermal energy into mechanical motion. This motion stretches and releases a 26.5-micrometer titanium-nickel-iron (TiNiFe) refrigerant film, triggering a reversible phase transition that produces cooling. In laboratory tests, Joule heating the actuator to 86°C produced a 12.9 K temperature span across the refrigerant film — the difference between its hottest and coldest states during the cooling cycle — and a 4.0 K span across the assembled cooling device, measured between its hot and cold sides. When the researchers replaced the resistance heating with an external 130°C heat source, the prototype still maintained a 2.2 K device-level temperature span, demonstrating that an external thermal source could drive the cooling mechanism.

Conventional refrigeration and air conditioning systems mostly use vapor-compression cooling. A compressor raises the pressure and temperature of a refrigerant, which then dumps heat in a condenser before expanding and evaporating at low pressure to absorb heat from the space being cooled. The technology is mature and efficient, but requires an electrically driven compressor and relies on refrigerants with significant global warming potential.

Solid-state cooling moves heat without the conventional compressor-and-refrigerant loop. Thermoelectric coolers, for example, use electrical current to create a temperature difference across semiconductor materials and are already common in compact electronics. However, the researchers note that thermoelectric devices typically reach only 10% to 15% of the theoretical reversed-Carnot efficiency limit, roughly one-quarter that of modern vapor-compression systems.

Elastocaloric cooling takes a different route. Certain shape-memory alloys change crystal structure when mechanically loaded and unloaded. Applying stress induces a phase transition that releases latent heat and warms the material. Once that heat is rejected, releasing the load reverses the transition, causing the material to absorb heat and cool. The solid alloy effectively becomes the refrigerant.

The problem is that the material still has to be repeatedly stretched and released. Existing elastocaloric systems generally use motors, hydraulic systems, or electromechanical actuators to provide the required force, adding electrical consumption, bulk, and mechanical complexity — particularly troublesome for miniature coolers. Instead, the KIT-Tsukuba team made one shape-memory alloy drive another. Heating the TiNi actuator film causes it to recover its original shape and contract. Mechanically coupled to the TiNiFe refrigerant film, that contraction supplies the force required for the cooling cycle. As the actuator heats and cools, it loads and unloads the refrigerant without an electric motor.

Broken down further, the system works as follows: the researchers take a shape-memory alloy (TiNiFe refrigerant film) that cools when released after being stretched. Instead of using a mechanical system to repeatedly stretch and release that alloy, they use another shape-memory alloy (TiNi) that contracts when heated and mechanically couple it to its cooling counterpart. When the TiNi is heated, it contracts, stretching and “loading” the refrigerant film. Once the heat is removed, the TiNi relaxes, releasing the film and triggering the phase transition that causes it to cool. Under cyclic heating, the TiNi alloy therefore provides the repeated stretching and releasing motion the TiNiFe requires for elastocaloric cooling.

The thermal actuator delivered a force-to-displacement ratio of 14.5 N/mm, compared with 1.1 N/mm for a commercial electromechanical actuator the researchers used as a reference. The thin films also provide a high surface-to-volume ratio for rapid heat transfer. Under Joule-heated actuation, the integrated device reached a steady 4.0 K temperature span after 20 cycles and a specific cooling power of 4.43 W/g. When driven from the external heat source, those figures fell to 2.2 K and 3.32 W/g, respectively.

This external-heat result is the real proof of concept. While the technology is still an early-stage laboratory experiment, a scaled, perfected version could have interesting implications. Typically, the heat the system needs to operate is generated from electricity or another form of energy. However, an ideal scenario would be to repurpose existing waste heat — a setup already attainable in data centers. Therefore, the technology has the potential to cool processors using the heat they generate!

However, applying it as a data center cooling technology is far from the technology's current state. The prototype produced just 2.09 milliwatts of cooling power at zero temperature lift. That is nowhere near the heat loads of modern processors, much less AI accelerators or data-center racks. The researchers also cite relatively slow actuation, limited strain rate, and the current heat-exchanger geometry among the factors constraining performance. There also needs to be a way to make the heating cyclical.

The team is now working to connect multiple films in parallel to increase cooling capacity. Further progress will require scaling the active material, improving heat transfer and operating frequency, and proving long-term durability. For now, the researchers have demonstrated the underlying energy chain in which heat can be converted into mechanical motion, and that motion can be turned into useful cooling without an electric motor driving the refrigeration cycle.

Micron Taiwan faces a potential strike as support for this labor action is increasing among its workers. Local media reports that a preliminary survey showed 80% of the people involved support such a move, with its labor union demanding that the current bonus system be replaced with a profit-sharing scheme, similar to what SK hynix employees receive. Micron, alongside Samsung and SK hynix, have been raking in a lot of cash because of the AI-driven memory chip shortage, and its workers want a piece of the action.

“Unions say peers offer far higher profit-based bonuses, with Samsung paying out 10.5% of profits and SK Hynix 10%,” Dan Nystedt said on X. “Formal mediation talks are expected to begin by mid-September at the latest, a mandatory requirement before an official strike vote can be held under Taiwan law.”

Earlier this year, Samsung narrowly avoided a strike after it agreed to a deal that gave its foundry workers bonuses that could go as high as $400,000 per employee. It seems that Micron Taiwan’s labor union is following suit, especially as the company’s last earnings call reported a revenue of $41.46 billion and a net income of $28.24 billion, with its Cloud Memory Business Unit, which provides high-bandwidth memory (HBM) to data center customers, accounting for more than 33% of the total revenue.

It’s unclear if Micron workers from other parts of the world will be affected by any deal made in Taiwan, but 60% of its global production capacity is centered around the island, with most of its HBM output coming from its facilities there. With the two sides beginning talks in September, we have yet to see what kind of deal they will reach. But given how lucrative the data center boom has been for the memory industry and the precedent set by workers at Samsung and SK hynix, it’s likely that they will have comparable demands.

While we’re unsure how much bonuses the workers in Taiwan will get, it’s likely that management will negotiate as production disruptions could mean massive losses for the company. Nevertheless, it would also raise complications for workers in other countries, as it could build jealousy and resentment among other workers who will not be part of the bargain, similar to what some Samsung workers are experiencing with the recently closed labor deal.

Intel's confidence in its 14A (1.4nm-class) fabrication process is rising as defect density drops. The company's own design teams are at work developing products that will use the technology, while external customers are now asking about 14A volumes Intel can get them, according to David Zinsner, chief financial officer of Intel, who spoke at Deustche Bank's 2026 Technology Conference. The CFO went as far as saying that 14A is Intel's best process since 22nm technology from the 2010's. But while the comment is optimistic, there is a caveat.

"When you look at the defect density, 14A is tracking better than the target curve we had for 14A," Zinsner said at Deutsche Bank's 2026 Technology Conference. "It is also doing better than any of the previous nodes in terms of how quickly we are bringing down the defects. In fact, we have not seen this performance since 22nm, which is arguably one of the best nodes Intel has ever put out."

Intel intends to begin risk production of its own products on 14A fabrication process in the second half of 2027 and then initiate its high-volume manufacturing in 2028, so 14A is two years away from mass production. So, what Intel's CFO said at the conference is that at this point in 14A's development, its defect reduction trajectory looks at least as healthy as the trajectory of the company's exceptionally successful 22nm process at a comparable point in its development (i.e., in 2010). There are a couple of catches with such phrasing, though. Firstly, the defect density on 14A now is not necessarily equivalent to a defect density on 22nm two years away from mass production. Secondly, due to advances of wafer processing and inspection equipment, what Intel counts as a defect now may not be the same thing as what it counted as a defect 16 years ago. Furthermore, defect density itself does not directly equal product yield.

Intel's 22nm was the company's first manufacturing process to rely on FinFET transistors and at the time was the most advanced process technology in the world; the rest of the industry moved to FinFET devices only with their 14nm and 16nm-class nodes in 2014 and 2015. By contrast, 14A will use second-generation gate-all-around (GAA) RibbonFET transistors, second-generation backside power delivery called PowerDirect, and will be able to use High-NA EUV lithography due to extremely complex patterning. Keeping all of that in mind, Intel's claim that 14A defect density reduction is progressing unusually well this far ahead of HVM is certainly good news.

Meanwhile, comparing 14A to 22nm's successors should be quite comforting for Intel as 14nm mass production was delayed by a year due to insufficient yield, the first-generation 10nm node was a failure, 20A was cancelled, 18A defect density was high even as it hit HVM milestone, while the company did not share almost any information about the progress of its Intel 4 and Intel 3 nodes.

There are good signs for 14A though: external customers are already developing products for this node, whereas the interest from external customers is now practical rather than theoretical.

"We are now seeing demand from our internal customers on 14A [and] they are actually probably the most cynical bunch out of anybody," Zinsner said. "The fact that they are now designing products on 14A was a good confidence boost for us as well. Then, engagements with customers externally, from a foundry perspective has significantly increased. Lip-Bu and the team are now meeting on weekly basis with customers. They are moving away from just looking at data to thinking about 'how much capacity can I get?' 'what does that supply look like?' So, we are now at a point where we have conviction around customers on 14A externally as well."

Intel initiated mass production of its 22nm-based Ivy Bridge processors in late 2011 and early 2012 and released them commercially in late April 2012. The product was highly successful (though overclockers did not like it because of inefficient thermal interface material between the die and integrated heatspreader), and 22nm fabrication technology served the company for many years, first for CPUs in 2012 through 2016, then for other products. Intel's 14A also promises to be a long-lasting node for Intel.

Glass-core substrates, the replacement for organic chip packaging that Intel promised in September 2023 with more than $1 billion behind it, are now in final qualification. However, the product is still not in a single commercial product. SKC, a material manufacturer and chemical affiliate of the SK Group, said on its July 27 earnings call that embedded glass substrate samples from its Absolics plant in Covington, Georgia, are undergoing package-level reliability evaluation in Taiwan, with results possible before year-end.

Samsung Electro-Mechanics formalized a 482.1 billion won ($310 million) glass-core joint venture with Sumitomo Chemical's Dongwoo Fine-Chem on July 2, targeting first production in the second half of 2027. Intel, which started the race, has shifted to licensing its patents and showing demo vehicles, with its own deployment now pointed at around 2030. However, and rather predictably, every timeline in the segment has slipped. Absolics originally planned mass production for the first half of 2024, and reported claims that AMD would adopt glass substrates for CPUs between 2025 and 2026 have come and gone unfulfilled.

Why glass?

The technical case for glass core substrates leans heavily on numbers Intel published a while back, such as 10 times the interconnect density of organic substrates and a 50% reduction in pattern distortion. Glass cores can be tuned to a thermal expansion coefficient of roughly 3 to 10 ppm per degree Celsius against silicon's 2.6, which cuts warpage by about half compared with organic cores, and rectangular panels in the emerging 510mm x 515mm format use more than 75% of their area for large die against roughly 50% for round 300mm wafers. Through-glass vias have been demonstrated at six microns in diameter with aspect ratios beyond 15:1 at ECTC 2025, and Georgia Tech has shown stacked glass running at 220 GHz with 0.3 dB of loss.

Glass also chips and cracks at the edges during drilling and dicing, and MIT Technology Review reported in March that early Absolics production runs broke hundreds of panels every couple of days during early testing days. Edge-coating work has cut measured edge stress from 95 MPa to 49 MPa, and low-temperature dielectrics that cure below 180°C have been developed to reduce thermal stress during build-up, but metallizing vias below 10 microns and holding nanometer-scale flatness across half-meter panels remain open manufacturing problems.

Intel's program

Intel demonstrated a working system booting Windows on a glass-core substrate in early 2025, and Rahul Manepalli, Intel's VP of module engineering, told MIT Technology Review that the benefits of glass cores are "undeniable" and that Intel wants "to be one of the first ones who do it." The commercial plan around that engineering has changed shape, however. DigiTimes reported in late July that Intel is in early-stage talks with Chinese cover-glass maker Lens Technology about a packaging partnership, but there has been no solid agreement made to date.

At NEPCON Japan in January, Intel Foundry showed its first thick-core glass substrate with two EMIB bridge dies embedded directly in the glass: a 78mm x 77mm package with two 800-micron-class glass layers, 10 redistribution layers on each side, and around 1,716 mm2 of silicon on top, roughly two full reticles, with no micro-cracking reported in testing.

There’s currently no production time to this, however, with TrendForce placing Intel’s commercialization somewhere around 2030, alongside co-packaged optics prototypes built on glass at its Rio Rancho, New Mexico site. Meanwhile, Amkor, Intel's packaging partner on the optics work, put commercialization within three years at an industry event in Seoul in April.

Korea aiming for 2027

Samsung Electro-Mechanics moved its glass program from advanced R&D into a business-execution unit in February and has been sampling from a pilot line at its Sejong plant since late 2024. The GLASEM joint venture announced on July 2 splits ownership: 66% to Samsung Electro-Mechanics and 34% to Dongwoo Fine-Chem, with site production in Pyeongtaek, and targets an operating plant in the second half of 2027, with the joint venture making the drilled and metallized glass core that feeds Samsung's substrate line. Korean industry reporting says samples have gone to AMD and Broadcom, and also puts Samsung's overall glass maturity at 40 out of 100, a gap between marketing dates and process readiness worth keeping in mind.

Absolics' $600 million plant in Covington, Georgia, backed by $75 million in CHIPS Act funding plus a further $100 million through the government's advanced packaging R&D program with Georgia Tech, has a Phase 1 capacity of 12,000 m² of substrate per year, enough for roughly two to three million H100-sized packages. The company produced mass-production samples in the first quarter, began customer qualification that reportedly includes AMD and AWS, and is targeting mass production by the end of 2026. LG Innotek runs a third Korean program from its Gumi plant, with prototypes delivered in 2024 and production targeted for 2027 to 2028.

TSMC and Japan

TSMC's CoPoS line in Chiayi, built around 310mm x 310mm rectangular panels, received tools in February, completed its pilot line around June, and is aiming for pilot production in 2027, with mass production in the second half of 2028. Equipment supplier SCHMID has described glass integration in that platform as under review, not committed, and TrendForce puts TSMC's commercial-scale glass-core production after 2030, meaning that the industry's biggest packaging operation is going panel-level first and glass later, if at all. TSMC revived glass substrate research a couple of years ago after earlier deprioritizing it, reportedly under pressure from Nvidia, whose accelerator packages are the main thing outgrowing current packaging.

Japan’s Dai Nippon Printing began phased operation of a TGV glass-core pilot line at its Kuki plant in Saitama in December 2025 on 510mm x 515mm panels, with sample shipments from early 2026 and full mass production targeted for fiscal 2028. Toppan's pilot line for glass cores and interposers at its Ishikawa plant was scheduled for commissioning in July, and Nippon Electric Glass has scaled its ceramic-reinforced GC Core panel to 515mm x 510mm at 1mm thickness. Rapidus is studying panel-level packaging on 600mm x 600mm glass as part of its 2nm program, with viability put at the late 2020s. China has its own entrants, led by display maker BOE, whose pilot line is sampling.

TrendForce estimates Nvidia's Rubin Ultra package at roughly 7,470 mm², about nine reticles' worth of silicon and memory, against around 2,739 mm² for Blackwell, and CoWoS interposer wafers cost on the order of $10,000 each, comparable to a processed 7nm wafer.

Organic substrates warp and lose dimensional stability at those sizes, and round interposer wafers waste a growing share of their area, which is what glass panels are meant to fix. SEMI's first dedicated market report on glass cores, published in May with Global Net, projects initial production around 2028 in select high-performance applications and a 67.2% compound annual growth rate from 2028 to 2040, while Yole puts the advanced IC substrate market at $31 billion by 2030 with glass cores among the drivers.

At the moment, no production design exists, and all customers that have been named or otherwise attached to the tech, including AMD, Broadcom, AWS, and Nvidia, come from wider industry reporting. Nothing has been confirmed officially. As for whether we might see more substantial progress next year, Absolics will need to publish some solid package-level reliability results by the end of this year, or a first officially named customer at any of the Korean manufacturers. We also need to hear from TSMC on whether glass cores will go into CoPoS or remain under review into the 2030s.

The Environmental Protection Agency (EPA) has recently given two new chemicals approval for use within the U.S., but the non-profit organization Earthjustice is suing the agency for not doing its job. According to a lawsuit reported by The Guardian, these chemicals are photoacid generators used for semiconductor manufacturing, but they’re also quite toxic for humans. Exposure to them could allegedly lead to “sudden death,” as well as various health risks like cancer, eye corrosion, neurological damage, and reproductive harm. It is alleged that the approvals are "part of a broader pattern around data centers."

The current administration wants to bring semiconductor manufacturing back into the United States, with President Donald Trump issuing an executive order last year to fast-track the approval of chemicals needed for data centers. It’s unclear if these two chemicals are related to the order, but Atty. Jonathan Kalmuss-Katz, who works with Earthjustice, said that the EPA “does not know the level at which the chemicals are ‘acutely lethal’ or cause other serious health damage.” It also said that “the chemicals may present an ‘unreasonable risk’ to workers and the public.” Nevertheless, the agency allegedly approved them for import and use in the U.S. with minimal and non-protective restrictions.

“If a chemical may present an ‘unreasonable risk,’ then the Toxic Substances Control Act requires the EPA to ‘prohibit or limit the manufacture, processing, distribution in commerce, use, or disposal of such substance or to prohibit or limit any combination of such activities to the extent necessary to protect against an unreasonable risk,” Kalmuss-Katz told The Guardian. He also added, “This is turning the new chemical review process on its head. You have a situation where the EPA has failed at its most fundamental obligation when it comes to new chemicals, and that is to protect the public from unreasonable risk.” Aside from the health and environmental risks posed by the photoacid generators, these chemicals also appear to be PFAS or per- and polyfluoroalkyl substances. These are synthetic chemicals with unusually strong bonds that could last a long time, even if they’re exposed to the environment, which is why they’re often referred to as “forever chemicals.”

The semiconductor industry has long been known for its use of PFAS, and researchers are working hard to find ways to help clean up these chemicals used to make chips. So, the approval of new chemicals without rules and regulations that require them to be removed from discharges is potentially disastrous for the environment. This is the same concern that a lawsuit against Micron’s planned New York fab raises, which says that the approved wastewater and air permits that the company received could still allow these “forever chemicals” to leak into the Oneida River. Environmental groups acknowledge that domestic chip production requires the use of these potentially harmful chemicals, but the EPA must ensure that these do not find their way into the water system and that the people working with them are protected from exposure.

Groq's former chief architect stood on stage at Hot Chips 2026 and presented his former company's inference chip as Nvidia silicon. Igor Arsovski, now Nvidia's VP of hardware, presented the Groq 3 LPX rack's architecture and published the first third-party benchmark of the hardware: Artificial Analysis measured it at 3,431 output tokens per second on a 100K-context Gemma 4 31B reasoning workload, roughly four times the 870 tokens per second of the next-fastest public endpoint. Arsovski said the rack is already in production, built on the LP30 chip Nvidia obtained through its $20 billion Groq deal in December 2025, the same deal that pushed the Rubin CPX it replaced off Nvidia's roadmap.

SRAM without HBM

Artificial Analysis ran the comparison on a private, pre-release Gemma 4 31B endpoint served through Google Cloud, taking the median of 50 sequential client requests at a concurrency of one, while the public providers it measured against ran shared production serverless endpoints. Serving one request at a time produces the highest per-user token rate the hardware can post, and it's not directly comparable to the multi-tenant conditions the other endpoints run under.

Nvidia's on-stage demo showed a higher figure still, 10,996 tokens per second on the same 31B model, which Igor Arsovski, Nvidia's VP of hardware, flagged on stage as "self-reported" before telling the audience the aim was "third-party verified independent benchmarks that you guys can trust." Gemma 4 31B is also a dense model small enough to sit inside a single LPX rack, and the picture at trillion-parameter mixture-of-experts scale, where memory capacity becomes the main constraint, went unaddressed.

Each LP30 carries roughly 500MB of on-die SRAM and no HBM, so a full LPX rack of 256 chips holds 128GB of memory delivering 40 PB/s of aggregate bandwidth against 315 PFLOPS of FP8 compute, with 350 ns of chip-to-chip latency in a Vera Rubin-compatible, MGX liquid-cooled rack that scales past 1,000 LPUs.

Keeping model weights resident in SRAM rather than streaming them from HBM removes the memory-access latency that dominates single-token decode, and the design drops caches, branch prediction, and out-of-order execution in favor of a fully deterministic pipeline that the compiler schedules at clock-cycle granularity. The architecture descends directly from the Tensor Streaming Processor that Groq, founded by ex-Google TPU engineer Jonathan Ross, described in a 2020 ISCA paper titled Think Fast, the same title Arsovski and Raghavan reused at Hot Chips.

A Rubin GPU carries 288GB of HBM4, roughly 576 times the memory of a single LP30, so a 31-billion-parameter model at FP8 needs on the order of 62 LPUs to hold its weights, and a large mixture-of-experts model runs into four figures of chips across several racks. Capacity is the cost of the SRAM-only design, and it's why Nvidia is describing the LPU as for decode rather than as a general-purpose replacement for its GPUs.

Determinism lets the compiler predict power draw cycle by cycle, which Nvidia uses to pre-order current from the rack's regulators ahead of demand, cutting voltage droop by more than 60% and overshoot by more than 70% against an uncompensated load. The same per-block scheduling lets the hardware equalize heat instead of throttling to the hottest tile, which Arsovski put at roughly 10% to 11% additional performance under a fixed thermal limit. "By doing this, we can actually get more utilization of the chip under the same thermal limit, basically. So we can actually get, again, about 10 to 11% more performance under the same thermal limit. So this is another benefit of deterministic execution."

Across racks, Nvidia synchronizes chips to a single virtual clock in what it calls a plesiosynchronous network, with each chip acting as both processor and router so the fabric needs no adaptive routing or congestion sensing, and clock drift between chips is compensated at the chip-to-chip links. Asked during Q&A about the blast radius of a chip that fails mid-workload, Arsovski said users "would experience the exact same as any other hardware in the industry" and would "just checkpoint it or reconfigure the hardware."

Splitting inference with Rubin

Nvidia is pitching the LPX rack as a decode co-processor bolted onto Vera Rubin NVL72, with Rubin GPUs handling the compute-heavy prefill phase and building the KV cache while the LPUs generate output tokens. Nvidia showed three ways to divide the work: disaggregated prefill and decode; attention-FFN disaggregation, which keeps attention and its cache on GPU HBM while the LPU runs the feed-forward layers; and external-draft speculative decoding, where a small model on the LPU proposes tokens that the GPU verifies in parallel, with only draft tokens crossing the link.

An FPGA bridges the synchronous LPU domain and the asynchronous world of host I/O and GPU hand-offs, and Nvidia's Dynamo runtime, together with an LPU extension to CUDA, orchestrates the split. The company put the gains from these modes at roughly three-to-five-times over Rubin alone on a two-trillion-parameter workload with a 400K-token cached context, all Nvidia-measured.

Cerebras CS4

Cerebras used the same Hot Chips session to present its CS4 wafer-scale system, which chief system architect Jean-Philippe Fricker said runs up to 30 times faster than GPUs and doubles the token rate of the CS3 while carrying 10 times the token capacity. Each CS4 rack packs three wafer-scale engines into a new modular platform Cerebras calls Nexus, built around pluggable compute "backpacks" that separate power, compute, and I/O, and Fricker put its memory bandwidth at 43 PB/s, which he told the audience was "2,000 times higher memory bandwidth than Nvidia's next-generation Rubin chip." Cerebras also has a partner for the prefill side of the same problem: it agreed in July to pair AMD Helios GPUs for prefill with its wafer-scale engines for decode, the same division of labor Nvidia now builds in-house with Groq.

Nvidia pulled the Rubin CPX, its own GDDR7-based long-context accelerator, to focus on shipping the LPU this year, a decision VP Ian Buck laid out at GTC 2026. The $20 billion deal that produced the LP30 was structured as a non-exclusive IP license plus the hiring of Ross, president Sunny Madra, and most of Groq's engineers, a form that avoided a formal merger review. Arsovski opened the Hot Chips talk by calling it "a pinch me moment for the Groq team that's now integrated into the Nvidia group."

Senators Elizabeth Warren and Richard Blumenthal wrote to the FTC and to Nvidia in early 2026, arguing the arrangement acquired Groq "in all but name," and no formal, deal-specific investigation has been confirmed as of late August.

Full Nvidia Groq Hot Chips 2026 presentation

d-Matrix presented Raptor, which it calls the first 3D DRAM accelerator for generative inference, at Hot Chips 2026 this week, showing a TSMC 4nm compute die bonded face-to-face at a 36-micron pitch on top of a custom-designed DRAM die that delivers 100 TB/s of bandwidth from 32GB per card.

Co-founder and CTO Sudeep Bhoja put the vertical interface's energy cost at 0.37 pJ/bit against roughly 2.4 pJ/bit for moving data into an HBM4 base die, calling it "a measured number" from working silicon, and the accompanying ISCA 2026 paper, written with the University of British Columbia, projects around 4.7 times higher throughput per card than HBM-based designs. Bhoja, however, didn't disclose who manufactures the DRAM die.

CEO Sid Sheth told CNBC in June that Raptor is slated to launch in 2027, but at Hot Chips the company gave no firm date or information on volume and pricing, and every performance figure shown, including 988 tokens per second per user on the 2.8-trillion-parameter Kimi K3 model at 1M-token context, is a d-Matrix projection built on early silicon.

The custom DRAM die

Raptor inverts the usual 3D stacking arrangement by putting the logic die on top and the DRAM underneath, so a cold plate sits directly on the compute silicon and the DRAM die doubles as the interposer, carrying PCIe and die-to-die signals down through its TSVs. "For the same amount of power, you can drive the bandwidth up," Bhoja said during the session, "and so we were able to drive the bandwidth up here to 100 terabytes per second."

The one-high stack achieves a power density of roughly 0.5W per square millimeter, which liquid cooling can handle, but the DRAM is designed for a junction temperature of 105°C, where retention collapses from a standard 32ms to 4ms, and the memory must refresh eight times more often. d-Matrix absorbed that penalty by shrinking each microbank to 1,366 rows and about 5.33MB, so a full refresh sweep costs only 1.37% of overall bandwidth.

The die carries 840 banks per chiplet, of which 72 (around 9%) are spares wired into a two-level mux chain the company calls bank chaining, letting any two failed banks anywhere on the die be switched out while channels stay symmetric. A [132,128] Reed-Solomon code on the logic die corrects two symbol errors per 128 bytes, with a CRC behind it. The interface has no PHY, no burst structure, and no sideband pins, so conventional data-bus inversion was impossible; d-Matrix instead compares each 128-byte flit to the previous one and stores a 1-bit inversion tag alongside the ECC metadata, recovering roughly 20% of the I/O power DBI would have saved. At full tilt, the vertical interface still burns 296W of the 422W per-package budget the ISCA paper discloses.

The bank geometry, spare-bank mux tree, refresh behavior, and interleaved ECC columns were all co-designed with the compute die. The 0.37 pJ/bit figure exists only because of that pairing, and no memory maker has anything like this die in its catalog.

Who's supplying it?

d-Matrix has named TSMC for the N4P logic die and Alchip as its ASIC design and 2.5D/3D packaging partner, but neither company operates a DRAM fab, and across the Hot Chips talk, the ISCA paper, and every public announcement since the Pavehawk 3DIMC test silicon came online last September, the firm has never identified who fabricates its custom DRAM. Only three companies make leading-edge DRAM at volume, and all three are allocating capacity to HBM4 lines that are effectively sold out through 2026.

J.P. Morgan estimates DRAM prices will have risen more than 400% between the start of 2024 and the end of 2026. In addition, analysts have recorded contract price increases of 90% to 95% in Q1 2026 alone, and SK hynix CEO Kwak Noh-jung told Reuters in July that "customer demand will remain higher than our supply capacity even beyond 2030."

Nvidia, the memory makers' largest and most leveraged customer, is reportedly testing Rubin Ultra configurations with as little as 192GB because it may not be able to source enough HBM4E. A startup with roughly $450 million raised, asking a memory maker to run a bespoke die with non-standard bank geometry on capacity that could otherwise print HBM, is negotiating from a far weaker position than that, and until the supplier is named, Raptor's 2027 volume plan rests entirely on an unknown, undisclosed dependency.

The die's 11.4 MB/mm² density is roughly half of HBM4's 21.9 to 26.3 MB/mm², Bhoja acknowledged during the Q&A: "A lot of the drop for us was also because we used a not-so-advanced DRAM. And so if we used a more mainline DRAM, just like the HBM4 guys are doing, we would be able to push that up almost all the way to the HBM4 numbers." The density penalty therefore tracks back to whatever foundry arrangement d-Matrix currently has.

32GB per card

Raptor's 32GB per card stands against 192GB to 288GB for HBM4-equipped accelerators, so d-Matrix sizes deployments at rack scale instead: 72 cards carry 2.3TB, enough to hold Kimi K3's weights at 4-bit precision with headroom for around 54 concurrent users at 1M context, by its own calculations. "Even with 32 gigabytes of memory capacity, we are able to solve SOTA models in a scale-up network, so no bits are wasted," Bhoja said.

KV cache growth works against that arithmetic over time, and co-presenter Aayush Ankit, who led Raptor's SoC architecture at d-Matrix before joining Meta's MTIA team, explained the failure mode while dismissing SRAM alternatives: "We are making this unit of compute blazingly fast. Communication becomes a bottleneck soon enough." Once models and context no longer fit a single rack, inference spills into the inter-card synchronization overhead that the vertical bandwidth was meant to eliminate, which the ISCA paper concedes.

Cerebras claimed 969 tokens per second on Llama 3.1 405B in November 2024, with third-party firm Artificial Analysis verifying the figure on live hardware, and that remains the closest published reference point for Raptor's numbers. d-Matrix's 988 tokens per second per user on a model seven times larger would be a step change if it holds water, but no third party has measured Raptor, and the comparison points in the ISCA paper are simulations anchored to early silicon characterization.

Asked by an Nvidia employee about multi-layer stacking plans, Bhoja said: "Our roadmap is still a work in progress, and we have a hard enough time trying to get one-high to work and work around all of the thermal issues of that." Samsung brought its own version of the idea to Hot Chips with zHBM, a concept that stacks HBM directly on the processor and carries a claimed 70% power-efficiency gain over an HBM4E setup, with no production timeline before HBM5. The problem for d-Matrix is that Samsung runs its own DRAM fabs; d-Matrix doesn't. Whether it can get wafers at volume remains to be seen.

Full d-Matrix Hot Chips 2026 presentation

Just over one week after Nvidia agreed to backstop up to $105 billion in financing for its data centers, OpenAI arrived at Hot Chips on Tuesday with benchmarks claiming its first in-house chip beats Nvidia's GB300. Jalapeño, the inference ASIC OpenAI co-developed with Broadcom, delivered 1.5 times to 1.9 times more throughput per kilowatt and 1.7 times to 3.6 times lower end-to-end latency than Nvidia's GB200 and GB300 rack systems on SemiAnalysis's public InferenceX suite, with a 700W part going up against accelerators rated at 1,200W and 1,400W. OpenAI plans to begin deploying the chip in its own data centers later this year.

The tests covered three open models: GPT-OSS 120B, DeepSeek R1 670B, and Moonshot AI's 1-trillion-parameter Kimi K2.5, with OpenAI reporting its widest leads at low-latency operating points, where it claims 8.6 times to 104.3 times more throughput per kilowatt at the GB300's fastest previous time-between-tokens settings.

OpenAI normalized the results to each accelerator's published package TDP, though it said Jalapeño's measured sustained power stayed at or below 550W in testing. An appendix comparison using all-in utility power per accelerator, 1.18kW for Jalapeño against 2.55kW for the GB300, produces narrower gaps, as does pitting Jalapeño against a GB300 running multi-token prediction, where the peak efficiency lead shrinks to roughly 1.5 times.

Jalapeño wasn't tested against Vera Rubin, the Nvidia platform that's slated to power the first gigawatt of Nvidia systems OpenAI agreed to deploy in the second half of 2026. The chip also doesn't train models, the workload where Nvidia's hardware remains unchallenged. In addition, the major comparisons ran Jalapeño's single-token prediction against GB300 configurations doing the same, even though Nvidia deployments commonly use multi-token prediction in production. SemiAnalysis, which said it ran InferenceX with OpenAI engineers in the company's lab, described the part as "beating every Nvidia, AMD, and Google chip we have been able to test."

Each Jalapeño package, unveiled in June after a nine-month RTL-to-tapeout cycle, pairs its compute die with six HBM4 stacks, totaling 216 GiB at 15.4 TB/s. The GB300 carries 288GB of HBM3E at a 1,400W rating, so per watt of rated power, OpenAI's chip packs roughly 50% more memory. The company's Hot Chips presentation states that the main bottleneck its architecture targets is exposing aggregate HBM bandwidth, not adding more of it.

That memory is of course the tightest commodity in the semiconductor industry. Samsung, SK hynix, and Micron have sold their HBM capacity through 2027, a shortage so severe that Nvidia is reportedly testing cut-down Rubin Ultra configurations with as little as 192GB, and SK hynix CEO Kwak Noh-jung has warned that 2027 will be the worst year of the crunch. Micron told the same Hot Chips conference on August 23 that HBM consumes roughly three times the wafer area of DDR5 for equivalent capacity, a penalty that widens with each generation. Scaling Jalapeño across the 10GW deployment agreement that OpenAI signed with Broadcom last October would make the company a substantial new claimant to HBM4 supply, which Nvidia currently dominates through multi-year allocation deals with SK hynix.

A second-generation chip is approaching tapeout, expected within months, according to Bloomberg, and concept work on a third generation is underway. The first part reportedly uses a TSMC 3nm-class process, keeping OpenAI in the same wafer, memory, and advanced packaging queues as Blackwell and Rubin for the foreseeable future.

OpenAI is procuring those inputs while deepening its financial dependence on the company it just benchmarked. On August 17, Nvidia agreed to provide up to $105 billion in financing for an OpenAI-leased data center campus in Ohio. "Nvidia is a really good partner, and we continue to need a lot of Nvidia," Richard Ho, OpenAI's vice president of hardware, told Bloomberg in an interview following the announcement.

Raghu Sreeramaneni, Micron HBM Design Architecture Fellow, told the Hot Chips 2026 conference on August 23 that the silicon penalty HBM carries against DDR5 is growing with every generation, and that it can’t be closed without giving up performance, something that Micron won’t do. Pressed on whether newer parts narrow the gap, he said it’s "definitely not getting better," putting the current overhead at roughly three times the wafer area of DDR5 for the same capacity. That trajectory is the reason why PC memory hit record prices this year, with conventional DRAM contract prices up 90% to 95% quarter over quarter in the first quarter of 2026.

Wafer mathematics

An HBM4 die fits 256 banks against 32 on a DDR5 die, and it reaches its bandwidth by running those banks in parallel, which requires far more die area for data paths, power delivery, and the through-silicon vias (TSVs) that link each layer to the base die. A single HBM3E die can feed 256 GB/s, while one DDR5 die supplies about 8 GB/s, so matching a given bit count makes the HBM die much larger.

Micron has publicly put that overhead at about three times the wafer area of DDR5, and Sreeramaneni said it traces to an HBM3E-against-DDR5 comparison that grows with each generation as pin speeds, bank counts, stack heights, and die sizes climb higher. He called HBM "probably the most cross-functionally complex solution that we make," and noted that in a two-GPU package, memory accounts for around 90% of the silicon, roughly eight times the area of the GPU dies. HBM already sells for about five times the price of DDR5 per bit, so each wafer a maker moves to it removes a disproportionate share of commodity memory from the market, pushing consumer product prices higher as a consequence.

Conventional DRAM contract prices rose 90% to 95% quarter over quarter in the first quarter of 2026 and a further 58% to 63% in the second. To put that into perspective, a mainstream 32GB DDR5-6000 kit sold for around $392 this month, up from $110 to $140 a year earlier, and 128GB of DDR5 passed $3,399, a 500% rise over 12 months, with European prices up 345% since September last year. Back in February, HP told investors that DRAM now makes up 35% of its PC build cost, up from 15% to 18% a quarter earlier, and Gartner expects PC shipments to fall more than 10% in 2026. While it’s true that the surge eased to 13% to 18% in the third quarter, that happened because consumer electronics makers reached the ceiling of what they could pass on, not because supply improved.

Bandwidth, stacking, and heat

Micron's presentation put compute performance scaling at roughly three times every two years and HBM bandwidth at under two times, a divergence Sreeramaneni summarized by saying "the memory wall is still present, and, in fact, maybe getting worse." HBM4 doubles the host interface to 2,048 I/Os, and Micron's parts run above 11 Gb/s per pin for more than 2.8 TB/s per stack, beyond the 8 Gb/s and 2 TB/s baseline in JEDEC's HBM4 standard.

Each of those gains comes from more parallelism, which enlarges the die again and widens the silicon ratio. Sreeramaneni said there is "a good path to 16 layers of DRAM," but "after 16, there's still a lot of work to be done." Meanwhile, SK hynix used its own Hot Chips talk to describe a 775-micron total-thickness ceiling that caps stacking until the industry adopts hybrid bonding, which SK hynix doesn't expect before HBM5. Micron's HBM4E, due around 2027, moves the logic base die to a TSMC foundry process, part of a wider shift toward customization as AI workloads fragment, which Sreeramaneni described by saying "disaggregation is the buzzword now."

The base die runs hottest because it sits at the bottom of the stack, doing the highest-speed interface work while the heatsink sits at the top. Every layer added between the base and the heatsink then raises thermal resistance. Sreeramaneni said Micron is now "architecting solutions around thermals rather than the other way around," a reversal from a few years ago. Micron highlighted how Meta's Llama 3 paper attributed 17.2% of unexpected interruptions during a 54-day run across 16,384 H100 GPUs to HBM3 memory, the second-largest cause behind failed GPUs, a reliability burden that scales with the amount of HBM in each package.

Share and supply outlook

SK hynix leads the HBM market with a share that Counterpoint Research places at around 58%, down from a high of around 69%, and Micron overtook Samsung for second place last year. Micron began volume shipments of 36GB 12-high HBM4 earlier this year for Nvidia's Vera Rubin platform, with Nvidia CEO Jensen Huang saying in June that all three suppliers are qualified and in production.

Analysts expect DDR5 per-wafer profitability to overtake HBM3E by the end of this year, leaving the three makers little reason to add commodity capacity when server DRAM pays nearly as well per wafer. SK hynix said last October that it had already sold its entire 2026 output, while Apacer's chief executive told investors that supply to independent module makers could fall to 30% of 2026 volumes next year. SK hynix CEO Kwak Noh-jung has called 2027 the worst year for memory supply in the industry's history, with demand outrunning production into 2030.

In response, Nvidia has raised its DGX Spark desktop from $3,999 to $4,699 and warned large customers of AI server price rises above 15%, both tied to memory. Realistically, the squeeze will loosen only under three conditions, none of them likely before new fab capacity comes online and scales up: China's CXMT ramping competitive DDR5 in volume, hybrid bonding arriving before HBM5, which SK hynix has already all but ruled out, or DDR5 profitability slipping back below HBM.

Full Micron Hot Chips 2026 presentation

SK hynix doesn't expect hybrid bonding to be ready for HBM4E, Jaesik Lee, VP of package engineering at SK hynix America, said during a presentation at Hot Chips 2026 on August 23, pushing the industry's most anticipated memory packaging transition out to HBM5 at the earliest.

The problem, as he describes it, is that HBM cubes are capped at a total thickness of 775 microns — the standard thickness of a 300mm logic wafer — so every additional DRAM layer must come from thinner dies and narrower gaps. 16-Hi HBM4, now in customer qualification at 48GB per cube while 12-Hi is in mass production, thins its core dies to around 50 microns and halves the gap between them compared with 12-Hi. Lee's session also went into detail about the company's iHBM cooling architecture three months after its May unveiling. Attaching a constraint to it, Lee explains that the heat blocks can't be applied to any HBM generation already in design.

The 775-micron limit

The JEDEC HBM4 standard raised the package thickness ceiling from 720 microns, which held through HBM3E, to 775 microns, easing the pressures associated with adopting hybrid bonding. When a GPU package gets its cold plate attached, both the logic die and the memory stacks are ground back to expose bare silicon, Lee said, and because logic wafers are 775 microns thick, a memory cube that grew any taller would stand proud of the processor beside it. "That's the kind of limit that we can go up so far, because the logic wafer thickness is also 775 microns," Lee said.

Thinner dies leave the stack with proportionally more oxide, which conducts heat poorly compared with silicon, while pin speeds that have risen from 1 Gbps in early HBM to 8 Gbps in HBM4 concentrate more power in the same footprint. SK hynix's own figures put the thermal burden at 2.2 times higher across the HBM generations shown, while stack counts double every two generations.

The company's mass reflow-molded underfill (MR-MUF) process, which stacks all dies via pick-and-place and joins them in a single reflow, already trades away margin here: filling gaps that have shrunk by half while controlling warpage on sub-50-micron dies is, per Lee, the main manufacturing challenge of 16-Hi.

Hybrid bonding keeps slipping

Samsung publicly committed to hybrid bonding for HBM4 in May last year, with SK hynix holding the copper-to-copper technique as a backup behind advanced MR-MUF. The JEDEC thickness relaxation then removed the immediate need, and industry discussions now weigh a further move to 825 to 900 microns for 20-Hi stacks, which would push the copper-bonding crossover out again.

Back in March, it was claimed by industry sources that SK hynix placed its first mass-production hybrid bonding order, a single inline system pairing Applied Materials and Besi tools worth around 20 billion won ($15 million), and Counterpoint Research expects the technique to enter full-scale HBM production with HBM5 around 2029 to 2030.

Hybrid bonding remains at the research stage for stacks of 20 layers and above, per the deck's roadmap, and SK hynix is still deciding which product gets it first. Lee didn't name a target generation, but ruling out HBM4E leaves HBM5 as the earliest slot. The technique joins flattened copper pads and oxide surfaces at room temperature, then relies on copper's thermal expansion during a cure step to form the bond.

"This is a very simple process, but in reality it's really challenging," Lee said. "We are talking about 16 layers and 20 layers that we need to make the hybrid bonding, so it's very different from the one-layer stacking." Removing micro-bumps entirely lets core dies grow up to 24% thicker at 20-Hi, cuts thermal resistance by roughly 35% versus MR-MUF at that height, and takes bump pitch below 18 microns, per the deck, against the 30 microns where MR-MUF is today. At HBM4's bump pitch, conventional micro-bumps still work, and each time JEDEC has raised the thickness ceiling, MR-MUF has stayed viable for another generation.

iHBM

The iHBM concept embeds thermally conductive, electrically insulating blocks into the base die's die-to-die PHY region, the interface hotspot where power density peaks, for a claimed thermal resistance reduction of more than 30%.

Lee's slides benchmarked it directly against Samsung's Heat Path Block approach, which routes heat out of the stack through dedicated pillars, and Micron's base-die circuit redesign, which claims over 20% better energy efficiency. All three are vendor claims measured on different metrics, and the SK hynix and Samsung designs are both slated for HBM5, with neither expected in mass production before 2028. Because the blocks sit inside the package alongside the D2D PHY, they need optimization with the customer's design and can't be applied to generations already in design, Lee said, which makes iHBM a co-design effort. "It's a kind of good option that we can do, but this is not something that we can apply [to] the generation that we already [have] in design."

During Q&A, Tanj Bennett of SemiAnalysis argued that stacking taller dilutes the silicon's own throughput. DRAM operating at the cell level delivers on the order of 20 TB/s per square centimeter; a 20-Hi stack tops out around 4 TB/s, and HBM takes far more manufacturing capacity than equivalent DDR5 or LPDDR. "As you get to 20 high, the average speed of that memory is slower than DDR5," Bennett said. "Why is it better to be using the height of the HBM stack instead of intelligently placing cheaper memory around it?"

Lee answered that training workloads demand both bandwidth and capacity, then added that inference may split the difference, keeping the KV cache in high-bandwidth memory while offloading to LPDDR. That split already exists in products like Nvidia's Vera Rubin platform, which pools LPDDR5X with HBM4 over NVLink-C2C for exactly this purpose, and the High Bandwidth Flash spec that SK hynix co-developed with Sandisk extends the tiering idea to NAND.

"That's a kind of question that we need to also look at in the future," Lee said of the tiered approach. SK hynix holds around 70% of Nvidia's HBM orders for the Vera Rubin generation, according to reporting from January, and all of it will be stacked with MR-MUF. Which product moves off it first, Lee says, is a decision that the company hasn't made.

Full SK hynix Hot Chips 2026 presentation

A research team at Kyoto University has built a silicon carbide (SiC) transistor that operates at 600°C (873 K) using ion implantation, the doping step used across commercial chip fabs, and cut the gap between its designed and measured threshold voltage to under 0.1V at 400°C, down from more than 2V in the conventional layout. The work, announced earlier this month by Mitsuaki Kaneko, Shunya Shibata, and Tsunenobu Kimoto and published in APL Electronic Devices, pairs a bottom-gate transistor with a double-well isolation structure to keep an ion-implanted SiC junction field-effect transistor stable at temperatures where silicon stops working above roughly 250°C.

The bottom-gate layout puts the gate electrode beneath the channel, which captures the dopants that scatter deeper than intended during implantation, a channeling effect that had thrown conventional top-gate JFET thresholds off by more than 2V. With the channeling tail compensated, the design-to-measured threshold voltage gap dropped to under 0.1V at 400°C.

The double-well structure isolates each device inside a pn junction instead of leaning on the semi-insulating SiC substrate underneath, which loses its insulating behavior as it heats and bleeds leakage current through the wafer. Kaneko's group reported the remaining leakage sits close to the theoretical floor set by SiC's own material properties, so there's little room left to improve it at the device level.

In contrast, NASA Glenn Research Center has run SiC JFET integrated circuits carrying more than 175 transistors for over a year at 500°C in air, and 60 days on a simulated Venus surface at 460°C and 9.3 MPa with no shielding, per the agency's published testing. Those chips are built from epitaxial JFET-resistor devices on a bespoke process. The Kyoto researchers reached a higher device temperature through ion implantation, a method already standard across mainstream fabs, which fits existing mass production.

SiC is already scaling as a power-device wafer material, but high-temperature logic is a separate and far smaller niche. It isn't the only wide-bandgap material chased for these conditions either, with researchers recently showing devices that hold up from 500°C down toward absolute zero in other compounds.

The transistor is normally-on, so it conducts with no gate voltage applied and draws standby power. Efficient logic needs complementary pairs built from normally-off devices, which this isn't. Kimoto's group has already demonstrated complementary SiC JFET logic gates at 350°C, and its next step is to design normally-off devices in the new structure to build low-power complementary circuits. Long-duration reliability at temperature and heat-tolerant packaging are just two challenges the researchers will need to overcome before any of it reaches a gas turbine, or Venus.

As advanced packaging technologies like EMIB and CoWoS become yet another battlefield in the semiconductor industry, outsourced semiconductor assembly and test (OSAT) companies are trying new tools in a bid to offer services that others do not. This week, LG Electronics Production Technology Institute (PRI) signed a contract with an OSAT to supply it with a maskless laser direct imaging (LDI) lithography tool that can be used to build metal-interconnect patterns in semiconductor packaging, potentially with higher yields than currently available tools, reports ETNews.

LG-PRI's Laser Direct Imaging (LDI) system is a maskless lithography machine designed to pattern fine metal interconnects for advanced semiconductor packaging. While the company has developed various versions of the machine, its highest-resolution version can produce 1.5-µm line-and-space (L/S) patterns, which should be fine to 'print' wiring pitches of about 3 µm. The production equipment uses a 405-nm laser-diode light source and can process substrates as large as 600 × 600 mm, according to various media reports.

LG positions its LDI system primarily for advanced semiconductor packaging using organic and then glass substrates, displays, and MEMS. Yet, it can also be used to build high-density printed circuit boards (PCBs) for mobile devices or prototype purposes.

LG is entering an already established direct-imaging market led by KLA, Screen Holdings, Limata, and ORC, but participated in by a dozen manufacturers from Germany, France, Switzerland, Japan, and even China. What is notable is that LG is offering LDI systems with line/space capability down to 1.5-µm, which means it is targeting the higher end of the market. Yet, to establish a position, LG plans to price its LDI system competitively.

Another way to pattern substrates

There are many ways to pattern substrates, including photolithography, e-beam lithography, nanoimprint lithography, and laser direct imaging, just to name a few. LDI is a maskless lithography process in which a laser exposes a digitally generated circuit pattern directly onto a photoresist-coated substrate, rather than transferring the pattern through a physical photomask. Instead of a photomask, an LDI system uses a digitally controlled pattern generator and projection optics, which resembles how a laser cinema projector projects a digital image onto a screen, except that LDI projects a circuit pattern onto photoresist. After development, the exposed resist leaves a pattern that defines where metal interconnects will be formed in later processing steps.

LG's LDI machine can produce down to 1.5-µm line-and-space patterns, which is good enough for chip substrates and even redistribution layers (TSMC CoWoS-R/-L uses RDL interposers with a minimum 4-µm pitch, or 2-µm line width/spacing). In contrast, competing products offer 1-µm, 3-µm, and 5-µm versions for different applications. While LDI in general cannot boast the resolutions offered by modern DUV, EUV, or e-beam lithography machines, LDI trades ultimate resolution for vastly higher throughput and large-area processing, exactly what the doctor ordered for production of PCBs or chip packaging. With current-generation LDI, the RDL interposer is the most advanced thing that these devices can pattern, as both CoWoS-S and CoWoS-L/EMIB-like technologies require considerably higher resolution. Meanwhile, there is another advantage that LDI machines have over lithography systems that rely on photomasks.

Because the image is generated digitally, the system can create and calibrate patterns in real time, while projection optics, precision alignment, and stage control position the pattern accurately on the substrate. This capability is important because the dimensions/geometry of packaging substrates can vary and because organic substrates can expand, contract, or warp during processing. Given that we are talking about a 1.5-µm-scale wiring pattern, any shift from the nominal layout may result in a solder bridge or a faulty contact, which means yield loss. That said, the ability to adjust the pattern to the peculiarities of a substrate may be a game-changer for OSATs, their customers, and a major selling point for devices that can do it.

A new kid on the block? Not really

Unlike competing chaebols Samsung and SK Group, LG Group does not produce chips, even though its divisions supply various materials and components for the semiconductor industry. To that end, it is perfectly reasonable for the company to enter the market for tools for the production of chip packages or PCBs.

In fact, LG PRI is not entering the exposure-equipment business from scratch. PRI traces its history to the Goldstar Production Technology Research Institute, established in 1987, and has worked on manufacturing and productivity technologies covering semiconductors, displays, and rechargeable batteries. Specifically, LG had already commercialized LDI technology for display manufacturing and supplied such equipment to LG Display.

As a result, the semiconductor packaging-grade equipment represents an expansion of an existing LG technology into a new market rather than the development of an entirely new exposure platform from scratch and without any experience. Interestingly, LDI is the first element of LG's semiconductor equipment ambitions. The company reportedly plans to expand its portfolio into high-bandwidth memory (HBM) inspection equipment as well as through-glass-via (TGV) laser systems for glass substrates.

Prospects

Before LG expands to inspection or laser drilling tools, it will have to establish itself as a producer of reliable tools used for chip packaging, which will likely take years. For now, the significance of the deal with the undisclosed OSAT is less about business and more about the fact that an OSAT decided to give LG's LDI machine a try. Moving from university R&D installations to an OSAT mass-production facility provides LG a starting point for pursuing additional external orders.

Whether LG can establish itself alongside Applied Materials, ORC, or Screen will depend on how its equipment performs in production and whether its emphasis on competitive pricing proves sufficient to persuade more packaging companies to adopt the system.

In any case, a new Wafer Fab Equipment player is here, which is good news considering shortages of virtually all chipmaking tools. While LG's entrance will hardly have any noticeable impact on the market for at least a couple of years, another supplier could eventually add much-needed capacity, increase competition, and give chipmakers and OSATs another source of advanced packaging equipment.

Nvidia has rejected a report claiming that it plans to begin small-batch shipments of a language processing unit tailored for Chinese customers by the end of 2026, with several Chinese orders already placed. "The reporting in The Information on NVIDIA's LPU is incorrect. We have no LPU sales in the China market today, and no China-specific LPU product in our roadmap," an Nvidia spokesperson told Tom's Hardware on Thursday. The Information's story, which cited two Nvidia employees, said the chip is a variant of the Groq 3 LPU Nvidia announced at GTC in March, and that its silicon is unchanged because it already falls within U.S. export rules.

The LPU was designed as a decode co-processor for the Vera Rubin platform, and Vera Rubin can't be sold in China. The Information's sources said Nvidia rewrote the software that splits work between the GPU and the LPU so the accelerator can run alongside processors that are available in the country.

The publication said Nvidia didn't respond to requests for comment over several days before publishing, and that it's unclear whether Beijing would allow the orders to proceed. Chinese officials blocked purchases of the H20 last year and only recently told companies they'd permit some H200 imports, so U.S. compliance alone doesn't guarantee the chips can be delivered.

Back in March, it was reported that Nvidia was preparing LPUs for China, with Jensen Huang saying two days later that the story was "totally false.” Thursday's statement is narrower than Huang's, addressing current sales and a China-specific product. Nvidia hasn’t clarified whether the standard LPU will ship to Chinese buyers. Huang told CNBC in May that Nvidia had "largely conceded" China's AI chip market to Huawei.

The Groq 3 LPU is built on Samsung's 4nm process with 512MB of SRAM per die and no HBM, and Nvidia said at GTC that it would ship in Q3 2026 to customers including OpenAI. U.S. export thresholds for China are set on compute density and bandwidth, and an SRAM-only decode accelerator with no HBM stack is the kind of part that can still be exported under them without a cut-down SKU, which is the mechanism The Information's sources described.

Huawei's Ascend 950DT, which the outlet named as the LPU's direct competitor, is optimized for decode and training and is due in Q4 2026, with the prefill-focused 950PR already in production since April. ByteDance and Tencent each took delivery of roughly 10,000 H200s in recent weeks, according to a Financial Times report this week, the first meaningful Nvidia accelerator volume to enter mainland China since December's U.S. approval.

Synopsys has published silicon results for what it calls the first 3D PCIe 6.0 test chip, a 5nm PHY built into a face-to-face stacked package that runs 64 GT/s per lane and up to 128 GB/s across an eight-lane link using PAM4 signaling, with receiver eyes clearing the standard's bit error rate requirement. The company says it got there by pulling apart an existing 2D PCIe 6.0 test chip, adding through-silicon vias, and redoing circuit design and signoff against 3D process design kits, according to its blog post.

In monolithic chips, PCIe PHYs sit at the perimeter of the die, right next to the package I/O connections, helping keep traces to the substrate short so attenuation and reflections remain manageable. A 2.5D package preserves that layout by parking the PHYs along the outer edge of the outermost chiplets. Face-to-face hybrid bonding removes the option. The bottom die is flipped so its redistribution layer meets the redistribution layer of the logic die above it, which leaves the PCIe PHYs facing away from the substrate they need to reach. Instead, the signals travel down through vias cut into the silicon.

Every TSV passes through active silicon and needs a buffer around it, so the vias can't be dropped wherever the PHY happens to sit. "You rarely drill straight down into the package substrate," Manmeet Walia, executive director of product management at Synopsys, said to Electronic Design, which reported that routing has to climb to one of the upper metal layers and reverse direction before descending.

Walia told the publication that electromigration and layout rules change substantially in 3D, that via count is a tradeoff between bandwidth and signals corrupting each other, and that customer logic sitting over the PHY's path down to the substrate is a challenge that Synopsys expects to work through iteratively, design by design. PAM4 leaves less room for that kind of error than the NRZ signaling used through PCIe 5.0, since it packs two bits into each symbol.

Fujitsu's Monaka processor takes the opposite approach, instead stacking four N2 compute chiplets carrying 144 Armv9 cores face-to-face on N5 SRAM chiplets using hybrid copper bonding, then putting the memory controllers and the PHYs for its 12 DDR5 channels on a separate and comparatively large I/O die rather than inside the bonded stack.

PCIe generations arrived roughly five to seven years apart for most of the standard's life and now release about every two years, with the Gen 8 specification due in 2028 at 256 GT/s per lane. Walia told Electronic Design that a further shift is coming with 3.5D packaging, where the PCIe PHYs get stripped out of the bottom die entirely, replaced with UCIe, and relocated to a side chiplet on the interposer that acts as a multi-protocol hub for Ethernet, PCIe, and CXL. Synopsys hasn't put a date on that. Its blog says leading-edge customers are evaluating angstrom-class process technologies for the top dies in their stacks.

SMIC posted its first $3 billion quarter earlier this month, with revenue up 36.1% year on year, net profit nearly tripling to $479.2 million. Co-CEO Zhao Haijun told analysts the next day that the Shanghai foundry will charge more for wafers processed in the third quarter after price negotiations concluded in the first. Utilization hit 93.7% against demand Zhao said SMIC can't fully meet, driven by Chinese AI data center buildouts that U.S. export controls have cut off from TSMC and Samsung at the leading edge. "Since there's still a big gap between industry-leading wafer prices and SMIC's current prices, we need to negotiate with customers for fairer pricing," Zhao said on the call.

The quarter blew SMIC's own out of the water on every front. The company had guided to 14% to 16% sequential revenue growth and a 20% to 22% gross margin; it delivered 20% growth to $3.01 billion and a 25.3% margin, up from 20.1% in Q1. Wafer shipments rose 14% quarter-on-quarter to 2.9 million 8-inch equivalents, blended selling prices climbed 5.7%, and Q3 guidance calls for a 26% to 28% gross margin. China accounted for 90% of revenue.

Demand isn’t coming from GPUs, however, with Zhao commenting that the surge came mostly from AI chips other than CPUs and GPUs, such as logic ICs, BCD power-management parts, and optical transceiver components, all in short supply. Meanwhile, growth in SMIC’s AI peripheral segment is expected to be around 40% for the quarter, while industrial and automotive chips rose to 16.5% of wafer revenue from 10.6% a year earlier.

From bust to boom

SMIC's utilization sat at 68.1% in the first quarter of 2023 and averaged 75% that year as net profit fell more than 60% and gross margin dropped 16.4 points to 21.9%. As late as early 2025, it was reported that SMIC and Hua Hong were cutting mature-node prices to defend share against a wall of new Chinese capacity. The company that spent 2023 and 2024 discounting into overcapacity spent 2026 raising prices by around 10% in December, negotiating targeted increases in capacity-constrained segments in February, and applying another round to Q3 wafers.

Export controls did most of the work, with Washington’s restrictions keeping China's AI accelerator demand away from TSMC. Beijing has been redirecting that demand inward: the government wants 70% of silicon wafers sourced domestically this year, and a Bloomberg Intelligence survey of 60 Chinese tech executives in June found firms plan to spend 46% of their AI accelerator budgets on local chips over the next 12 months, up from 30% now. SMIC is the only Chinese foundry that mass-produces 7nm-class logic, which makes it the sole domestic route to silicon for Huawei's Ascend line and Cambricon's accelerators. A protected buyer pool, along with a mandated shift to domestic supply and a single qualified supplier at the leading edge, produces a textbook seller's market.

Hua Hong, China's second-largest foundry, reported utilization of 102.8% in the same week, with record revenue of $717.5 million, up 26.8% year on year. TrendForce data shows foundry prices across China rose 5% to 15% between Q1 and Q2, with a third round of increases being prepared for the second half. TSMC is reportedly raising prices across all its advanced nodes too, so SMIC's hikes track a global trend, but SMIC is doing it from a captive position TSMC doesn't have: its customers have no other choice.

China's AI chip designers post record first halves

Cambricon's first-half revenue rose 108% to 6 billion yuan (c. $890 million) with net profit up 122.6% to 2.3 billion yuan, per its Shanghai Stock Exchange filing reported by the South China Morning Post. Moore Threads grew first-half revenue 147% to 1.74 billion yuan and cut its net loss by 96%, and Biren projected first-half revenue growth of more than 1,850% off a small base ahead of a Hong Kong IPO. Memory maker CXMT raised $8.6 billion in Shanghai's biggest-ever semiconductor listing last month and surged 466% on debut to become the most valuable company on any mainland exchange. Every one of these firms sits on the U.S. Entity List or depends on suppliers that do, and every one just posted record or near-record numbers.

Beijing had until recently been blocking Chinese imports of U.S. accelerators. The US approved around 10 Chinese firms to buy Nvidia's H200 in May, but China had been blocking the purchases to protect domestic suppliers. Under Secretary of Commerce Jeffrey Kessler told a congressional hearing on July 14 that "very few" H200s had actually shipped. Officials have relented as of August 19, with ByteDance and Tencent each having received around 10,000 H200 chips, the first meaningful deliveries since the U.S. approved around 10 Chinese firms as buyers.

Some 20,000 delivered accelerators against Huawei's target of 600,000 Ascend 910Cs this year leaves Chinese cloud spending, which Goldman Sachs pegs at roughly $102 billion for 2026 in combined AI capex across Alibaba, Tencent, ByteDance, and Baidu, landing overwhelmingly on domestic silicon.

SMIC's 7nm yields and the HBM shortage

SMIC's leading-edge economics remain brutal, however, with industry sources cited by the Financial Times putting SMIC's 5nm and 7nm prices 40% to 50% above TSMC's with yields of less than a third, a consequence of running multi-patterned DUV on nodes designed for EUV. The wafers SMIC is repricing are overwhelmingly mature-node parts, where its cost position is sound; the advanced capacity that feeds Ascend production stays yield-limited and expensive per good die regardless.

Memory, not logic, caps accelerator output anyway, and SemiAnalysis estimates Huawei has been drawing down a stockpile of roughly 13 million Samsung HBM stacks acquired before the late-2024 controls, and domestic HBM from CXMT will cover only a fraction of 2026 Ascend targets.

SMIC's own profit surge also comes with a glaring asterisk: CFO Wu Junfeng said the near-tripling was boosted by a one-time gain from a subsidiary. Demand for its silicon rests largely on policy rather than proven end markets, with an analyst tally cited by Asia Times putting China's top 11 listed chip firms at a combined average of roughly 122 times projected 2026 earnings.

Japanese chemical maker Ajinomoto has reportedly told customers in mainland China that it will cut supply of ABF, the insulating build-up film that's used in nearly every high-end processor package, by 30%, according to a report from the Chinese outlet JW Insights, which cites unnamed supply chain sources.

If true, that would be painful for Chinese customers like Shennan Circuits, Xingsen Technology, and Shenghong Electronics, who rely on Ajinomoto's reported 95% global market share of the film. In contrast, China's self-sufficiency rate is thought to sit below 5%.

JW Insights attributes the cut to Ajinomoto prioritizing Japanese customers and core overseas accounts, which supply the FC-BGA substrates under Nvidia, AMD, and Intel accelerators, over mainland buyers. Whether or not the 30% figure holds up, the squeeze is well documented, and China's response was underway long ago.

A confirmed shortage

Ajinomoto's ABF production ran at roughly 2 million square meters per month at full utilization in the second quarter. The company has committed ¥25 billion (around $156 million USD) since 2023 to expand capacity by about 50% by 2030, and land purchased in Kani City, Gifu Prefecture, hosts a third plant not expected to come online until around 2032.

In the fiscal year ended March 31, Ajinomoto reported that ABF sales grew 25% with margins above 50%, and the share of its film going into servers and networking silicon reached 70%, up from 40% in fiscal 2017. According to Goldman Sachs, the gap between ABF substrate supply and demand will widen from around 10% in the second half of 2026 to 21% in 2027 and 42% in 2028.

Ajinomoto notified substrate makers in May of a roughly 30% price hike taking effect this quarter, two months after UK activist fund Palliser Capital disclosed a top-25 shareholding on March 31 and publicly demanded the company raise ABF prices by more than 30%. That hike is confirmed, even if the volume cut isn't. ABF material accounts for about 30% of a substrate's bill of materials, so the increase flows directly into the cost of every FC-BGA package built on it. We've been tracking ABF crunches since the shortage that constrained GPU production in 2021 and 2022, and the current cycle looks to be extending a pattern that's already hit BT resin substrates and T-glass cloth, where single Japanese suppliers also dominate.

China has three films in qualification

Huazheng New Material's CBF, developed with the Shenzhen Institute of Advanced Electronic Materials, is the most mature of China's three named alternatives. The film uses a modified epoxy resin with spherical silica filler, which routes around Ajinomoto's IP rather than copying it. According to reports coming from Chinese media, its mass-production yield sits at above 85%, with reliability testing reportedly having passed inside Huawei Ascend systems and validation underway at Xingsen and Shennan Circuits. Huazheng's first production line of 3 million square meters per year is said to be running at full utilization, and a second line doubling that is slated to come online at the end of 2026.

Lotus Holdings, best known in China as a producer of MSG, acquired 51% of Shenzhen Newface, the developer of NBF, in April for roughly ¥103 million. Newface is said to have qualified all products below nine build-up layers, with nine- to 11-layer films in development and validation underway at Taiwanese substrate makers. Ajinomoto itself is a food and seasonings company that derived ABF from its amino acid chemistry in the 1990s.

Hongchang Electronics' GBF, co-developed with Taiwan's Jinghua Technology, has been validated at a leading domestic OSAT and is in small-volume trial production, with scale-up targeted for the fourth quarter. All three films face the same challenge of downstream reliability qualification taking one to three years of thermal cycling, damp-heat aging, and electrical testing, often longer than the R&D itself, and the highest layer-count films under flagship AI accelerators remain unmatched domestically. Upstream inputs, including specialty resins and spherical silica filler, are themselves partly import-dependent.

Huawei's Ascend packaging sidesteps ABF

Huawei's Ascend 910C reportedly connects two compute dies on separate silicon interposers through an organic substrate, an approach SemiAnalysis has described as trading die-to-die bandwidth for yield and cost against Nvidia's CoWoS.

That architecture makes Huawei less dependent on the high layer-count ABF-based FC-BGA substrates that Nvidia's B200 and GB200, AMD's MI300X, and Intel's accelerators sit on, and Chinese reporting seems to position Ascend as the anchor qualification target for both CBF and GBF. Cambricon, Biren, Moore Threads, and Alibaba's T-Head, which package on conventional FC-BGA, are directly exposed to any mainland ABF supply disruptions.

China banned exports of dual-use items to Japanese military-linked end users back in January through Ministry of Commerce Announcement No. 1, following Prime Minister Sanae Takaichi's November remarks on a Taiwan contingency, with measurable fallout. Chinese exports of restricted rare earths to Japan fell roughly 51% year-over-year in the first half of 2026, Nikkei Asia reported, and Japan imported just 13 tons of dysprosium in the period, down 82% from two years earlier, per TrendForce.

Ajinomoto's move to cut ABF supply to China eight months later has obvious retaliatory optics, despite every account of the alleged cut attributing it to capacity allocation under AI demand. China's rare-earth controls have so far targeted materials where China holds the leverage, and ABF is a market where it holds none.

Meanwhile, BOE signed a three-year glass substrate agreement with Corning in May and designated glass-core packaging a strategic business in July, and Lens Technology announced a through-glass-via collaboration with Intel the same month, extending China's push into glass substrates as the longer-term route around Japanese film.

A glass core swaps out the middle layer of a substrate, but the chip package still needs insulating film built up on either side, so glass doesn't remove the need for ABF or its substitutes. None of China's glass projects has reached mass production either. Until that changes, China's answer to the reported cut depends on whether Shennan, Xingsen, and Shenghong qualify their domestic films.

Intel is set to raise $19.7 billion by selling new common stock in a bid to finance the building out of new production capacity, the development of next-generation leading-edge process technologies like 14A and others, and day-to-day operations. While the company does not assign money to a particular project, Intel needs to build capacity to land orders from large external clients, so capacity expansion will likely be a priority. According to Bloomberg, the share sale attracted $100 billion in demand.

Intel will sell 210,526,315 shares for $95 apiece through an underwritten public offering. In addition, participating banks have 30 days to acquire as many as 31,578,947 more shares at the same $95 price, minus applicable underwriting discounts. Should they exercise all their options, Intel could sell approximately 242.1 million shares altogether and increase the proceeds to roughly $23 billion. Without the additional shares, Intel expects net proceeds of approximately $19.7 billion after underwriting discounts, commissions, and estimated expenses. The transaction is scheduled to close on August 12, 2026.

Intel's market capitalization increased from roughly $90 billion last August to $491 billion at press time, so the time is right to sell some shares and raise some much-needed cash, as the company must compete against giants like TSMC and Samsung, which spend tens of billions of dollars every year on new fabs and advanced process technologies. Meanwhile, Intel's capitalization reached its all-time high of $673 billion on June 20, 2026.

Intel has not assigned the money it is going to raise to particular projects and says the capital can be used across the business, including for capital expenditures and working capital. The company is currently ramping up its Fab 52 in Arizona and is on track to start using adjacent Fab 62 when it needs to. In addition, the company still has to build its fab complex in Ohio, which is expected to cost over $100 billion when fully built, so it badly needs money.

In its risk disclosures, the company specifically mentioned Intel 14A — which is due to enter mass production in 2028 — and other advanced process technologies, manufacturing expansion required to support them, and the need to secure design wins and volume commitments from major external foundry customers. While Intel does caution that these long-term investments amounting to tens of billions may not generate adequate returns, it is impossible to land sizeable contracts from external customers without having production capacity readily available.

Interestingly, Intel also mentioned alternative financing arrangements, government grants, and the U.S. government's significant equity position in the company among relevant factors.

House Select Committee on China Chairman John Moolenaar has written a letter that demands the U.S. government enforce an existing export control measure designed to prevent Chinese companies from getting advanced chips produced by contract chipmakers like TSMC or Samsung Foundry.

When Joe Biden was about to leave the office in early 2025, he signed a law that required chipmakers to determine their end customers in a bid to prevent contract manufacturers from unknowingly fabricating chips using American technologies for companies that served as intermediaries for restricted organizations tied to the Communist Party or the People Liberation Army. The regulation was introduced after it was discovered that chips made by TSMC for China-based Sophgo were actually Huawei's Ascend 910B AI accelerators. However, the Trump administration announced in May 2025 that it would not enforce this semiconductor regulation known as the 'Foundry Due Diligence Rule,' which naturally created uncertainty about whether the foundry-focused requirements would be actively implemented at all.

"The announcement created ambiguity as to whether front-end fabricators like TSMC can export unpackaged advanced dies to non-approved designers located outside of China, without performing the due diligence specified in the Foundry Due Diligence Interim Final Rule (IFR)," the letter reads. "These exports enabled Huawei to obtain millions of controlled Ascend logic dies from TSMC, using its front company, Sophgo in 2023 and 2024."

Before the adoption of the 'Foundry Due Diligence Rule,' Chinese companies and/or their intermediaries could misrepresent chip specifications and end users when placing orders with TSMC or other foundries, which allowed restricted devices to be made despite U.S. export controls. Under the rule, foundries and OSAT providers exporting chips produced on 14/16nm-class process technologies or more advanced nodes must presume those devices qualify as controlled AI processors subject to a global licensing requirement, unless they meet an exemption. According to the letter, foundries have generally complied with these requirements since the rule took effect. As a result, Moolenaar essentially asks the government to enforce the existing controls rather than introduce stricter export controls.

Moolenaar believes that the Commerce Department's Bureau of Industry and Security could clarify its current stance on the Foundry Due Diligence Rule in two ways:

• By issuing guidance, which confirms that the worldwide Regional Stability (RS) licensing requirement still applies to exports from front-end foundries;
• By formally annulling the AI Diffusion IFR and amending §744.23 to explicitly restore that requirement for both foundries and OSAT providers.

Either approach would eliminate conflicting interpretations and reinforce enforcement of existing export controls, Moolenaar believes.

Apple used to be among a few companies willing to buy memory and other components worth billions of dollars under long-term supply contracts at fixed prices. But the artificial intelligence era represents a new reality with new purchasing champions, marking a tectonic shift in the high-tech world. Alphabet, Microsoft, Meta, and Amazon have purchase commitments totaling about $2 trillion, and a significant portion of these commitments are for memory, according to estimates by analyst Claus Aasholm. While the commitments are approximate, span many years, and should be generally taken with a grain of salt, they still reflect the direction the industry is moving.

Combined purchasing commitments from the four major hyperscalers shown in the chart —Amazon, Alphabet, Meta, and Microsoft — reached nearly $2 trillion by Q2 2026, with Alphabet and Microsoft accounting for the overwhelming majority of the total.

The rapid expansion suggests several major findings. Firstly, the AI infrastructure race is accelerating, not stabilizing. Secondly, AI infrastructure investments are driven by a handful of hyperscale cloud service providers (CSPs) whose long-term procurement commitments now vastly exceed those of traditional consumer electronics companies such as Apple.

Thirdly, memory has become a strategic asset — perhaps a competition weapon — rather than a commodity. Fourthly, suppliers of memory — both 3D NAND and DRAM — are gaining pricing power. Finally, demand for memory will likely drive major capacity expansion at Micron, Samsung, and SK hynix, even though so far these companies have been exceptionally disciplined about their capacity investments.

Almost $2 trillion commitments

Google shows by far the most aggressive increase in purchasing commitments, rising from roughly $140 – $150 billion in Q3 2025 to around $811 billion by Q2 2026 (though these are total purchase commitments by Alphabet, not specifically memory purchase commitments), while Microsoft follows a similar trajectory and reaches approximately $678 billion in total obligations, which includes, but is not limited to memory.

Meta is also ramping commitments substantially to around $349.3 billion (again, these are total commitments), whereas Amazon increased its commitments more gradually to roughly $130 billion. By contrast, Apple — which makes the world's most popular smartphone, and which was the largest consumer of memory just a couple of years ago — remains almost flat throughout the period at approximately $57 billion (of which $56.2 billion is payable within 12 months). Apple's commitments fall well short of Nvidia's commitments of $119 billion.

Again, we are talking about total purchase commitments, which include foundry capacity, 3D NAND, and DRAM memory, but are not limited to them. Alphabet, Amazon, Meta, and Microsoft all build custom silicon and custom servers, so a significant portion of these commitments is to various EMS providers.

While $1.968 trillion of purchase commitments for memory and storage alone would be an absurdly large amount of money, a huge portion of these commitments consists of contract manufacturing obligations as well as memory chips. This suggests that the foundry, 3D NAND, and DRAM markets are entering a new phase in which hyperscalers are willing to make vastly larger forward purchasing commitments than traditional consumer-electronics companies, giving suppliers a strong incentive to prioritize customers prepared to secure future capacity on that scale.

Strategic assets

While Claus Aasholm's chart is explicitly dedicated to memory, it does describe total purchase commitments of tech giants, so the chart can reasonably be read as evidence that memory and capacity at TSMC, Samsung Foundry, and GlobalFoundries are becoming a strategic asset rather than merely another component to procure at the best available price.

AI infrastructure requires enormous quantities of AI accelerators, DRAM (including HBM), and 3D NAND. Meanwhile, the supply of high-end memory (HBM) is constrained by fab capacity at major DRAM makers, whereas the supply of AI accelerators is constrained by both wafer capacity and foundries and packaging capacity at foundries and their OSAT partners. As a result, hyperscaler CSPs have an incentive to lock in supply years ahead, even if doing so requires exceptionally large purchasing commitments.

That also changes the relationship between semiconductor suppliers and their customers. In theory, a company willing to guarantee hundreds of billions of dollars of future purchases can effectively help underwrite expansions of foundry, memory, and advanced packaging capacity and, in return, secure priority access to scarce products and future process technologies. In reality, TSMC can well afford capacity expansion using the money it gets from hyperscalers and give priority to its largest customers. In this environment, access to DDR5, HBM, and 3D NAND memory becomes part of the competitive advantage rather than merely a procurement exercise.

This is also what makes Apple's position in the graph interesting: its purchasing commitments barely move while those of Alphabet, Amazon, Meta, and Microsoft surge. If the trend continues, Apple may remain one of the world's largest semiconductor buyers in absolute terms, but the question is whether it will be among the key customers that foundries, memory makers, and OSATs plan their future capacity expansions.

An inflection point

Perhaps the most interesting takeaway of the findings revealed by long-term purchase commitments is that the industry's center of gravity appears to have shifted.

During the smartphone era, foundries (well, TSMC has won) and memory suppliers often competed aggressively for Apple's business because of its enormous purchasing power. Today, hyperscalers building AI infrastructure are making purchasing commitments that dwarf those of traditional CE companies like Apple, which may well represent a strategic inflection point akin to the one Andy Grove described in his 'Only the Paranoid Survive' book.

Will this tectonic shift result in prioritization of customers capable of enabling future capacity expansions through massive long-term purchase agreements, or will foundries and memory makers remain more or less disciplined with their capacity expansions so as not to lose a lot when demand declines? This is a question that has yet to be asked.

In any case, the AI megatrend has transformed semiconductors — from foundries to advanced packaging and from DDR5 to HBM4 — into strategic assets that can no longer be treated as ordinary components procured on demand. And this is something that will continue in the long run.

Most megafactories usually take years to go from conceptualization to construction, but a recent drone flyover of the Terafab showed that progress has already started on the ground less than five months after Musk unveiled it. However, the video doesn’t do justice to the true scale of its footprint, so X user Nic Cruz Patane created a visualization to help us understand how large the chip manufacturing facility is.

The site will reportedly have a floor space of at least 100 million square feet, making it larger than Giga Texas (10 million sq. ft), the Pentagon (6.6 million sq. ft), Apple Park (2.82 million sq. ft), and the Mall of America (5.6 million sq. ft), combined. This also makes it significantly larger than New Century Global Center in Chengdu, China, with has an interior space of “just” 18.9 million sq. ft.

While this might seem like an absurd amount of space for chip manufacturing, it appears that the Terafab will need it because it’s going to be more than just a chip fab making AI processors — instead, it will be an all-in-one facility that will produce logic and memory chips, as well as have lithography, packaging, and testing under one roof.

Elon Musk started talking about building his own chip manufacturing facility in late 2025 and officially announced the project in March of this year. The reasoning behind this project is that both SpaceX and Tesla will require at least 1TW of compute, which is more than ten times that current global chip supply.

Nvidia CEO Jensen Huang warned that a project like this will be an “extremely hard” challenge, but it appears that Musk is willing to put his massive resources on the line for this. Intel CEO Lip-Bu Tan even said that he can “think of no better partner than Elon Musk” to explore “unconventional” ways of chip manufacturing.

Musk is no stranger to both leading and funding projects that otherwise seemed impossible. Although he did not found Tesla, his investment and leadership in the company led it to become an industry trailblazer in EVs. He fundamentally changed commercial space travel with SpaceX; something that used to be the purview of NASA and other national governments, and he also broke a record when he set up 100,000 Nvidia H200 GPUs in just 19 days back in 2024 — a process which Huang said usually takes four years.

It seems that Terafab is Musk’s biggest project to date, which is estimated to require up to $119 billion in investments. But even though it seems that this project already has a secure customer base in SpaceX and Tesla, the former acknowledges that there’s a risk that this ambitious megafactory may not be successful.

SpaceX and Tesla on Thursday formally unveiled plans for the initial phase of their Terafab project. The first stage of the plan — which is expected to use Intel's 14A process technology — is expected to require $16.8 billion in investment, while the completed campus is planned to encompass more than 100 million square feet of manufacturing space. The massive semiconductor manufacturing complex will be built in Grimes County, Texas, at a location that belongs to SpaceX.

According to SpaceX and Tesla, their combined demand for semiconductors is projected to exceed 1 terawatt (TW) of compute per year, which significantly exceeds today's global supply. In fact, SpaceX, Tesla, and xAI already consume a significant portion of contract chipmaking services available today, and can potentially justify building a dedicated fab that will exclusively serve Elon Musk's companies. In particular, Terafab is envisioned to exclusively produce AI inference processors for Tesla Optimus humanoid robots and Cybercab autonomous vehicles, as well as 'high-power' processors intended for SpaceX's space-based data centers. Meanwhile, SpaceX and Tesla have not disclosed when their combined demand for compute per year will hit the 1 TW benchmark.

Unlike a conventional semiconductor fab, Terafab is envisioned as a vertically integrated manufacturing campus where advanced logic devices, memory chips, packaging, and testing operations are housed together. Normally, logic and memory are produced at different fabs using different process technologies, whereas packaging and testing services are performed at different facilities. However, the companies believe that consolidating logic, memory, packaging, and testing in one location will not only shorten the production cycle but will also shorten time-to-yield by enabling faster iterative improvements.

Given the description and the goal of the project, this will be a massive facility. Yet, SpaceX and Tesla have said little about its capabilities; The only thing they have disclosed is that the current facility in Grimes County will feature '100 million square feet of manufacturing space.' The 100 million square feet (9.3 million square meters) figure immediately stands out because it is far beyond anything ever announced for a semiconductor manufacturing facility. For example, the total area of Samsung's Pyeongtaek campus is approximately 2.89 million square meters, or 31.1 million square feet. A single Samsung fab occupies about 120,000 square meters, or 1.29 million square feet. However, the key wording is important: '100 million square feet of manufacturing space' does not mean 100 million square feet of cleanroom space.

Based on an image published by SpaceX, the Terafab facility will occupy four massive buildings. It is unclear whether these buildings will be four phases of the project (i.e., logic, memory, and packaging will be made under one roof) or will serve different purposes (i.e., one building makes logic, another produces memory, yet another does testing and packaging). In any case, when fully built, Terafab will be a massive semiconductor production campus that will require significantly more than $16.8 billion.

The announcement claims that the Terafab facility will employ at least 3,000 people, and that between 60% and 80% of them will be Grimes and nearby Brazos County residents.

Terafab is expected to use water from Gibbons Creek Reservoir instead of local groundwater and feature on-site wastewater treatment as well as water recycling and conservation measures.

The announcement follows Tesla's groundbreaking earlier this year on a research semiconductor facility at the North Campus of Tesla's Giga Texas campus, which the companies describe as a precursor to Terafab.

China has started low-volume production of domestically developed immersion DUV lithography machines, with around five systems planned for this year and roughly 20 in 2027, according to a report from July 27, which wiped roughly $44 billion off ASML's market value.

The manufacturer was named as Shanghai Aishengna Electronic Technology Group by Reuters the following day, a state-owned company established in August 2023 with 7 billion yuan (around $1 billion) in registered capital that absorbed engineering teams from Huawei-affiliated startup Yuliangsheng and state scanner maker SMEE. The first units are due at SMIC, Hua Hong, and CXMT this year for production-line validation rather than volume output, but the machines are far from matching ASML's models and still require further testing.

Five machines represent about 3.8% of the roughly 130 immersion systems ASML deploys in a typical year (a class of lithography machines that put a super thin layer of liquid over the wafer during processing), with the European giant holding an estimated 98.7% of the total immersion market. The Chinese tool reportedly prints 28nm-class features in a single exposure and reaches 7nm, and theoretically 5nm, through multipatterning, the same route SMIC already runs on its installed ASML fleet. Neither company has confirmed the report, no machine has been shown publicly, and no throughput or overlay figures have been disclosed, against the 330 wafers per hour and 2.5nm overlay of ASML's current flagship immersion tools.

SMIC’s scanner

SMIC has been running a domestic immersion scanner, developed under the codename Mount Everest, since September last year, when the Financial Times reported that China's largest foundry had begun testing the Yuliangsheng tool with production-line integration targeted from 2027 after qualification.

FT compared the machine against ASML's Twinscan NXT:1950i, a system that entered the market in 2008, which puts the design roughly a decade and a half behind the tools ASML sells today. Yuliangsheng, founded in Shanghai in 2022 with 1 billion yuan (around $149 million) in registered capital, is understood to have delivered three lithography machines to fabs for testing by late last year, but this hasn’t been officially confirmed.

Many of the critical components needed for the machines are still imported from Japan, and delays at those local suppliers are what have capped 2026 output at around five units. The 2027 target of 20 machines is therefore an ambitious target that assumes a domestic component base that hasn't been established yet, and the imported parts remain within reach of any future export-control round.

18% litho localization

SMEE's most advanced shipping product remains the SSA600 series, a 90nm-class dry ArF scanner that was in mass production as of May last year. The 28nm-capable SSA/800 immersion tool the company announced in 2023 has never been deployed, and a state-media claim of its successful development was deleted shortly after publication. In December last year, SMEE won a roughly RMB 110 million ($16 million) single-source government contract for a KrF scanner specified at 110nm resolution and 15nm overlay, a useful indicator of where its production-grade capability is at the moment.

Domestic equipment took 35% of Chinese fab purchases in 2025, beating Beijing's 30% target and up from about 10% three years earlier, according to figures from China's semiconductor industry association. Etch and thin-film deposition passed 40% localization, and metrology reached 25%, while lithography managed just 18%, most of it in trailing-edge and packaging tools. From the end of 2025, new fab capacity additions are required to source at least half their equipment domestically, a mandate that guarantees the new scanners a customer base, whatever their specs turn out to be.

Etch, deposition, and everything else

Naura Technology became the world's fifth-largest chip equipment maker by 2025 sales, trailing only ASML, Applied Materials, Lam Research, and Tokyo Electron, and moving ahead of KLA. The company booked RMB 27.14 billion ($4 billion USD) in revenue in the first three quarters of 2025 against RMB 6.05 billion ($589 million USD) for all of 2020, holds an order backlog stretching into 2027, and has started developing lithography tools for the first time. AMEC grew revenue and net profit by more than 30% in 2025, while cleaning specialist ACM Research posted $901.3 million for fiscal 2025 and guided to as much as $1.18 billion for 2026.

SiCarrier, the Shenzhen firm widely linked to Huawei, debuted around 30 tools at SEMICON China in March 2025, spanning etch, deposition, metrology, and test, and was reportedly valued at RMB 65 billion ($9.63 billion USD) by September 2025 with more than RMB 10 billion ($1.48 billion USD) in orders. The U.S. added it to the Entity List in December 2024.

The sector's own leadership is less triumphant than the numbers suggest. In March, SMIC co-founder Wang Yangyuan and the heads of YMTC, Naura, and EDA firm Empyrean jointly described China's tool industry as "small, fragmented, and weak" and called for national consolidation, with Big Fund III's $47.5 billion being redirected toward lithography and EDA. The Aishengna venture, which merged the Yuliangsheng and SMEE teams into one state-owned vehicle, looks like the first product of that pressure to consolidate. China spent a record $49.5 billion on wafer fab equipment in 2024, per SEMI, and remains the world's largest buyer through 2027, even after a pullback in 2025.

EUV lithography

A Reuters investigation in December 2025, dubbed China’s “Manhattan Project,” described an operational prototype EUV light-source machine in a high-security Shenzhen lab, completed in early 2025, that generates EUV photons but hasn't yet exposed a wafer. More than 3,000 researchers are said to work across the program, with Huawei playing the coordinating role and SMEE handling system integration.

Two teams are chasing the light source: one led by Lin Nan, a Beihang University professor who worked at ASML from 2015 to 2021, has a solid-state laser-produced-plasma design running at 3.42% conversion efficiency against the roughly 5.5% commercial viability required, while Zhao Yongpeng's laser-assisted discharge plasma group at Harbin Institute of Technology has reached around 100W of EUV output against the roughly 600W that ASML's production sources deliver.

Beijing's target is chip output from the machine by 2028, with Reuters' sources calling 2030 more realistic. A separate strand of reporting described a reverse-engineered prototype built around an intercepted Cymer light source that has likewise produced nothing. Earlier claims of a Huawei EUV trial production in 2025 and mass production in 2026, which circulated via Chinese media in March 2025, were never confirmed by any primary source. Tsinghua University's accelerator-based SSMB concept, which would need a synchrotron of a staggering 100 to 150 meters in circumference, remains an academic project.

Export controls

The MATCH Act, introduced in the House and Senate in early April, would ban not only sales of immersion DUV tools to SMIC, Huawei, Hua Hong, CXMT, and YMTC but also the servicing of machines already installed, and would give the Netherlands and Japan 150 days to align.

At this stage, it remains a bill, not a law, but servicing restrictions would strike directly at the installed ASML base that produces every advanced chip China currently makes, including SMIC's N+3 node in Huawei's Kirin 9030. Chinese fabs have been preparing for exactly that scenario by using third-party engineers and gray-market parts to Frankenstein older ASML machines.

ASML's China exposure is already shrinking on schedule, with the country representing 20% of system sales compared to 41% in 2024 and 33% last year, even as the company raised full-year guidance to €43 to €45 billion in July. Likewise, Applied Materials expects to lose $600 to $710 million in China revenue this fiscal year.

Ultimately, three markers will indicate whether China’s domestic DUV program is a legitimate rival or yet more state-sanctioned hot air. The main one would be validated production wafers from an Aishengna tool at SMIC, Hua Hong, or CXMT with published throughput and yield, followed by delivery of anything close to the 20 machines planned for 2027, and a first exposed wafer from the Shenzhen EUV prototype before the 2028 target set by Beijing.

China has revised its regulations concerning protection of integrated circuit (IC) layout designs developed domestically. The new regulations are intended to make it harder for Chinese companies to obtain legal protection for copied chip designs by tightening originality requirements, registration procedures, and infringement remedies, Reuters reports. As a result, it will get harder for underperforming China-based companies to copy ICs developed by their more successful rivals. Meanwhile, Chinese chipmakers will be allowed to produce their designs in Taiwan or South Korea.

The amended regulations apply to physical chip layout designs that define how circuit elements are arranged on silicon and represent a substantial amount of engineering work, including for companies that specialize in chip design rather than manufacturing. To qualify for legal protection of their IC layouts, applications must now demonstrate that the layout was independently developed, formally attest that the design is original, and specify which parts of the layout constitute their own creative contribution (perhaps to give authorities an idea about which were licensed or obtained from open sources). As a result, authorities will be able to filter out weak claims and distinguish companies with strong technological capabilities. Furthermore, they will also be able to determine whether a particular chip was designed and built in China, or its alleged developer obtained its product elsewhere and attempts to disguise it as a 'Made in China' silicon.

The new rules also strengthen enforcement. From mid-October and onwards, in infringement disputes, courts may determine compensation based either on losses suffered by the rights holder or profits earned by the infringing party. Punitive damages will also become an option. In addition, the regulations clarify procedures for licensing, transferring, or using IC layout-design rights as collateral. Organizations that develop protected layouts are also required to provide reasonable rewards and payments to personnel responsible for creating those designs.

Chinese policymakers were also reportedly considering measures to prevent strategically important domestic technologies from being transferred abroad, acquired by foreign companies, or produced overseas. If these strict proposals were approved by the CCP and the government, Chinese chip designers would be unable to produce their chips at TSMC in Taiwan or Samsung Foundry in South Korea, and would force them to exclusively make them domestically at Semiconductor Manufacturing International Corp., Hua Hong, and other domestic contract producers that are decades behind market leaders.

The report claims the updated regulations reflect the strategic importance China now assigns to technologies developed by its domestic semiconductor industry. For now, it is evident that China is not ready to implement overseas production of advanced designs. However, the report does not explicitly claim they are completely off the table, according to Reuters.

The updated rules were signed by Premier Li Qiang on July 23 and will take effect on October 15.

Drone flyovers of Elon Musk’s new Advanced Technology Chip Fab (ATCF) in Texas have been shared on social media. The videos and photos come courtesy of flight instructor and drone videographer Joe Tegtmeyer (@JoeTegtmeyer), who regularly documents the progress of Giga Texas and Starbase projects. Considering that the fab was only announced on March 21 this year, groundwork has been brisk, and in a follow-up post, Tegtmeyer states progress has “hit another gear.”

In the above post, the all-things-space enthusiast notes that the ATCF already appears to be very well defined. “GeoPiers are at work on the part of the foundation that has had gravel mix placed on the top soil and more work grading on the south end,” he explains.

A few days later, there was an update with progress clearly accelerating, according to Tegtmeyer. As well as commenting on the shape of the facility becoming ever clearer and construction materials continuing to be stockpiled, there are other important signs of progress. Specifically, “4 new permits were filed in the past few days for the large trailer complex that houses the general contractors and the temporary Tesla offices for the new facility,” noted the enthusiast. “[It is] a great sign that things are progressing rapidly behind the scenes to fit out the facility when construction progresses far enough to allow for equipment installation.”

Importantly, Tegtmeyer reminds his followers that the ATCF is not the Terafab that will be built ~ 100 miles or so to the East in Grimes County, Texas. Rather, the ATCF is the equally vital but “smaller development facility that is the first part of the overall AI Chip development program.”

The ATCF is a joint Tesla–SpaceX facility, and is a keystone for the Tesla, SpaceX, and xAI accelerator chips to come. It is expected that chips from the ATCF will combine logic, memory, and advanced packaging under one roof. Projects spawned from here are destined to support Tesla cars and Robotaxis, Optimus robots, and upcoming Earth-orbiting AI data centers.

That’s distinct from the much larger, multi-phase, vertically integrated, 1TW/year Terafab megaproject, which will see collabs between Tesla, SpaceX, xAI, and Intel. It has massive AI computer production ambitions to feed terrestrial and space-based AI systems. We are still awaiting evidence that the Terafab has moved beyond the pre-construction phase.

If there were any doubt about the seriousness of Musk's semiconductor design and fabrication ambitions in these and other progress reports, ASML CEO Christophe Fouquet recently confirmed talks with Musk had taken place about building one of the largest chip manufacturing operations ever attempted. SpaceX has also already filed for a $55 billion facility in Grimes County, Texas, with potential expansion costs reaching $119 billion, and phase one potentially beginning before the end of the year. Meanwhile, it is good to see the rapid progress of the smaller ATCF.

Lumentum CEO Michael Hurlston told an audience at the RAISE Summit in Paris earlier this month that indium phosphide, the compound semiconductor behind every laser in an AI data center, is heading into a supply squeeze worse than what we've already seen with DRAM / NAND, and that Nvidia's decision to fund Lumentum and its biggest competitor at the same time was a response to exactly that.

In his remarks, Hurlston said that telecom customers bought lasers in the hundreds, while Nvidia and the hyperscalers are asking for hundreds of millions. While Lumentum runs five indium phosphide fabs, it's still shipping more than 30% below what customers want. Nvidia's answer, in March, was to write $2 billion checks to Lumentum and Coherent, the two suppliers that, between them, make most of the world's high-speed datacom lasers, with purchase commitments and future capacity access attached to both.

"Between the two of us, I don't think we can service the demand that Nvidia and others are now putting on us to solve this resistance problem in the data center," Hurlston added.

Silicon doesn't emit light

Indium phosphide has a direct bandgap of roughly 1.34 eV, which lets it convert electrical current into photons efficiently. Silicon's bandgap is indirect, so it can guide, split, and modulate light but can't generate it. Every silicon photonics platform in production, including those of Nvidia, Broadcom, Marvell, and Cisco, still needs an indium phosphide laser somewhere in the package to supply the light for silicon to manipulate. Moving from pluggable transceivers to co-packaged optics changes where that laser sits and how it's mounted, but it doesn't remove it from the bill of materials.

Nvidia's marketing claims its photonics switches use four times fewer lasers than an equivalent pluggable deployment, alongside 3.5 times better power efficiency and ten times better network resiliency, all of which are vendor figures. Those savings are per port, and it's that port count that's exploding.

The high-end Spectrum-X Photonics configuration runs 512 ports at 800 Gb/s for 400 Tb/s of switching, and Quantum-X Photonics InfiniBand runs 144 ports at 800 Gb/s. Co-packaging also shifts the laser type toward high-power continuous-wave sources and external laser modules that feed multiple channels, which are harder to build than the electro-absorption modulated lasers inside a conventional pluggable. Coherent's Nvidia agreement covers that category of high-power CW lasers, external laser source modules, and fiber array units.

Capacity at Lumentum and Coherent

Lumentum posted record revenue of $808.4 million in its fiscal third quarter, up 90% year over year, with components revenue of $533 million and pump laser shipments up 80%. On the May earnings call, Hurlston told analysts the company expects its supply line to increase 50% measured from one December quarter to the next, and in the same breath said the supply-demand imbalance on EMLs had widened from the 25% to 30% given a quarter earlier to "somewhere greater than 30%," with pump lasers tighter still. A supplier growing output by half a turn per year and losing ground anyway is a clean measure of how steep the demand curve is.

Coherent's 6-inch indium phosphide line yields more than four times as many devices as its 3-inch line at less than half the cost, CEO Jim Anderson told investors on the company's fiscal Q3 call. Anderson said EMLs, CW lasers, and photodiodes are all in production on the 6-inch line with yields above the legacy 3-inch lines, and that internal capacity would double by the end of the June quarter, one quarter ahead of plan, then more than double again by the end of 2027. Coherent's revenue hit a record $1.8 billion, up 21%, with data center and communications now 75% of the total against roughly 41% a year earlier, and backlog stretching into 2028.

Logic and memory moved to 300mm wafers in the early 2000s. Indium phosphide is a brittle, expensive, small-boule material where the industry-wide upgrade currently underway is 3-inch to 6-inch, roughly the transition silicon completed in the 1980s. Lumentum's fifth fab, announced in March, is a converted Qorvo gallium arsenide plant in Greensboro, North Carolina, described as 4-inch and 6-inch compatible and ramping around 2028.

Running through China

Indium is recovered as a byproduct of zinc refining, so its output can't be scaled independently of zinc economics, no matter how much laser demand there is. The USGS Mineral Commodity Summaries 2026 put China at an estimated 760 tonnes of roughly 1,100 tonnes of global primary refined indium in 2025, about 69%, and recorded a 72% year-over-year fall in unwrought indium exports between September 2024 and September 2025 after Beijing placed the metal under export controls in February last year. The U.S. warehouse price averaged about $390 per kilogram in 2025 against $340 in 2024.

AXT's Chinese subsidiary Tongmei had to obtain Ministry of Commerce export permits, granted in June and August 2025, before it could resume shipping indium phosphide substrates out of China. The fabs Nvidia is funding sit downstream of that licensing regime, and the wafers going into them aren't made in the United States in meaningful volume.

DRAM contract prices rose 90% to 95% quarter over quarter in Q1 2026, the largest quarterly increase TrendForce has recorded, and the firm forecast a further 58% to 63% in Q2 with NAND up 70% to 75%. HBM is sold out for 2026. Hurlston is measuring his warning against a genuinely historic crunch, which makes it a strong claim rather than a throwaway one, and he runs a company whose valuation depends on the shortage persisting.

LightCounting's April 2026 market forecast puts current transceiver demand about 30% above supply, matching Lumentum's own figure, but states that the shortages should be gone by the end of 2026 and cuts expected Ethernet transceiver growth to 65% for the year after 82% in 2025 and 93% in 2024. Coherent, hitting its capacity doubling a quarter early, points the same way. The distinction from memory is that the fix here is a wafer-size transition already running in production with yields ahead of the old node, not a new fab that takes three years to build.

Reuters has named Shanghai Aishengna Electronic Technology Group as the state-owned company producing China's first domestic immersion deep ultraviolet lithography (DUV) scanners, a day after news of the program broke without identifying the manufacturer. Aishengna was established in August 2023 with RMB 7 billion, around $1 billion, in registered capital and is thought to have absorbed engineering teams from Shanghai Yuliangsheng Technology and Shanghai Micro Electronics Equipment.

Aishengna has been named by a single source who declined to be named, and its shareholders, SMEE and Yuliangsheng, didn’t respond to requests for comment. SMIC has been testing a Yuliangsheng immersion tool since September 2025, and first deliveries are slated for SMIC, Hua Hong Semiconductor, and ChangXin Memory Technologies.

Photoresist, coater tracks, and light sources

Tokyo Electron held an 89% share of the global coater/developer market in 2022, per Shared Research's analysis of the company's own disclosures, with its chief executive putting the figure near 90% and at 100% for EUV production. A scanner only exposes the wafer, however. It’s the track that's responsible for applying the resist film, baking it, and developing the pattern after exposure, and it has to be mechanically and thermally matched to the scanner in a single cluster, which is why the two are bought together. Shenyang Kingsemi has reached 28nm-class track capability and is currently targeting 14nm.

JSR, Tokyo Ohka Kogyo, Shin-Etsu, and Fujifilm hold a combined 72.5% of the ArF photoresist market, while Chinese suppliers hold under 1% of ArF immersion resist specifically. Nata Opto-electronic built a 25-ton ArF line, later expanded to 50 tons, passed customer qualification in December 2020, and completed project acceptance in 2024 with little volume to show for it. Xuzhou B&C says its ArF immersion products cover 45nm to 28nm and can stretch to 14nm, and chairman Fu Zhiwei has put mass production of China's core advanced resists five years out.

Cymer, Gigaphoton, and Coherent hold more than 80% of the ArF excimer laser market between them, and Cymer has been an ASML subsidiary since 2013. Beijing RSLaser shipped China's first high-power domestic excimer laser in 2018 under the national Project 02 program and has a 4 kHz 193nm ArF prototype aimed at 90nm and 65nm-class tools, generations behind what 28nm immersion requires.

Carl Zeiss SMT has been ASML's sole projection optics supplier since 1983, and Zeiss SMT revenue grew from €1.2 billion in 2016 to €4.1 billion in 2024. It’s not currently known what, if any, Japanese tooling is inside the Aishengna machines, but excimer sources and precision optics are areas where Chinese substitution is believed to be lacking.

CXMT

CXMT is projected to reach around 350,000 wafer starts per month by the end of 2026, roughly 25,000 short of Micron, up from 40,000 in 2020. DRAM scaling at 1a and 1b-class nodes runs on immersion multipatterning because CXMT has no EUV access, which makes any potential ramp lithography-gated rather than cleanroom-gated. DRAM contract prices rose 93% to 98% quarter on quarter in Q1 2026, and TrendForce projected a further 58% to 63% in Q2, lifting DRAM industry revenue 81% to $97 billion. A domestic immersion source is therefore worth having to CXMT, even at inferior overlay and throughput.

A DUV-only 7nm flow needs roughly 19 spacer-defined multipatterning masks from the front end through the second metal layer, against about 10 for an EUV-based N7+ process, by one published comparison of SMIC's process. SemiAnalysis has put SMIC's 7nm defect density near 0.14, around double TSMC's N5 and N6. ASML CEO Christophe Fouquet told analysts during the company's July earnings call that rising DRAM lithography intensity partly reflects "the increased replacement of multi-patterning with more cost-effective single-exposure EUV." As such, every exposure China adds to compensate for the missing EUV burns scanner hours a thin domestic fleet doesn't have.

The MATCH Act

China fell to about 14% of ASML's sales in Q2 2026 from 33% across 2025, and installed base management, the service and upgrade business, brought in €2.8 billion of ASML's €9.3 billion in second-quarter revenue. H.R. 8170 would ban both the export and the servicing of immersion DUV systems to any destination in China and designate SMIC, Hua Hong, Huawei, CXMT, and YMTC as restricted entities by statute. Former ASML chief executive Peter Wennink has said the company can service most Chinese tools, but not with spare parts of U.S. origin that fall under export control, which is the mechanism that the MATCH Act would widen.

The bill remains in committee after clearing the House Foreign Affairs Committee in April, with a Senate companion filed as S. 4281, and no floor vote yet scheduled. Its 150-day allied-alignment clause would also reach Nikon, which sold 11 ArF immersion systems in FY2024 and none in the first three quarters of FY2025, and which plans to deliver a new immersion prototype to a major chipmaker by 2027. ASML expects about 130 immersion shipments this year and intends to raise immersion capacity 30% in 2027, with a further 30% under investigation for 2028.

As for Chinese providers, SMEE prototyped its SSA600 ArF tool in 2011 and never reached sustained commercial sales, and a late-2023 shareholder claim that the company had developed a 28nm machine was subsequently retracted. SiCarrier showed etch, CVD, PVD, and ALD tools at SEMICON China 2025 without a lithography system, and a December 2025 government contract reported as a lithography award turned out to cover a KrF tool at 110nm. We’ve previously assessed that China’s toolmakers are more than a decade behind the market leaders.

The AI Futures Project's June forecast puts a commercially viable domestic 7nm-capable immersion scanner between 2032 and 2038, with a median of 2035, and claims ASML holds 98.7% of the immersion market today. Five machines in 2026 would be under 4% of ASML's annual immersion output, and each would still need a coater track, a qualified ArF immersion resist, and an excimer source to print a single wafer; China leads in none of those three.

A state-backed company in Shanghai has begun mass-producing immersion deep ultraviolet lithography machines and is due to deliver the first units this year to SMIC, Hua Hong Semiconductor, and memory maker ChangXin Memory Technologies, according to The Information, citing two people familiar with the program. Output targets around five machines in 2026 and roughly 20 in 2027, and all three named recipients sit on the list of Chinese firms that a bill now moving through Congress would cut off from ASML sales and servicing by statute.

The Information didn't name the manufacturer, but its sources described the operation as having pulled DUV development teams from several Chinese companies, one of them the state-backed startup Shanghai Yuliangsheng Technology. SMIC has been testing a Yuliangsheng immersion tool since September 2025. Most components in the new systems are domestic, though some critical parts still come from Japan, and delays at local suppliers have held back output this year.

U.S. House Resolution 8170 designates SMIC, Hua Hong, CXMT, Huawei, and YMTC as restricted entities in law, and three of those five are the named first customers for the domestic scanner. The MATCH Act, introduced in April, was reported out of the House Foreign Affairs Committee on April 22 and has a Senate companion filed as S. 4281. Its immersion DUV provisions cover servicing and technical assistance, not just new exports, which extends its scope to installed tools already operating in Chinese fabs, fabs which have spent the past two years stretching that installed fleet through secondary-channel upgrades.

ASML expects to ship about 130 immersion systems in 2026, matching 2025, CFO Roger Dassen told analysts during the company's July earnings call. Dassen added that ASML intends "to increase capacity by 30% in 2027" for immersion, and is investigating another 30% for 2028. China accounts for around 20% of ASML's net sales this year, down from 33% in 2025, driven mainly by mainstream logic demand.

Immersion DUV prints 28nm-class features in a single exposure and reaches 7nm through multipatterning, at a cost in overlay errors and yield. ASML CEO Christophe Fouquet told the same call that rising DRAM litho intensity partly reflects customers replacing multipatterning with cheaper single-exposure EUV.

Independent analysis from the AI Futures Project in June put commercial-scale Chinese immersion DUV in the mid-2030s, with ASML holding 98.7% of the immersion market. Qualifying the new machines for production lines could take many months, and they trail ASML's tools on performance and build quality. China's domestic EUV effort, which Reuters first reported as a working prototype in December, remains years away.

A new semiconductor firm, TYLsemi (pronounced Tile Semi), publicly revealed itself this month, alongside $43 million in early-stage funding and an ambitious plan to simplify the development of custom processors for AI infrastructure.

Dozens of contract chip designers can develop custom processors of different complexity. However, only a few companies can offer custom silicon design services using standard chiplets to speed up and derisk the development cycle. TYLsemi is aiming to join their ranks. We spoke to their founders to find out how the nascent business might pull it off.

Emerging from stealth

Rather than compete solely as another custom ASIC design house, TYLsemi intends to offer reusable, standards-based connectivity, power delivery, and eventually memory chiplets that customers can combine with their own differentiating compute silicon to build a unique system-in-package. For companies that do not intend to conduct semiconductor development themselves, TYLsemi will also provide an end-to-end service that includes design and implementation of a differentiating chiplet, packaging, qualification, and high-volume production, essentially enabling companies without any silicon development skills to offer their own multi-chiplet processors.

TYLsemi was co-founded by Mohit Gupta and Sunil Bhardwaj, semiconductor veterans who have led global engineering, operations, and business teams at Alphawave, SiFive, Cadence, Rambus, and other chip companies, and who collectively have plenty of experience with both standard and custom silicon. Mohit Gupta, a co-founder and chief executive of TYLsemi, believes that the time to establish a company that specializes in pre-approved chiplets and custom ASIC design is right now.

"Chiplets have been discussed for seven or eight years, but several things have changed in the last three or four years," Gupta told Tom's Hardware Premium. "First, advanced packaging has matured significantly. There are now multiple 2.5D and 3D integration options in volume production. Customers are not limited to one packaging technology or supplier; there are options from foundries and OSATs, including TSMC, Intel, ASE, and Amkor. Second, die-to-die standards have arrived. In the past, most chiplet implementations relied on proprietary interfaces. UCIe is now moving into production deployments, including at hyperscalers, which makes heterogeneous integration much more practical. Third, supply-chain resilience has become critical. Customers increasingly want modular and potentially multi-source strategies rather than a single point of failure. Those factors have created an environment that did not exist four or five years ago."

AI accelerators will be among the primary applications to benefit from multi-chiplet design, as we have already learned from AMD and Nvidia.

"The AI accelerator market is on track to reach $604 billion by 2033, and custom silicon XPUs built for specific hyperscaler workloads are the fastest-growing segment," Gupta said. "At that scale, chiplet-based design is no longer optional, yet there is no pure-play chiplet company serving this market with a full portfolio. TYLsemi closes that gap with standards-based chiplets combined with UCIe-based die-to-die connectivity, XPU-aware design, packaging, and integration — giving customers a fast, proven path to AI-era silicon."

Chiplet economics

The vast majority of AI and HPC accelerators today feature large die sizes, in many cases approaching the size of a reticle. However, as modern process technologies are becoming more complex, foundries tend to increase their quotes for new nodes. A leading-edge wafer used to cost around $15,000 to process around five years ago, but today that price is around $30,000. As a result, large chips at a size close to the reticle limit implemented on a leading-edge node become an option for a select few chip designers who can afford it. For newcomers, multi-chiplet designs enabled by advanced packaging and standardized interconnects such as UCIe start to make a lot more sense.

"Once dies get into the 500 – 600 mm² range, the yield curve becomes increasingly difficult. Timing closure on a reticle-sized die is also challenging," Gupta explained. "I have worked on a reticle-sized accelerator, and getting from 99% to the final 1% can require disproportionately more engineering effort."

TYLsemi estimates that its chiplet approach could reduce total cost of ownership by 57% at a volume of 100,000 devices, from $350 million for a monolithic 700 mm² 3nm-class chip to $150 million for a design combining a 500 mm² 3nm-class compute die with four 100 mm² I/O chiplets built on an N-1 process. TYLsemi believes that the unit price of a monolithic chip would be $3,000, whereas the cost of an SiP would be around $600. The company attributes the saving to higher yields, reusable I/O silicon, lower IP licensing and engineering costs, and substantially lower per-unit silicon costs. However, the company stresses that the figures are illustrative estimates rather than actual manufacturing costs. Additionally, multi-chiplet designs can enable faster product refreshes compared to large monolithic dies as they are faster to develop and yield.

"Compute may move to 2nm or A14, while high-speed I/O can remain on 3nm, since I/O does not scale in the same way as logic," Gupta said. "Our power-delivery chiplets can use an even less advanced process. Customers therefore do not have to use the most expensive silicon real estate for every function. […] There is no single answer for every design. You have to determine the right disaggregation points based on the architecture, thermal requirements, package, and how multiple accelerators communicate. […] The exact partitioning will vary by application, but you still get a better total cost of ownership."

TYLsemi primarily targets AI infrastructure, so it generally envisions multi-chiplet designs to be used for AI accelerators, data-center CPUs, high-performance computing, networking and telecom silicon, and heterogeneous SoCs. However, TYLsemi has also ignored the fact that multi-chiplet designs are already widely used for consumer CPUs and GPUs.

Foundation chiplets

At the core of TYLsemi's proposition are its foundation chiplets, which are reusable building blocks intended to handle common non-compute functions in custom AI and infrastructure processors and are implemented using various process technologies from TSMC. The foundation chiplets include the following:

• TYL.IO — a family of connectivity chiplets that includes TYL. IO PCIe, a 32-lane PCIe 7.0/CXL chiplet connected to the compute die via UCIe; TYL.IO Scale, a 224G+ SerDes for ESUN/UALink scale-up connectivity; and TYL.IO EIC for co-packaged optics.
• TYL.Power — a 16nm in-package IVR chiplet with embedded passives, designed to provide power closer to compute dies and use closed-loop control and die telemetry to improve power delivery.
• TYL.Mem — a planned family of memory-connectivity chiplets. TYLsemi has not yet disclosed the architecture or specifications, though it is safe to assume they are talking about memory controllers and PHYs.

Not all of these chiplets will be available immediately, as the company has certain priorities amid limited resources.

"The first TYL.IO product disaggregates the PCIe functionality that would normally sit on a large server processor, it is a 32-lane PCIe Gen7/CXL chiplet connected to the host compute die using UCIe," Gupta explained. "The idea is that the CPU cores can move to 2nm, A14, or another leading-edge process, while the I/O chiplet remains on 3nm. The next product in the family will address scale-up connectivity between XPUs within a rack using high-speed SerDes. That device will be considerably larger, with around 72 lanes and approximately 14 TB/s of bandwidth. We also have an EIC roadmap for co-packaged optical connectivity. We expect samples of our first I/O product in the second half of 2027."

These chiplets can be used as standalone components or integrated with a customer's compute dies designed by the customer to TYLsemi through TYL.Forge, TYLsemi's end-to-end custom silicon platform.

TYL.Forge

TYL.Forge is arguably one of the key enablers of TYLsemi's business, as the program is aimed at companies that have their own compute architecture or even a compute die, but cannot build their own SiP or manage the entire semiconductor supply chain.

"There are larger custom silicon companies in the market, but many of them focus on a relatively small number of customers that can generate billions of dollars in annual business," Gupta explained. "We see an opportunity among emerging AI companies and system companies that need advanced custom silicon but also need a partner capable of taking responsibility for the entire implementation and supply chain."

Such clients can provide their proprietary compute RTL, while TYLsemi handles physical implementation and integrates the resulting compute die with its pre-validated connectivity, power, and eventually memory chiplets. The company then manages tape-out, packaging, assembly, testing, qualification, and high-volume production.

"For example, a customer building a large accelerator can bring us its matrix-multiplication engine," Gupta said. "We can implement the custom compute die and integrate it with our chiplets, so the customer does not have to reinvent the I/O and other common functions. This reduces risk and time to market."

In fact, TYL.Forge appears flexible about where the customer enters the development process. TYLsemi describes the platform as covering everything from architecture and front-end design through implementation, tape-out, assembly, qualification, and production. So instead of RTL, customers can come to TYLsemi with an architecture/concept, and then the company will help develop and implement the silicon. Nonetheless, TYLsemi does not intend to invent the customer's core compute architecture itself. In addition, customers can bring in an existing compute die, which TYLsemi can combine with its chiplets, package, test, and bring to production.

"We are also talking with companies developing Arm- and RISC-V-based server processors," Gupta said. "They can develop the architecture, while we implement the rest of the silicon and bring the product to production. That gives customers an economic and engineering advantage because they do not need to build teams for every part of the chip."

The key advantage of TYL.Forge is the reuse of pre-validated components. Instead of developing common functions such as PCIe connectivity and power delivery for every new processor, customers can use TYLsemi's pre-validated foundation chiplets and focus engineering resources on differentiated compute architectures, software, and system design. In theory, TYLsemi could integrate third-party chiplets (not from a customer, but from a third-party chiplet provider). Still, the company's focus remains on offering its own pre-validated chiplets and custom silicon with SiPs it builds.

"Potentially, [we could integrate third-party UCIe chiplets into a TYLsemi-based system], UCIe has done a very good job defining the electrical interface, but the ecosystem is still maturing at the protocol level," Gupta explained. "In some cases, if we provide a chiplet to a customer, we may also need to provide or enable the UCIe IP on the other side of the connection. We are committed to UCIe and industry standards because standardization ultimately wins. […] We can consider customization for a large strategic customer or hyperscaler, but we do not want those projects to derail our standard product roadmap. […] Even when customers buy our standalone chiplets, I expect many of them will ask us to handle packaging and testing because heterogeneous integration and supply-chain management are difficult "

TYLsemi estimates that its approach can cut development time and cost by up to 50% compared with traditional custom silicon programs. In the best-case scenario, TYLsemi envisions that the development cycle can shrink considerably compared to today's cycles that can be two, three, or more years long. According to TYLsemi, once a customer provides sufficiently mature RTL or a netlist, the company can take a custom compute die to tape-out in around six to nine months or so, which includes fabrication, assembly, testing, and qualification.

"If a customer provides mature final RTL or a netlist and uses our standardized I/O chiplet, we believe we can take the custom compute die to tape-out in approximately six months in some cases," Gupta explained. "More generally, our target is six to nine months from a mature design to tape-out. The architecture and front-end phase is more customer-dependent. For a first-generation product, that can take around six months; for a more mature second- or third-generation design, it could be closer to three months, and some of that work can overlap with implementation. After tape-out, fabrication can take roughly four to five months depending on the process, followed by perhaps another two months for assembly, testing, and qualification. If the architecture is already mature, it may therefore be possible to reach production samples in about a year."

Still, the company stresses that architecture development and implementation typically include feedback loops, which greatly slow the development process. This is why the company provides the relatively conservative '50%' figure.

TYL.IO and TYL.Power samples will be available to qualified customers in 2027, in partnership with TSMC, and the company is looking forward to designing processors for its clients in time for them to reach the market in 2029 – 2030.

Speaking of TSMC, TYLsemi will initially only offer designs and services adhered to the TSMC ecosystem, though eventually it may offer other options for packaging technologies, such as Intel's EMIB and Foveros, or Amkor's packaging methods.

"We are initially focused on the TSMC ecosystem, but we also intend to explore other advanced-packaging supply chains," Gupta said. "We do not want to limit ourselves to one packaging option. Over time, that could include other OSATs and packaging technologies. […] That could include Intel, ASE, Amkor, or others. Amkor, for example, is building significant packaging capacity in Arizona."

Intel will design, package, and fabricate Fortinet's sixth-generation Security Processor (SP6) on its Intel 4 node, the companies announced on July 21, giving the process its first named external foundry customer, roughly three years after it entered production. Intel told Tom's Hardware the agreement reflects "the strategy Intel outlined for Intel 4 several years ago," including support for custom networking ASIC workloads. Intel's own record from those years reads differently, however, with the company's 2021 roadmap having scoped Intel 4 to two internal products. And through 2022, it told engineers and investors that Intel 3, not Intel 4, would be its first process offered to foundry customers.

Intel 4's record

Intel's Accelerated announcement back in July 2021 said that Intel 4 would reach production readiness in the second half of 2022 for products shipping in 2023, naming "Meteor Lake for client and Granite Rapids for the data center." The release and its accompanying fact sheet, however, contained no reference to foundry customers, networking, or custom ASICs on the node.

At VLSI 2022, Intel disclosed that it wasn't building a high-density library for Intel 4 and that Intel 3 would be the first new node offered through what was then Intel Foundry Services. A 2024 post on Intel's own foundry blog describes Intel 3 as "Intel Foundry's first leading-edge process node," and Intel's fiscal year 2024 annual report listed the processes available to external customers as 18A, Intel 3, Intel 7, Intel 16, and a 12nm node co-developed with UMC. Intel 4 appears nowhere on that list.

Ericsson's RAN Compute processors, announced in November 2023, were built on Intel 4, so Fortinet's part won't be the first third-party silicon to come off the node. That work grew out of a bespoke Intel-Ericsson collaboration, though, and Ericsson's formal foundry agreement with Intel, announced in July 2023, covered 18A. Fortinet is the first named customer buying Intel 4 as a foundry service, and the first cybersecurity vendor on any Intel node. The Ericsson engagement is also the closest thing in the public record to networking silicon on Intel 4, two years after the strategy Intel now says it outlined for the node.

Fab 34 economics

Intel 4 entered high-volume manufacturing at Fab 34 in Leixlip, Ireland, in September 2023, producing the compute tile for Meteor Lake-based Core Ultra chips, and shares the fab with Intel 3. Intel sold a 49% stake in the facility to Apollo-managed funds for $11.2 billion in June 2024, then bought it back in April 2026 for $14.2 billion, funded with $7.7 billion in cash and $6.5 billion in new debt. That buyback returned 100% of Fab 34's wafer economics to Intel at a premium of roughly 27%, and it only pays off if the fab's EUV capacity stays loaded.

Meteor Lake is aging out of Intel's lineup as 18A-based Panther Lake ramps through 2026, which leaves open the question of what fills Intel 4 capacity next. A multi-generation firewall ASIC program is a reasonable answer with mature yields, a customer that values supply stability over bleeding-edge density, and a part Intel described as tailored for cost-sensitive applications. Intel said in April that yields were improving across Intel 4, Intel 3, and 18A.

Fortinet's supply chain

Fortinet's 2025 annual report names Renesas and Toshiba America as the contract manufacturers for its ASICs, utilizing foundries in Taiwan and Japan operated either by TSMC or by the contract manufacturers themselves. The current SP5, a monolithic 7nm Arm-based SoC announced in February 2023, sits in that supply chain, so SP6 on Intel 4 moves Fortinet's next flagship security processor out of a TSMC-linked flow and into Intel's. The disaggregated design language in the announcement points to a chiplet-based part, a first for Fortinet's SP line.

Fortinet re-engineered three FortiGate models in 2022, the 70F, 600F, and 3700F, to accept alternative components during the chip shortage, and CMO John Maddison told SDxCentral at the time that the company wouldn't wait for parts to arrive in 2023. The "resilient and diversified" supply chain used in the SP6 press tracks back to that experience. Ken Xie called Fortinet "the #1 firewall leader with a 55% unit market share" in the company's 2025 results in February, with approximately six million FortiGates deployed, so there’s real, substantial volume here even if the parts are relatively inexpensive.

Intel Foundry reported $307 million in external revenue for 2025, up from $159 million the year before, against total foundry revenue of $17.8 billion and an operating loss of $10.3 billion. External revenue in Q1 2026 was $174 million. Fortinet's hardware business runs at roughly 30% of its revenue, and, per analysis from ServeTheHome, SP6 is ultimately a component of a portion of an annual hardware stream around $2 billion, so the deal won't move Intel's foundry line materially, even at full production.

CEO Lip-Bu Tan told CNBC in May that he expected commitments from multiple foundry customers in the second half of 2026, and Intel told investors in January that two prospective customers were evaluating 14A test chips. The SP6 announcement comes inside Tan's stated window, and it finally gives Intel something its foundry marketing has lacked in a named customer with shipping volume on a node with mature yields. Meanwhile, an 18A or 14A commitment from a major external customer is still missing, and Fortinet's cost-sensitive parts on a 2023 node don't substitute for one.

Intel and Fortinet have announced a strategic collaboration to develop the Fortinet Security Processor 6 (SP6), the next generation of the custom silicon behind Fortinet's FortiGate firewalls. Intel will contribute chip design, advanced packaging, and manufacturing to the SP6 program, making Fortinet the first cybersecurity vendor named as an Intel silicon customer.

Intel told Tom's Hardware that SP6 will be built on Intel 4, the EUV process the company has so far used only for its own products, making Fortinet both the first cybersecurity vendor named as an Intel silicon customer and the first named external customer for the node. However, the announcement specifies no production timeline, and it comes in the same Q3-Q4 2026 window in which CEO Lip-Bu Tan said the company expects commitments from multiple foundry customers.

SP6 will draw on what the companies described as Intel's expertise in disaggregated semiconductor design and advanced packaging tailored for both AI-enabled and cost-sensitive applications.

That points to a chiplet-based part, which would be a departure from the current SP5, a monolithic 7nm Arm-based SoC that Fortinet launched in 2023 for its entry-level and mid-range FortiGate appliances. The deal will help Fortinet "accelerate and strengthen our ASIC strategy," said Ken Xie, founder, chairman, and CEO of Fortinet, in the announcement.

Intel 4 was the company's first process node to use EUV lithography and entered high-volume manufacturing at Fab 34 in Ireland in September 2023, where it produces the compute tile for Meteor Lake-based Core Ultra chips. The node didn't appear among the processes Intel listed for external foundry customers in its fiscal year 2024 annual report, which named 18A, Intel 3, Intel 7, Intel 16, and a 12nm process co-developed with UMC. Intel said the SP6 work reflects plans it laid out for Intel 4 several years ago, including support for custom networking ASIC workloads.

Microsoft agreed in early 2024 to build an unnamed custom chip on Intel's 1.8nm-class 18A node, a deal that reportedly covers a next-generation Maia AI processor. That announcement followed a similar pattern, with no product details, node variant, or timeline at signing and specifics emerging over the following 18 months.

Intel CFO David Zinsner said in March that the company was fielding inbound interest in 18A-P from prospective foundry customers as yields improved, and Tan told CNBC in May that foundry commitments were expected in the second half of 2026.

Fortinet does bring real volume, though, if not marquee volume. IDC ranked Fortinet first in firewall appliances shipped as of early 2023, with a 48% unit share. The company ships its own ASICs across its entry-level and high-end FortiGate ranges, and SP6 extends a silicon program now in its sixth generation.

Neither company committed to work beyond SP6, though the release described the agreement as a starting point, with further collaboration on chip technology and manufacturing under discussion.

TSMC intends to raise base quotes on advanced chip production services by up to 10%, according to Nikkei, which cites people with knowledge of the matter. The price hike reflects increased demand for sophisticated processors by the AI sector, raising costs of tools and materials, as well as amplified investments in new production capacities.

For advanced process technologies — which TSMC considers 7nm-class and below — TSMC plans to raise baseline prices by 5% to 10%, depending on the particular production node and customer, the report claims. Furthermore, customers that need additional HPC chip capacity beyond their original volume requirements will reportedly have to pay another 10% to 15% premium on top of the standard increase, which means that some services will get a price hike of around 25%, if the report is accurate.

TSMC also intends to increase prices for mature manufacturing technologies, including its 12nm, 16nm, and 28nm-class nodes as well as other legacy fabrication technologies, the report claims. Increases could reach 10%, although certain nodes will reportedly see smaller adjustments, according to Nikkei.

Advanced technologies generated around 77% of the foundry's revenue in Q2 2026, whereas mature nodes accounted for 23%, which essentially means that TSMC is hiking prices on all of its services.

The company reportedly began discussing the new pricing with customers around June and completed negotiations in July. Rather than introducing higher rates immediately, TSMC opted to implement them from the beginning of 2027 to give clients like Apple, AMD, Nvidia, and MediaTek additional time to accommodate the changes and adjust their prices accordingly.

Since TSMC produces the lion's share of advanced processors for AI, HPC, networking, and smartphone applications, its price hikes will inevitably create a ripple effect in the industry and will make almost all electronics more expensive.

TSMC is not alone in raising prices these days. Vanguard International Semiconductor has also raised prices, while UMC began implementing increases in July. Also, memory makers have increased prices significantly, making TSMC management jealous. Intel also recently increased prices of its client and data center CPUs, citing market demand.

"I am really jealous about memory companies' 86% gross margin," said C.C. Wei, chief executive of TSMC, during the company's earnings call with financial analysts and investors. "86% [margin at memory makers] – 68% [margin at TSMC], I will be happy about that."

TSMC rarely comments on its prices to a large degree because they vary based on volumes and relationship with a particular client. Nonetheless, the head of the company stressed that the company has no intentions to increase prices suddenly or dramatically.

"So we do not suddenly increase our price by which I like to have 4x or 5x," Wei said. "You cannot survive for that kind of... for your customer to survive for that kind of price increase. So we earn our value, and we make sure that our profit, our gross margin, is enough for our long-term sustaining expansion, that is to the benefit of my customers and TSMC also, that is our philosophy."

An analysis of Huawei's Kirin 9030 system-on-chip (SoC) for smartphones conducted by SemiAnalysis revealed that SMIC's third-generation 7nm-class fabrication technology (N+3) has smaller metal pitch than Intel's 18A fabrication technology and that China's leading foundry has managed to achieve transistor density on par with manufacturing process that rely on EUV lithography. But does this make SMIC's N+3 node as competitive as Intel's 18A or TSMC's N2 and N3? Not really.

SemiAnalysis' teardown indicates that SMIC's N+3 fabrication process supports a minimum metal pitch of 32.5nm, which is nominally tighter than the approximately 36nm pitch used for many high-performance cells in Intel's Panther Lake CPU, even though 18A can support approximately 32nm metal pitches. The video from SemiAnalysis and High Yield does not reveal other important characteristics of SMIC's N+3, such as contacted gate pitch (CGP), standard cell height (tracks or nm), or fin pitch, so we cannot make direct comparison of this node to Intel's or TSMC's technologies. What it does reveal is estimated transistor density of around 113.4 million transistors per square millimeter (Mtr/mm²), which is even higher than transistor density of TSMC's N6, 107.7 Mtr/mm².

TSMC's N6 uses multiple EUV layers, so achieving higher transistor density without using EUV lithography is an indisputable technological achievement of SMIC. The foundry achieves this density by using DUV multi-patterning, including self-aligned quadruple patterning on the tightest layers, and extensive design-technology co-optimization (DTCO). In addition, SemiAnalysis believes that SMIC used techniques like reduced fin counts, placing contacts directly over active gates, and tightening cell isolation. Such methods allow for increased transistor density, but at the cost of increased process complexity, cost, yield risks, and design constraints.

Meanwhile, transistor density does not equal overall process competitiveness. Despite its compact layout, the Kirin 9030 reportedly delivers performance comparable to flagship application processors from roughly three years ago and has a substantial energy-efficiency disadvantage compared with modern Apple, Qualcomm, MediaTek, and Samsung designs. Huawei's highest performing CPU core is characterized as roughly Cortex-X2-class in IPC, while Apple's much smaller efficiency cores reportedly outperform it in integer workloads and consume considerably less power.

Given the fact that Kirin 9030 is neither a performance nor efficiency champion, the provocative comparison with Intel 18A is not exactly justified. Although SMIC N+3 has 32.5nm minimum metal pitch that is nominally tighter than the approximately 36nm pitch used in Panther Lake, 18A offers both higher transistor density and considerably higher performance efficiency. In addition, 18A uses gate-all-around transistors and backside power delivery, which make it particularly suitable both for mobile SoCs and for data center applications.

SemiAnalysis concluded that while export restrictions have slowed China's technological progress, progress is still being made. SMIC could potentially continue increasing density by tightening upper and lower metal layers, shorter standard cells, smaller gate pitches, and eventually backside power delivery. If the company continues scaling, N+4 could approach TSMC N5-class density, while N+5 with backside power might reach Intel 18A-class density, according to SemiAnalysis. Still, transistor density alone does not necessarily bring substantial improvements of performance or power efficiency.

ASML is reportedly considering increasing prices of its existing Low-NA EUV lithography tools. The potential price hikes have already upset TSMC, ASML's largest client, reports The Information. But can ASML rapidly raise the prices of the tools it sells to chipmakers? Price adjustments are not going to happen overnight, said Roger Dassen, CFO of ASML, said during the company's quarterly earnings call. But there is one thing to remember about ASML's price hikes: They are going to affect the entire semiconductor industry.

"When it comes to Low-NA [EUV tools] pricing, of course, you know that we keep on increasing the productivity of the Low-NA tool, [which] gives us a pretty strong runway for potential price improvements going forward," said Roger Dassen, chief financial officer of ASML, during the company's quarterly earnings call. […] Given the long order lead times that we have, that does not translate into pricing effects tomorrow."

ASML just reported record results for its second quarter of 2026, with total net sales of €9.326 billion ($10.67 billion) and net income of €2.918 billion ($3.338 billion). The company now expects net sales between €43 billion ($49.2 billion) and €45 billion ($51.5 billion) in 2026, a range that comfortably exceeds its own guidance and the expectations of industry analysts.

More complex, more expensive

ASML has generally increased the average selling price of EUV scanners over successive generations as they increase in complexity and productivity, so the idea of higher pricing of EUV tools is not particularly new. The company calls the concept of 'value-based pricing' and gradually increases its average selling price (ASP) based on the value that its tools provide to its operators.

Early ASML Twinscan NXE systems were commonly discussed in the roughly €100 million–€120 million ($115 million-$137 million) range, while later production models such as the NXE:3400C and NXE:3600D moved toward roughly €140 million–€170 million ($160 million-$195 million). The latest NXE:3800E is climbing even further. High-NA EXE systems represent another major step up, at more than €350 million ($400 million) per machine according to industry reports.

Meanwhile, both productivity and performance of ASML's Twinscan NXE systems have been steadily increasing: While the NXE:3400C and NXE:3600D can process 160 – 170 wafers per hour (WPH) and feature a matched machine overlay (MMO) of ≤ 1.1nm, the NXE:3800E and NXE:3800F increase productivity to 220 WPH and 260 WPH, respectively, while increasing MMO to 0.9nm. With the NXE:4200G and NXE:4200H (which will likely feature an all-new light source), we are looking at productivity beyond 300 WPH and MMO of ≤0.8nm - ≤0.7nm.

ASML makes no secret that more advanced EUV lithography tools carry a higher average selling price than their predecessors.

"You should also recognize that the tool mix that we are going to ship next year is a different tool mix from the tool mix that we shipped this year," Dassen said. "When it comes to EUV in particular, the tool mix that we are going to ship next year will be EXE: 3800E and EXE:3800F [tools], while this year it is a combination of EXE:3600D and EXE:3800E [systems]."

ASML's value-based pricing logic is essentially: if an upgraded Low-NA machine processes more wafers and therefore generates more economic value for a fab, ASML can capture part of that additional value through a higher system price. This is what ASML has been doing for some time. That said, the important nuance in Dassen's statement is that he appears to be discussing further price increases for Low-NA EUV tools.

ASML's EUV dominance

ASML is the only supplier of EUV lithography systems on the planet; every single chipmaker must comeq to the company to get one of these scanners, so the company sells these machines years in advance. This year, the company expects to have the capacity to build 65 EUV tools and intends to increase capacity by 30% next year to around 84 – 85 EUV systems. In 2028, ASML plans to produce 110 EUV scanners.

"For 2027, we are now close to being fully covered with orders for Low-NA EUV, and we are planning to increase our Low-NA EUV capacity by around 30%," Dassen said. "Looking ahead to 2028, we have already received a significant number of Low-NA EUV orders. Strong demand forecasts from our customers have led us to investigate a further 30% capacity increase for that year."

Given the strong demand for ASML's tools, the company is in a position to increase prices, at least according to the rules of the free market. As always, comments on pricing are made in opaque corporate language, so one can make a lot out of it, or nothing at all.

"Clearly, the environment that we live in today, with the [substantial] value that our products bring to customer, of course, gives us flexibility on pricing, more so than what you would have seen in the past," Dassen said. "Of course, we are executing on that as well."

However, it is not that easy for ASML to just hike prices overnight, and there are important things to note. When ASML gets an order, and that order enters the reported backlog (which totaled €38.8 billion as of late Q4 2025), it already carries a sales value (with a possibility of inflation adjustments), so there is necessarily an agreed or otherwise contractually determined price basis attached to it.

This makes Dassen's comments particularly interesting. If ASML is already close to fully booked for 2027 Low-NA EUV capacity and has received a significant number of Low-NA orders for 2028, then much of that capacity is already represented by customer orders with associated sales values. As a result, unless ASML can subsequently renegotiate the fixed price before delivery, it cannot hike prices on tools that it intends to ship in 2027 and part of 2028. This essentially means that ASML could only book new orders that it intends to ship in 2028 (presumably for the second half of the year) and onwards at higher prices. However, ASML has not confirmed this directly.

ASML's next-generation NXE:4200G is on track to arrive in 2029. That scanner is poised to be more expensive than the NXE:3800E/NXE:3800F anyway, so the company's Low-NA EUV ASPs will inevitably get a bump that year. Of course, ASML is set to continue shipping its EXE:3800-series after 2029. The big question is whether the company will adjust the prices of these units that will ship starting in 2028 or not.

We do not know the answer to this question, unfortunately. But during the conference call, ASML's financial chief implied that the company would like to get 'rewards' from its customers not only for improved productivity, but also for other improvements of its tools, which means that ASML will stick to its value-based pricing model, but may adjust the pricing for the EXE:3800 model.

"We have always been able to show customers not just productivity upgrades, but also the value from better imaging, the value of better overlay, etc.," Dassen said. [But] you got this very strong correlation between throughput improvements and ASP. That is just the way things panned out, which, put in another way, customers were paying for the productivity upgrade, and the value that we gave them for free was the value associated with, let us say, overlay improvement, imaging quality […]. In the current environment, with the value that we bring, we are also having conversations with customers on how we get rewarded for that additional value."

TSMC's backlash against the price hikes

As ASML will likely not be able to hike prices on EUV tools that have already been pre-ordered and which are set to be delivered in the course of the next two years, existing chipmakers like TSMC will not feel the effects of the price increase at least over the next 24 months. Of course, if ASML adjusts prices of its EXE:3800-series scanners due to ship in 2028 – 2028, nobody is going to be happy. And yet, TSMC seems to be so upset that the sentiment has made it to the press. There are several reasons behind the reported backlash, but the major one seems to be strategic.

For years, TSMC has said that ASML's High-NA EUV lithography tools were too expensive, and that the company's engineers can continue innovating using Low-NA EUV systems. Also, adopting all-new High-NA EUV scanners will not be an easy undertaking in general, as the transition also requires new photoresists, photomasks, pellicles, metrology equipment, design rules, computational lithography flows, and numerous other supporting technologies and process innovations.

TSMC's mid-term expansion strategy, as well as a leading-edge roadmap through 2030, has been built around extracting more performance and resolution from conventional Low-NA EUV scanners using techniques such as improved masks, computational lithography, and multi-patterning where necessary. If ASML proceeds with its price hikes for Low-NA EUV tools, this can seriously hit one of the key economic foundations of TSMC's strategy.

TSMC's intention to avoid the usage of High-NA EUV scanners until at least 10A-class (1nm) process technology has an important economic advantage. A High-NA EUV system costs more than €350 million, whereas Low-NA systems are substantially cheaper. As a result, TSMC can choose its own trade-off between additional process steps and buying much more expensive lithography equipment. In contrast, Intel, which is set to adopt High-NA EUV lithography for its 14A fabrication process, has much less flexibility: If a particular process technology is designed around High-NA EUV layers, these extremely expensive scanners, along with other ingredients, become part of the manufacturing flow and affect wafer pricing.

One of the reasons TSMC would like to extend usage of Low-NA EUV systems is that it already has the world's largest install base of these tools, as well as mature processes, established flows, and plenty of innovations that enable it to stay ahead of the competition. However, if ASML systematically ties productivity improvements of each new Low-NA EUV generation to higher prices, the cost advantage of staying with Low-NA EUV gradually narrows, which greatly undermines TSMC's strategy.

There are several other reasons for TSMC to be particularly unhappy with ASML. Firstly, TSMC needs an enormous number of tools for its upcoming fabs in Taiwan, the U.S., and Japan. All of the company's leading-edge process technologies through 2029 rely on Low-NA EUV production tools. Even a relatively modest percentage increase applied to dozens of Low-NA EUV scanners can add billions of dollars to the company's capital expenditure (CapEx).

Secondly, ASML is essentially arguing that it deserves a share of its customers' improved economics. From TSMC's perspective, potential price adjustments are very different from charging more because a new scanner is materially more expensive to manufacture. ASML is effectively saying: 'Your fabs are more profitable, and our tools are more productive, therefore we want a piece of that.'

Thirdly, pricing negotiated today determines pricing and TSMC's economy for years to come. Now that ASML is sold out for 2027 and a significant part of 2028, it is negotiating terms for the second half of 2028 and onwards. Accepting a substantial increase now will inevitably carry through to dozens or hundreds of future Low-NA EUV systems, which directly affects TSMC's economics.

TSMC's A14 (1.4nm-class) fabrication process has made rapid progress in the last three months and is well ahead of N2 at the same stage of development, according to the company's update provided at its earnings call this week. The technology also faces strong customer interest and engagement across both smartphone and AI/HPC applications.

"A14 technology development is on track and progressing well. Internal product-like vehicle demonstrated close to 90% device performance and close to 90% 256Mb SRAM yield," said C.C. Wei, chief executive of TSMC, during the earnings call with analysts and investors.

A14 — which is expected to enter mass production in 2H 2028 — is making rapid progress in terms of performance and yield improvements. This April, the company disclosed that the production node achieved >85% target transistor performance and >80% 256Mb SRAM yield. Roughly three months later, both figures are approaching 90%, which suggests a gain of around 5% in device performance and nearly 10% in SRAM yield.

For comparison, TSMC's N2 demonstrated more than 80% of its target device performance and over 50% yield on a 256Mb SRAM test chip in April 2023. By April 2024, the process had advanced to more than 90% of its target device performance and over 80% SRAM yield. While development trajectories are not directly comparable, the figures suggest that A14 is maturing considerably faster than N2 did at a similar stage of development.

The very rapid progress of A14 compared to the relatively slow maturation of N2 at similar stages of development can probably be attributed, at least in part, to TSMC's growing experience with gate-all-around (GAA) nanosheet transistors. Back in 2023, the company barely had enough experience with the production of gate-all-around (GAA) nanosheet transistors, as N2 is its first process technology to adopt such a structure. By contrast, A14 relies on TSMC's 2nd Generation of GAA devices, so it can probably benefit from the transistor-design improvements, process refinements, and manufacturing expertise accumulated during the development and ramp of N2.

It appears TSMC has likely eliminated many of the yield limiters with A14 and N2, though keep in mind that a high 256Mb SRAM yield merely indicates low enough defect density and good process uniformity across a highly repetitive test structure, but it is not directly representative of functional or parametric yield of a commercial processor.

Nonetheless, the close to 90% device performance and close to 90% 256Mb SRAM yield about 2.5 years away from expected mass production start put TSMC's A14 progress well ahead of N2. Such progress can potentially enable TSMC to start high-volume manufacturing (HVM) using A14 ahead of schedule, provided that customer designs are ready, or initiate HVM with better-than-usual functional and parametric yields.

Speaking of customer design readiness, Wei indicated that clients strive to tape-out their A14 designs ahead of schedule, which is a good sign. It is also interesting to note that despite the fact that A14 lacks Super Power Rail backside power delivery (A12 will gain SPR in 2H 2019), it is set to be adopted not only by client processors, but also by AI/HPC applications.

"We are observing a strong level of customer interest and engagement on both smartphone and HPC/AI applications, and customer new tap-out activity is ongoing and ahead of schedule," Wei said.

A14 is TSMC's next-generation process technology that combines the company's 2nd Generation GAA nanosheet transistors with a new standard-cell architecture to improve performance, power efficiency, and transistor density. Compared with N2, TSMC expects A14 to deliver a 10% – 15% performance uplift at the same power and transistor count, or reduce power consumption by 25%–30% at the same frequency and complexity. The node is also projected to increase transistor density by around 20% for mixed designs and by 23% for logic.

As the semiconductor industry increasingly relies on ASML's EUV lithography tools both for logic and memory production, ASML is considering increasing prices of these systems as they deliver greater productivity and better value for its customers. However, with 2027 production nearly sold out and substantial 2028 orders already booked, meaningful increases may primarily affect systems delivered from late 2028 onward. Nonetheless, the idea has already angered TSMC, ASML's largest client, reports The Information.

"When it comes to Low-NA [EUV tools] pricing, of course, you know that we keep on increasing the productivity of the Low-NA tool, [which] gives us a pretty strong runway for potential price improvements going forward," said Roger Dassen, chief financial officer of ASML, during the company's quarterly earnings call. "Given the long order lead times that we have, that... doesn't translate into pricing effects tomorrow."

Value-based pricing set to persist… in a new way

ASML has long followed what it calls value-based pricing and gradually increased its quotes based on output, patterning costs, power consumption, and other benefits its new tools offer to clients.

Usually, this involved increasing prices once in a while. For example, if early Twinscan NXE Low-NA EUV systems cost roughly €100 million – €120 million ($115 million–$137 million), the more advanced are priced starting at €170 million ($195 million. It is still well below rumored quotes for High-NA EXE scanners that exceed €350 million ($400 million). At the same time, Low-NA productivity has risen from 160–170 wafers per hour (WPH) and ≤1.1nm matched-machine overlay to 220 WPH/260 WPH with NXE:3800E/NXE:3800F at 0.9nm. Future NXE:4200G/NXE:4200H systems are expected to exceed 300 WPH and improve overlay to ≤0.8nm–≤0.7nm.

"Clearly, the environment that we live in today, with the value that our products bring to customer — it's substantial— of course, gives us flexibility on pricing, more so than what you would have seen in the past," Dassen said. "Of course, we are executing on that as well."

However, later during the call, Dassen emphasized that ASML intends to maintain its value-based approach even in the current environment of high demand and limited supply in the semiconductor world. Yet, he stressed that from now on, ASML might want to charge for things beyond just productivity.

"We have always been able to show customers not just productivity upgrades, but also the value from better imaging, the value of better overlay, etc.," Dassen said. "[But] you got this very strong correlation between throughput improvements and ASP. That is just the way things panned out," he said, suggesting that ASML shares value with its clients.

No price hikes in the short term

ASML will be unable to hike prices of Low-NA EUV systems for another couple of years. Since orders that are in ASML's backlog already carry a sales value, subject to inflation adjustments, prices for much of the 2027 and early 2028 output may already be contractually determined. Unless existing contracts can be renegotiated, higher pricing could therefore primarily apply to 2028 shipments and beyond, or for new orders that somehow get squeezed in in 2027. The NXE:4200G, due in 2029, should naturally lift average selling prices anyway as it gets major performance improvements.

TSMC is upset

For TSMC, however, the issue is strategic. The foundry's leading-edge roadmap through 2030 relies on extending Low-NA EUV with better masks, computational lithography, and multipatterning. Until then, TSMC's strategy has always been avoiding High-NA EUV until at least its 10A-class (1 nm-class) technology. If ASML hikes prices of its future Low-NA EUV lithography systems, it will likely affect all of TSMC's plans for the next several years.

TSMC already operates the world's largest Low-NA EUV fleet and needs many more scanners for fabs in Taiwan, the U.S., and Japan as it executes its global expansion strategy. Consequently, even modest increases beyond TSMC's projections could add billions to capital spending, reduce the economic advantage of postponing High-NA, and ultimately raise its manufacturing costs. Moreover, accepting higher prices now could establish the baseline for dozens or hundreds of future systems, which will allow ASML to capture a larger share of the economic value created by increasingly productive lithography equipment.

Can this force TSMC to transition to High-NA EUV tools earlier than planned? Moving to High-NA EUV requires not only €350-million-plus scanners but also new resists, masks, pellicles, metrology, design rules, and computational lithography flows, which are likely not ready at TSMC.

Tower Semiconductor has announced a dual-track expansion of its 300mm silicon photonics, silicon germanium, and advanced packaging operations in Japan, committing up to $3 billion net of grants with backing from the country's Ministry of Economy, Trade and Industry (METI). Alongside the announcement, the Israeli specialty foundry raised its 2028 business model to approximately $3.6 billion in revenue and $1.2 billion in net profit, and it says those targets rest entirely on the first of the plan's two tracks: reviving the shuttered Arai fab it inherited from Panasonic and maximizing its running 300mm fab in Uozu, Toyama Prefecture.

Two tracks, one committed

Track One converts the former Arai facility, designated Fab 6, into a 300mm silicon photonics and advanced optical packaging plant while expanding output at Fab 7 in Uozu, with full production readiness expected during the fourth quarter of 2027. The Arai plant ceased operations in July 2022 because it exclusively served Nuvoton Technology Corporation Japan (NTCJ) rather than Tower's foundry customers, according to Tower's SEC filings, leaving an intact fab shell sitting idle for four years.

Track Two calls for constructing a new 300mm fab adjacent to Fab 7, which Tower says would deliver a multi-fold increase in silicon photonics and silicon germanium capacity and become "highly accretive beginning in 2029." The company hasn't signed definitive agreements for it, however, and none of the new 2028 targets depend on it.

A restructuring of the TPSCo joint venture, announced in March 2026, cleared the way for all this. Tower entered Japan in 2014 by buying 51% of Panasonic's three-fab semiconductor manufacturing operation, and Panasonic sold its remaining stake to Nuvoton in 2020. Under the March agreement, Tower takes full ownership of the 300mm Fab 7, while NTCJ absorbs the 200mm operations and pays Tower $25 million, with closing expected on April 1, 2027. Sole ownership of Fab 7 removed the joint-venture structure that would have complicated a $3 billion buildout.

Tower CEO Russell Ellwanger contrasted the approach with greenfield construction and fab acquisitions, which he said typically require years of process development, customer qualification, and financial stabilization while ramping from zero revenue against high fixed costs. Reusing a dormant building next to a qualified, cash-generating photonics fab is why Tower can achieve production readiness roughly 18 months ahead; Rapidus, by comparison, broke ground on its greenfield Chitose site in September 2023 and doesn't expect mass production until 2027.

29% increase in revenue

Tower reported $1.566 billion in revenue and $220 million in net profit for 2025, up from $1.436 billion and $208 million in 2024. The new 2028 model more than doubles 2025 revenue and implies a net margin of around 33%, against roughly 14% today. Measured against the prior 2028 model of $2.8 billion in revenue and $750 million in net profit, which Tower reaffirmed in its Q1 2026 report in May, the new targets add 29% to revenue and 60% to net profit.

Silicon photonics revenue is doing most of the heavy lifting, with Ellwanger telling analysts on the company's Q4 2025 earnings call in February that silicon photonics revenue reached $228 million in 2025, up from $106 million in 2024, and hit a $380 million annualized run rate in the fourth quarter, a figure he noted includes some non-wafer engineering revenue. In May, Tower disclosed $1.3 billion in contracted silicon photonics revenue for 2027 from its largest customers, backed by $290 million in prepayments already collected.

Tower's photonics customer roster includes Innolight, which builds 400G, 800G, and 1.6T optical transceivers on Tower's PH18 platform family, and Marvell, which said in June it had shipped more than five million coherent photonic ICs manufactured with Tower. The company claims more than 50 active silicon photonics customers and supplies foundry capacity for 200 Gb/s-per-lane devices used in 1.6T transceivers.

Tower's forward-looking disclosures flag construction delays, equipment lead times, permitting, and METI grant covenants that "may result in loss of a portion or all of the grant funds." The implied margin expansion also assumes sustained AI and data center optics demand from a concentrated group of very large customers through 2028, a dependency Tower acknowledges.

Tower’s position in the photonics foundry race

GlobalFoundries paid $453 million in cash for Singapore's Advanced Micro Foundry in November 2025, according to its annual report, a deal the company said made it one of the largest silicon photonics manufacturers. TSMC's COUPE co-packaged optics platform is tracking Nvidia's optical interconnect roadmap, with 1.6 Tb/s optical engines arriving in 2026 products.

Tower occupies a different lane from TSMC, as a merchant foundry serving dozens of transceiver makers and chip designers, rather than a packaging platform aligned with one customer's rack-scale plans. GlobalFoundries competes with Tower far more directly, and the two are also in court, with GlobalFoundries pursuing patent infringement claims against Tower.

MarketsandMarkets estimates the silicon photonics market at $2.65 billion in 2025, growing to $9.65 billion by 2030 at a 29.5% compound annual growth rate. Demand for optical data movement in AI clusters underpins those forecasts, as interconnects shift from copper to light at 800G and 1.6T speeds.

METI's support for Tower joins a Japanese subsidy program that has committed up to ¥1.2 trillion to TSMC's JASM fabs in Kumamoto, roughly ¥536 billion to Micron's Hiroshima operations, and around ¥2.9 trillion in planned funding for Rapidus. Tower's award appears to be the program's first at this scale for a dedicated silicon photonics foundry.

Intel agreed to buy Tower for $5.4 billion in 2022, but abandoned the deal in August 2023 after Chinese regulators declined to approve it, paying Tower a $353 million termination fee. The Japan program is the largest capital commitment in Tower's history, well beyond the up-to-$300 million arrangement it struck with Intel in September 2023 for 300mm capacity in New Mexico. Three years after nearly becoming an Intel subsidiary, Tower is building its own flagship instead.

Intel has entered high-volume manufacturing using ASML's High NA extreme ultraviolet (EUV) lithography technology for a subset of its Intel Core Ultra Series 3 "Panther Lake" processors, becoming the first company to ship high-volume logic products manufactured with the technology. ASML announced the milestone in an official press release on Wednesday, July 15, confirming that Intel Foundry is running the qualified High NA layers on its Intel 18A process node in Oregon.

According to ASML, Intel is using High NA EUV to pattern selected Intel 18A layers, with products already shipping to customers at yields matched to those achieved on ASML's existing NXE EUV platform. These layers are dual-qualified, meaning the same layer can be exposed on either an existing 0.33 NA NXE scanner or a 0.55 NA EXE scanner, with the resulting wafers being interchangeable.

High NA EUV has long been viewed as the successor to today's EUV lithography, promising to extend semiconductor scaling by enabling manufacturers to print smaller, denser circuit patterns that are becoming difficult to achieve with existing tools. Until now, the platform had been confined to R&D work. ASML’s announcement marks the first time High NA EUV has been used to produce and ship a high-volume commercial logic product.

Panther Lake, built on the Intel 18A manufacturing process, is spearheading this transition. Rather than replacing the company's entire lithography flow, Intel is applying High NA EUV to specific layers while the remainder of the chip continues to be manufactured using conventional lithography.

High NA EUV builds on the same 13.5-nanometer extreme ultraviolet light used by today's scanners but increases the optical system's numerical aperture (NA) — how much light a lens system can collect and focus onto a silicon wafer — from 0.33 to 0.55. The higher value resolves finer features in a single exposure, allowing chipmakers to print smaller patterns with greater precision and process control.

This increased resolution is expected to reduce reliance on complex multi-patterning techniques for some of the industry's most demanding layers, thereby simplifying manufacturing and improving feature fidelity. In the long term, these capabilities are expected to support higher transistor densities and improved performance in future processors, particularly as AI workloads continue driving demand for increasingly advanced semiconductor technologies.

"With increased resolution and better process control, the introduction of High NA EUV marks a substantial development in semiconductor lithography," said ASML President and CEO Christophe Fouquet. "We are proud to play a role in enabling the smaller, denser patterning that will accelerate advancements in AI and other emerging technologies."

Intel and ASML have been working towards this milestone for several years. In 2024, Intel completed installation of one of the industry's first commercial High NA EUV lithography systems, the TWINSCAN EXE:5000, at its Hillsboro, Oregon, research and development facility. The company later became the first to qualify ASML's second-generation TWINSCAN EXE:5200B, which increases wafer throughput and overlay accuracy while incorporating an improved EUV light source over its predecessor.

While the announcement represents High NA EUV's commercial debut, it does not mean Panther Lake is manufactured entirely using the new lithography platform. Instead, Intel has qualified High NA for selected layers, an approach that mirrors how new lithography generations are typically introduced into advanced semiconductor production before broader adoption across future nodes.

Intel Foundry Executive Vice President and General Manager Naga Chandrasekaran said that qualifying the High NA process option on selected Intel 18A product layers enables the company's existing tool fleet to deliver higher manufacturing output while providing flexibility for future process technologies.

Panther Lake itself is not a future product. Intel launched Core Ultra Series 3 at CES on January 5, 2026, opened preorders the following day, and put systems on shelves globally from January 27. The Core Ultra X9 378H followed in April alongside the value-tier Core Series 3, code-named Wildcat Lake, and the handheld-focused Arc G3 parts arrived on May 28.

The announcement’s statement that the product is shipping to customers refers to wafer flow from the fab into the supply chain, rather than to a product launch. ASML says the two companies will continue working on High NA readiness, with the flexibility to incorporate the technology into future nodes based on customer needs — most immediately, Intel 14A, which Intel has designed to use High NA on a set of its tightest-pitch layers.

Google plans to use Intel's EMIB-T packaging for its next-generation TPU codenamed Humufish, according to SemiAnalysis. TSMC's portfolio of chip-on-wafer-on-substrate (CoWoS) technologies has become the de facto standard advanced packaging option for nearly all AI and HPC processors made in the industry. Competing offerings are usually considered as secondary solutions if CoWoS is in tight supply, but things are beginning to change.

Google is a long-standing CoWoS customer for TPUs, starting from the Third-Generation TPU, all the way to Google's latest Eighth-Generation TPUs. Assuming that SemiAnalysis's report about Google's decision to move to EMIB-T with its Ninth-Generation TPUs is accurate, it's a big decision for Google, as switching from one advanced packaging technology to another is a complicated endeavor, which involves plenty of changes and unknowns. Understanding Google's reasons for the switch could shed some light on the prospects of Intel's and TSMC's advanced packaging technologies, which will be used by leading chip designers and hyperscalers in the coming years.

Advanced packaging technologies at glance

For years, Google used TSMC's CoWoS-S, and later, CoWoS-L packaging. Initially, the company used CoWoS-S packaging, which relies on a silicon interposer up to 3.3X the reticle size, but with its 7th- and 8th-Generation TPUs, the company moved to CoWoS-L. CoWoS-L relies on a redistribution layer (RDL) interposer with embedded local silicon interconnect (LSI) bridges that enable high-performance die-to-die links, which can scale packages to 5.5X the reticle size today. TSMC promises to improve CoWoS-L's capabilities to scale over 14X the reticle size by the end of the decade.

Unlike CoWoS, Intel's embedded multi-die interconnect bridge (EMIB) technology does not use any interposers. The technology instead relies on tiny embedded silicon bridges within the substrate to enable high-density die-to-die interconnections, whereas everything else is routed through an inexpensive organic substrate.

EMIB-T adds through-silicon vias (TSVs) to the bridge, which enables power to flow vertically instead of going through the organic substrate. In addition, Intel's EMIB-T also integrates sophisticated metal-insulator-metal (MIM) capacitors and a dedicated ground plane into the bridge to improve power integrity. The latter is a particularly important feature of complex next-generation AI accelerators, which demand more, cleaner power, and for which power delivery is becoming as challenging as signal routing.

The main selling point of EMIB (and EMIB-T) is that it is not constrained by interposer reticle limits as it places small silicon bridges only where high-density die-to-die links are needed. Strictly speaking, CoWoS-L is not either, as it uses LSIs locally as well. The difference is that those bridges are embedded into a package-wide RDL interposer that connects everything and enables dense interconnections across the package.

Since both CoWoS-L and EMIB-T are designed to address the same applications and have many similarities in the way they do this, the choice between them is likely driven by a combination of factors rather than one single advantage or disadvantage. On the technology side of matters, these factors include interconnect performance and density, power delivery, scaling beyond very large package sizes, and mechanical rigidity. On the business side of things, costs, capacity availability, and supply chain diversification are also a significant factor.

Crucial differences

SemiAnalysis claims that the main advantage of EMIB/EMIB-T over CoWoS is the lack of reticle limit, but this argument does not fully hold against CoWoS-L, as it was invented specifically to escape the reticle limitation by replacing the monolithic silicon interposer with localized LSI bridges.

When it comes to dense, package-wide routing, CoWoS-L's RDL interposer is fundamentally superior to an ordinary organic substrate offered by EMIB-T. Organic substrate wiring has coarser line/space dimensions and larger vias, so it cannot provide the same routing density as CoWoS-L's fine-pitch RDL. Where an EMIB bridge connects adjacent dies, Intel can achieve very high interconnect density. But anything that needs to travel beyond those bridges must use the package substrate or cross a topology involving additional bridges.

By contrast, CoWoS-L gives the designer two levels of connectivity: LSIs provide extremely dense local die-to-die connections, while the global RDL interposer provides relatively dense and flexible routing across the entire package. This means the RDL can carry longer, lower-density connections without consuming valuable LSI resources, while still offering much finer routing than the underlying package substrate.

One scenario for Google's choice is that it potentially wanted better power delivery than what CoWoS-L could offer. EMIB-T integrates TSVs for vertical power delivery, sophisticated MIM capacitors for local decoupling, and a dedicated ground plane into its silicon bridges. The combination of these features substantially reduces power-delivery impedance and improves transient response and power integrity, which gives EMIB-T a major advantage over conventional EMIB for power-hungry AI accelerators. However, we have no idea how EMIB-T stacks up against CoWoS-L in the case of Google’s Humufish.

Of course, the larger the RDL interposer becomes, the greater its parasitics can become, potentially limiting scaling unless TSMC finds ways to mitigate them. However, EMIB does not eliminate long-distance wiring: If two distant dies must communicate, those signals still have to travel somewhere, and routing them through an organic substrate is not inherently electrically superior to routing them through a purpose-built RDL interposer. Therefore, it is difficult to claim that Google chose EMIB-T over CoWoS-L, simply because EMIB-T offers superior package-wide electrical characteristics.

After Nvidia suffered yield loss with its Blackwell data center GPUs due to an alleged mismatch in the coefficient of thermal expansion (CTE) among the GPU chiplets, LSI bridges, RDL interposer, and motherboard substrate, which led to warping and system failure, it is reasonable to question the mechanical rigidity of CoWoS-L packages. Nvidia has found a solution for its dual compute chiplet Blackwell packages, and so have other developers of AI accelerators. However, as package dimensions increase, they may behave differently, therefore causing yield losses.

By contrast, EMIB/EMIB-T eliminates the large RDL interposer and embeds small silicon bridges in the organic substrate, so most of the package consists of the substrate itself. This does not make EMIB/EMIB-T packages immune to mechanical failures, as large packages can warp and bend, causing various problems. However, as such packages lack the very source of global thermomechanical stress, they can potentially be more robust mechanically. However, EMIB-T can potentially complicate things because TSVs, additional metal structures, MIM capacitors, and their ground plane make the bridge more complex. Thus, Intel must manage both global package warpage and local stresses around each embedded bridge to ensure the mechanical rigidity of these packages.

Ironically, while CoWoS-L can offer denser package-wide routing, which is better for ultra-large processors, EMIB-T may potentially provide better mechanical rigidity required for such devices. Nonetheless, EMIB-T and its organic substrate do not eliminate package bending or cracking risks entirely.

Economics

If Google's Humufish TPU really moves to EMIB-T, the decision could well be both technical and strategic. Google has the engineering resources to opt for an all-new packaging technology in an effort to lower costs and eliminate dependence on TSMC's constrained CoWoS capacity. Nvidia tends to procure advanced packaging allocations years in advance, so it is possible that Google could simply not get enough CoWoS-L wafers for its 9th-generation TPU.

As a bonus, Google can also build relationships with Intel Foundry without using the company's fabrication technologies. In fact, keeping in mind that Intel and Google already have a strategic agreement covering Intel Xeon CPUs, it wouldn't be too surprising to learn that the cloud giant is courting Intel Foundry as well.

Both Intel's EMIB-T and TSMC's CoWoS-L have their own technological and economic advantages and disadvantages. Perhaps the biggest advantage of CoWoS-L is its predictability, as the company has experience with that tech. However, if Google has decided to drop that predictability in favor of an all-new packaging method, it may well have a combination of technological and strategic reasons to do so.

Researchers from Osaka Metropolitan University have developed a programmable thermal device that can control where heat is radiated while remembering its configuration even after power is removed, a capability that could one day contribute to smarter thermal management in high-performance chips, silicon photonics, infrared sensors, and energy-harvesting systems. The work, published in Laser & Photonics Reviews, overcomes two longstanding obstacles that have prevented the practical realization of nonreciprocal thermal devices.

The device combines a magneto-optical material — a material that changes its optical properties in the presence of a magnetic field — with a phase-change material known as germanium-antimony-tellurium (GST) to independently control how a surface absorbs and emits infrared radiation. Unlike previous designs that lost their functionality once power was removed or only worked when light struck the surface at extreme angles, the researchers say their device operates almost straight on while retaining its programmed state without continuous energy input.

Under normal circumstances, materials follow a principle stating that if a surface efficiently absorbs heat at a particular wavelength and direction, it must also emit heat equally well under the same conditions. This relationship, defined by Kirchhoff's law of thermal radiation, holds for conventional materials and limits how precisely engineers can manipulate heat. Rather than directing thermal energy where it is most useful, these materials simply emit heat based on how they absorb it.

Circumventing this relationship has become an active area of research, as it could give engineers an entirely new way to control thermal energy. Devices capable of independently steering absorption and emission could improve radiative cooling, thermophotovoltaic systems that convert heat into electricity, infrared sensing, thermal communication, and other photonic technologies where controlling heat is just as important as controlling light.

Researchers have explored several ways to achieve this by breaking Lorentz reciprocity, the physical principle that links incoming and outgoing electromagnetic waves. Most approaches rely on magneto-optical materials, magnetic Weyl semimetals, or actively modulated metasurfaces. However, these designs have generally encountered two major problems. First, they require light to strike the surface at very oblique, or grazing, angles to produce strong directional behavior. While this works experimentally, it significantly reduces the amount of usable thermal radiation and produces broad, inefficient emission patterns. Second, many existing designs are volatile. Their behavior disappears as soon as the magnetic field, electrical signal, or heating source controlling them is removed, making continuous power necessary simply to maintain their operating state.

The Osaka Metropolitan University team tackled both limitations by combining two materials that perform complementary roles. The first is indium arsenide (InAs), a magneto-optical semiconductor whose interaction with infrared light changes in the presence of a magnetic field. Rather than allowing light to behave identically in all directions, the material introduces a directional asymmetry that enables nonreciprocal thermal behavior. The second ingredient is GST, a phase-change material that can reversibly switch between amorphous and crystalline states, dramatically changing its optical properties while retaining whichever state it is written into, even after power is removed.

The researchers patterned GST into a microscopic grating above the InAs layer, forming what they describe as a magneto-optical metagrating. The InAs provides the directional control needed to separate heat absorption from heat emission, while the GST layer acts as a non-volatile switch that stores the device's operating mode. Applying a magnetic field tunes how infrared radiation interacts with the structure, while changing the phase of the GST permanently alters that behavior until it is intentionally rewritten. In effect, the device can be programmed to emit heat differently and retain that configuration without requiring continuous energy.

According to the researchers, the prototype achieved a nonreciprocity factor approaching 0.9 while operating at an incidence angle of just three degrees, much closer to normal incidence than the steep angles typically required by previous designs. The system also supports continuous tuning via changes in the magnetic field or incident angle, as well as digital on-off switching via the GST phase transition. The team further analyzed why the nonreciprocal effect weakens when GST changes state, concluding that the reduction results from a combination of optical field redistribution and increased damping rather than simple absorption losses alone.

Although the technology remains an early-stage research demonstration, the ability to program thermal radiation could eventually become valuable in computing hardware as processors continue to pack more transistors, chiplets, and photonic components into increasingly compact packages. Future thermal metasurfaces could give engineers another tool for directing heat away from hotspots, reducing thermal interference between neighboring chiplets, or stabilizing silicon photonic devices whose optical characteristics shift with temperature.

Beyond computing, the researchers also envision applications in radiative cooling, thermophotovoltaic energy conversion, infrared emitters, thermal communication systems, and photonic memory technologies. For now, however, the work remains a laboratory demonstration rather than a deployable technology. Considerable engineering challenges remain before programmable thermal emitters find their way into commercial electronics.

Micron committed up to $3 billion to the U.S. semiconductor supply chain last week. Of that, $500 million goes to GlobalWafers as strategic financing — subject to definitive agreements and closing conditions — for its 300mm raw silicon wafer plant in Sherman, Texas, and the two companies will sign a 10-year agreement for access to that plant's output. Ben Tessone, Micron's senior vice president and chief procurement officer, tied the move to securing "critical input materials." In a second announcement from Boise the same day, Micron raised its planned US spending to more than $250 billion through 2035, up from $200 billion, and poured the first load of concrete at its Clay, New York megafab a quarter ahead of schedule.

Running until 2035, the $250 billion spending target is attached to a goal of making 40% of Micron's DRAM in the U.S. by the mid-2030s. Only a relatively paltry $500 million of that $250 billion has been earmarked for buying wafer supply from GlobalFoundries, the only U.S. supplier that’s capable of producing 300mm wafers.

The 300mm wafer market

Roughly 85% of global 300mm wafer capacity sits with five suppliers, according to market research firm Mordor Intelligence: Shin-Etsu and SUMCO of Japan, Taiwan's GlobalWafers, Germany's Siltronic, and South Korea's SK Siltron. The two Japanese firms hold more than half between them.

GlobalWafers America opened the Sherman plant in May last year on an initial $3.5 billion investment. It’s the first fully integrated 300mm raw wafer facility built in the U.S. in more than two decades, and the company says it’s the only CHIPS-participating supplier capable of producing advanced 300mm wafers domestically. The site holds a CHIPS Act award of up to $406 million, finalized in December 2024 and shared with a silicon-on-insulator plant in St. Peters, Missouri. Commerce Department figures from 2022 put full-build capacity at around 1.2 million wafers per month across a six-phase campus, with one phase currently running.

Meanwhile, SUMCO is ending 200mm production at its Miyazaki site and has slowed new 300mm expansion. The leading-edge capacity Shin-Etsu and SUMCO added in 2025 was sized to match contracted demand rather than to build ahead of the market. Wafer suppliers have run this way for a decade, protecting margins instead of chasing volume, and with suppliers holding back, the capital for new capacity increasingly comes from their customers.

GlobalWafers chairperson and CEO Doris Hsu set out her terms for that at the Sherman opening, announcing an additional $4 billion for the site and telling Reuters that further phases depended on the first two turning a profit, on customers signing long-term contracts, and on reasonable pricing, prepayments, and government support. Micron's $500 million in financing and a decade-long supply commitment cover most of that list, and Hsu has since called the Micron agreement the largest long-term deal in her company's history and said a second Sherman phase is now necessary.

Micron is locking in its own customers on the same basis, having signed a strategic customer agreement with General Motors on July 1 and another with Ford on July 6, two of 16 such agreements the company cited on its fiscal Q3 2026 earnings call. Each ties future memory output to a named buyer.

We’ve seen the industry do this before. During the memory boom of 2017-2018, chipmakers signed prepaid, take-or-pay wafer agreements to guarantee supply, but those prepayments became balance-sheet liabilities when DRAM pricing fell through 2019. SK Group chairman Chey Tae-won told an audience at Nvidia's GTC conference that the current wafer shortage could last through 2030 with a deficit above 20%, which is the argument for signing now. Conversely, the 2019 write-downs are the argument against.

HBM packaging

High-bandwidth memory is of course the component that’s currently carrying the steepest premiums in the AI market, and a fabbed wafer isn’t yet HBM. The die has to be stacked and packaged using advanced 2.5D methods with through-silicon vias, the capacity for which is located almost entirely in Asia. Micron's committed HBM packaging anchor is a roughly $7 billion facility in Singapore, with operations starting in 2026. Per a June 2025 SEC filing, the company lists U.S. HBM packaging as an intention, but no committed site or date has yet been announced.

As for U.S. packaging capacity that is scheduled, it’s all clustered in or around 2028. SK hynix is building the first U.S. 2.5D advanced packaging plant in West Lafayette, Indiana, a roughly $3.87 billion project with mass production set for the second half of 2028. Amkor, meanwhile, has expanded its Peoria, Arizona campus to $7 billion, with production slated for early 2028. TSMC's Arizona fabs run leading-edge logic but don’t yet offer high-volume 2.5D packaging on U.S. soil — this is reportedly planned for 2029. While it’s true that a wafer fabbed in New York and packaged in Singapore counts toward domestic DRAM, it doesn’t make the finished HBM stack domestic.

Output timelines vs. 2035

Micron's Manassas, Virginia fab began producing 1-alpha DRAM in May, and it’s the only U.S.-made memory in volume, representing roughly 2% of the world’s supply. The first new Idaho fab should reach wafer output in mid-2027, and the second in late 2028, while the Clay, New York campus isn’t expected to produce until around 2030. The $250 billion capex figure runs five years past that, while conventional DRAM contract prices continue to rise at record amounts — more than 90% quarter over quarter in early 2026, according to TrendForce — and manufacturers increase prices. Apple raised MacBook, iPad, and Vision Pro prices last month, citing memory costs, and none of the announced U.S. capacity will do anything to alleviate such shortages.

Samsung and SK hynix committed a combined $880 billion under a South Korean government-coordinated chip and AI program announced last month, spread over roughly a decade. That spending is domestic to Korea and separate from Samsung's $37 billion Texas footprint. But set next to Micron's $250 billion, we’re seeing a pattern of more companies announcing more capex than construction projects can physically absorb.

HBM consumes roughly three times the wafer area per bit of standard DDR5, so shifting production to HBM removes more commodity memory from the market. DRAM already takes around a fifth of global 300mm capacity, and memory is the largest single application for 300mm silicon. Micron's Sumit Sadana told CNBC in January the company could meet “at most” two-thirds of some customers' medium-term demand.

Intel has unveiled Starfire, a space-grade system-on-chip designed for the U.S. government that pairs eight CPU cores and a three-tile NPU built on its Intel 18A node with an Intel 3 graphics tile, all in one Foveros package. Intel published the Starfire sell sheet, listing two versions that draw 10 W and 35 W and reach up to 45 and 75 TOPS, respectively, rated to run between -55 and 125 Celsius.

Both SKUs share the same layout of four Intel 18A P-cores, four low-power efficiency cores, a three-tile NPU also on 18A, and a four-core Xe GPU with 64 execution units built on Intel 3. The Low Power part runs its P-cores at 1.0 GHz, efficiency cores at 850 MHz, and the GPU between 800 MHz and 1.0 GHz. The Performance part clocks the P-cores to 3.1 GHz, efficiency cores to 2.1 GHz, and the GPU to 2.0 GHz. Both carry 12 PCIe Gen4 lanes, support LPDDR5 or DDR5, and are rated for a 10-plus year lifetime.

Intel builds the CPU and NPU on 18A and the GPU on the older Intel 3, the same node division it used for Clearwater Forest, the 288-core Xeon that stacks 18A compute tiles on Intel 3 base tiles. Smaller transistors hold less charge per stored bit, which makes leading-edge silicon more prone to radiation-induced bit flips, so committing 18A to orbit leans on RibbonFET and design-level hardening rather than a mature, inherently more tolerant node.

The market Starfire is targeting has run on BAE Systems' RAD750 for two decades. That radiation-hardened PowerPC part clocks 110 to 200 MHz, carries 10.4 million transistors, and is built on 150nm or 250nm lithography, per public specifications, and it flies on the Mars rovers, Kepler, and Fermi, among more than 150 spacecraft. BAE's multi-core RAD5545 and the Microchip-built processor NASA is developing to reach 100 times the throughput of current spaceflight chips are the more recent step up. Starfire's up to 75 TOPS and dedicated NPU put it in a different bracket, built for on-orbit AI inference rather than telemetry and control.

Intel lists the radiation data, covering total ionizing dose, single-event latch-up, and single-event effects, as characterization in process, so the part isn't radiation-qualified yet, and it notes the specs are subject to change. Intel Government Technologies is handling Starfire, with samples in Q3 2026 and a pitch of market-competitive pricing and domestic manufacturing. Intel Foundry is the only U.S.-based maker of leading-edge logic, holds Trusted Foundry status, and has tied its 18A and packaging roadmap to Pentagon programs including RAMP-C and SHIP, though 18A yields aren't expected to reach industry-standard levels until 2027.

Apple's planned M7 Ultra chip is being designed to support up to 1.5 TB of unified memory and to push AI performance toward the class of Nvidia's Blackwell accelerators, according to a new Bloomberg report published by Mark Gurman. But whether the lofty top memory config can ship at all will depend on the state of the memory market, and the part isn't expected until 2028. The same report says Apple has compressed its Mac silicon timeline, taping out the M7 roughly six months after the M6.

We've already heard that Apple plans to release a base M6 chip this fall for entry-level Macs, then skip the Pro, Max, and Ultra versions of that generation and move straight to the M7 line. However, Gurman now reckons that we'll see a base M7 in the first half of 2027, M7 Pro and M7 Max at the end of 2027, and the M7 Ultra in 2028. Apple reportedly began taping out the M7 about six months after it started the same process for the M6, which is what has enabled the company to pull the schedule forward.

The 1.5 TB target for the M7 Ultra is roughly twice the capacity Apple has planned for the M5 Ultra, per Gurman, who tied the configuration directly to memory availability. Apple already pulled the 128GB Mac Studio this year over supply constraints as DRAM prices climbed, and a 1.5 TB part would call for far more of the same scarce, high-cost memory.

Apple's current M3 Ultra already reaches 819 GB/s of memory bandwidth by fusing two Max dies, and it's the Ultra tier, not the base chips, that carries the heaviest local-AI workloads. Gurman describes the M7 Ultra as a large step up in AI performance rather than stated parity with Nvidia's data-center silicon. "I'm told the processor dramatically upgrades AI performance, bringing it closer to the class of dedicated AI accelerators such as Nvidia Corp.'s Blackwell," Gurman wrote in his report.

Apple is also preparing an AI server built on the M5 Ultra under the code name J246 for deployment soon, with a second server chip based on the M7 Ultra planned for 2029, according to the report. The 2028 generation, which includes an M8 chip code-named Soko and a high-end Mac part called Cardinal, moves to a 1.4nm process. That aligns with TSMC's A14 node, which the foundry has scheduled for mass production in the second half of 2028.

None of the dates or specifications have been confirmed by Apple.

SK hynix has completed the largest-ever foreign company IPO in U.S. history, raising $26.5 billion in its Nasdaq debut today, July 10. The South Korean memory giant sold 177.9 million American depositary receipts (ADRs) — a U.S.-listed stand-in for a foreign share — at $149 apiece, each representing a tenth of a Seoul-listed share. The offering was more than seven times oversubscribed and drew demand from more than 500 investment firms, according to Financial Times. Temporary Nasdaq trading is underway under the ticker SKHYV before regular-way trading begins as SKHY on Monday, July 13.

The offering was led by Bank of America, Citigroup, Goldman Sachs, and JPMorgan, with nine additional firms rounding out a 13-bank syndicate. Anchor demand came from heavyweight institutions including Baillie Gifford, Coatue Management, and Situational Awareness Partners, which together signaled interest in as much as $7 billion of stock, according to people familiar with the matter cited by Financial Times.

SK hynix is the world's leading maker of high-bandwidth memory (HBM), the vertically stacked DRAM that has become critical infrastructure for AI accelerators. The company has said it will steer the proceeds toward boosting its AI-memory manufacturing capacity. Confirmed build-outs include the first-phase fab at the massive Yongin semiconductor cluster, a new P&T7 advanced-packaging line in Cheongju, and EUV lithography equipment slated for delivery by the end of next year. Separately, SK hynix is constructing its first U.S. production site, a $4 billion advanced-packaging plant in West Lafayette, Indiana, targeted for completion around 2028. The facility is eligible for up to $458 million in CHIPS Act grants and up to $570 million in federal loans.

What display resolution do you use on your primary monitor?

SK hynix is seeing sensational growth thanks to the ongoing AI boom. The company is reportedly on track to post over 200 trillion won ($133 billion) in operating profit this year, a record-breaking figure that would see SK hynix employees earn around $400,000 each in bonuses. The company’s Seoul-listed stock is up roughly 220% year-to-date and has climbed more than sixfold over the past year.

In late June, SK hynix briefly surpassed Samsung as South Korea's most valuable company, closing at around 2,080 trillion won (about $1.35 trillion), a meteoric rise for a company that almost declared bankruptcy in 2001 and, more recently, recorded an annual operating loss of 7.73 trillion won in 2023. That rise doesn't seem like it will be slowing down any time soon. SK hynix has said its entire 2026 output of HBM, DRAM, and NAND is already sold out, with the crunch expected to extend into 2027.

Japanese chipmaker Rapidus will try to lure customers away from TSMC not only by offering a different kind of service, but also by offering its manufacturing services at lower prices, chief executive Atsuyoshi Koike announced this week. The company's plan to rival TSMC in terms of pricing appears on the surface as a risky move, as the company moves to develop leading-edge process technologies.

At present, Rapidus is looking at charging ¥3 million – ¥3.5 million ($18,550 - $21,635) per wafer processed using its 2nm-class fabrication process, which is significantly below TSMC's rumored quote of around $30,000 per N2 wafer, and is comparable to what Samsung is rumored to offer with its SF2 manufacturing technology, set at $20,000 per-wafer. Actual prices will depend on exchange rates, though Rapidus' general idea of offering significantly lower quotes than TSMC is immediately apparent.

Rapidus plans to start high-volume manufacturing (HVM) using its 2nm-class fabrication technology by the second half of 2027. The ramp of a new fab will take some time, so expect meaningful volumes from Rapidus to only be produced in 2028, when TSMC's N2 will no longer be its leading-edge node.

By the time Rapidus starts HVM at its IIM-1 in 2027, TSMC will have ramped production of chips using its performance-enhanced N2P manufacturing node, and the company will also absorb all the yield learning with gate-all-around the company will have with its N2 present at five fab modules. Furthermore, by the time Rapidus reaches meaningful volumes at IIM-1 in 2028, TSMC will have ramped up production using its advanced A16 fabrication process with Super Power Rail backside power delivery as well as a 3^rd-generation 2nm-class node named N2X.

In addition to the vast 2nm-capable capacity and process maturity that should be kept in mind when comparing Rapidus with TSMC, there is another factor to consider. One of TSMC's major advantages over its rivals is its Open Innovation Platform (OIP) ecosystem, which includes comprehensive electronic design automation tools, silicon-proven IPs, even for the latest nodes, a host of contract chip designers, and advanced packaging services not only from TSMC but also from its partners. For now, neither Rapidus nor Intel and Samsung Foundry can offer anything close to TSMC's OIP.

Given the advantages that TSMC will likely have over competitors with its 2nm-class fabrication technologies in 2028, lower pricing may be among the few ways to compete against the world's largest foundry. Rapidus' strategy of offering lower quotes while operating a single fab does not seem like the best way of earning money, but perhaps a certain way to lose it.

However, Rapidus may have another ace up its sleeve with single wafer processing across all process steps. The approach will greatly speed up the production cycle, which will be its indisputable advantage over other chipmakers, albeit at the cost of tool usage efficiency. Will lower quotes and shorter production cycles be enough for Rapidus to win customers from TSMC? Only time will tell.

Rapidus is reportedly negotiating with more than 60 potential customers, mainly overseas companies, which demonstrates the company's ambitions to become a viable rival to the global leader TSMC as well as contract chipmakers Intel Foundry and Samsung Foundry.

Researchers in Korea and Japan have presented two separate memory-integration proposals that aim to increase HBM (High-Bandwidth Memory) capacity and bandwidth without trapping more heat inside ever-taller DRAM (Dynamic Random Access Memory) stacks, one of the most pressing challenges facing future AI accelerators. Presented at the 2026 IEEE/JSAP Symposium on VLSI Technology and Circuits held in June, the two approaches — V-Die from a Korean research collaboration and MOSAIC from a University of Tokyo-led group — both explore the same broad idea of standing DRAM memory dies on their edges instead of stacking the memory dies only upward like conventional HBM.

The Korean proposal, called Vertical-Die (V-Die), was presented by researchers at the Ulsan National Institute of Science and Technology (UNIST). The design rotates custom DRAM dies upright, drops through-silicon vias to free die area for more memory cells, gives each die its own bottom-edge I/O, and runs liquid-cooling channels between adjacent dies. In simulations against an HBM4 system at equal capacity, the V-Die system reportedly achieved 540 tokens per second on a GPT-3-sized workload, compared to 296 tokens per second for HBM4.

The Japanese project, MOSAIC, takes a similar “sideways stack” idea but focuses on the practical difficulty of connecting so many vertical dies to a GPU or package substrate. Presented by University of Tokyo researchers, the MOSAIC work uses orthogonal die stacking and a contactless die-to-die interface, in which data is transferred through tiny inductive coils rather than requiring every signal pad to land perfectly on a physical contact. The researchers say the prototype interface achieved up to 4 Gbps per channel, while the memory structure could double HBM4-class capacity in a DRAM-on-GPU configuration.

Both projects aim to solve the growing problem of AI chips being held back by memory. Modern accelerators can perform enormous amounts of computation, but large, powerful models depend on moving huge amounts of data between memory and compute. This is why HBM has become one of the defining technologies of modern AI hardware.

The technology addresses the memory wall by stacking multiple DRAM dies vertically on a base die and placing that stack very close to the processor. Nvidia's Blackwell Ultra B300, for instance, carries up to 288GB of HBM3E memory, without which much of the silicon would sit idle waiting for data. The dies are connected via through-silicon vias (TSVs) — tiny vertical channels etched through the silicon and filled with metal.

The stack then communicates with the GPU over an extremely wide interface, often routed through a silicon interposer or an advanced package. This is the core reason HBM can deliver terabytes per second of bandwidth: it uses a very wide, very short data path instead of sending memory traffic across a motherboard, as with conventional DIMMs (Dual In-line Memory Modules), physical sticks of RAM used in computers.

However, that same structure creates several problems. While taller stacks add more capacity, they also make it harder to remove heat. Heat generated in the lower dies and at the high-speed interface must pass through layers of silicon, bonding materials, underfill, and package structures before it reaches a heat spreader. Furthermore, TSVs consume die area that could otherwise be used for memory cells, and as bandwidth rises, more routing and I/O place additional pressure on both signal integrity and packaging costs.

HBM4, the latest generation of HBM, addresses a number of these challenges. Meanwhile, companies such as SK hynix, Samsung, and Micron are racing to improve speed, capacity, base-die performance, and thermal management. SK hynix has already shown iHBM, which embeds cooling elements into the HBM interface area, and Samsung has shown an HBM5 mockup with Heat Path Block cooling to more directly extract heat from the stack. However, they all retain the same upward stacking structure.

This convention is what V-Die and MOSAIC are challenging. By standing DRAM dies upright, the researchers expose far more silicon surface area to the cooling path. In theory, this turns the memory stack into something closer to a heat-sink fin array, where heat can move laterally and escape more directly instead of being trapped in the middle of a thick vertical pile. It also opens the door to new connection schemes along the bottom or side of each die, rather than forcing every die to communicate through TSVs running vertically through the stack.

For V-Die, the key shift is removing TSVs from the memory dies and replacing them with bottom-edge connections. Each DRAM die gets its own I/O along the bottom edge and connects directly to the substrate, with links reportedly spaced every 20 microns. The team says this layout gives four times as many connections as HBM4 and cuts memory read time by 37%, although some signals must travel farther across the package to reach the processor.

Cooling is the other half of the V-Die argument. The proposal places microfluidic cooling channels between adjacent upright DRAM dies, allowing coolant to dissipate heat closer to its source. According to the researchers, this could keep the stack around 45°C, far below the 80°C-plus range associated with dense HBM systems. In a simulated 16-die stack matched to H100-class hardware on a GPT-3-scale model, V-Die hit 540 tokens per second, compared to HBM4's 296, and cut first-token latency by 32%, or about 24 milliseconds.

MOSAIC, meanwhile, is focused on making the sideways stack manufacturable. Because the dies are assembled flat and then turned on edge, even a few microns of die-thickness variation across dozens of dies can add up to an alignment miss where the signal pads no longer land. The Japanese team’s answer is a contactless interface based on inductive coupling. One side of the memory die carries oblong coils, while a corresponding set of coils sits on the substrate or mating chip. Current in one coil induces a signal in the other, allowing data to cross the small gap without a direct metal-to-metal signal contact. This eliminates the need for precise overlapping, giving the package greater tolerance for assembly variation. Power, which requires fewer, larger connections than data, can still be supplied via physical contacts on the sides of the memory cube.

The VLSI MOSAIC prototype achieved up to 4 Gbps per channel and demonstrated TSV-free 3D integration for a memory-on-GPU layout. The team says the approach can enable twice the memory capacity of HBM4 without significantly increasing peak temperature. A related bump-MOSAIC hardware demonstration at ECTC used 100-micron-pitch microbumps, achieved stacking alignment within 6 microns as verified by X-ray CT, and showed a configuration with three times the thermal conductivity of conventional stacking while adding up to 30% more memory capacity.

While the results look promising, neither V-Die nor MOSAIC is close to replacing commercial HBM. Neither is close to shipping. V-Die is still a proposed architecture, with a prototype in the works to validate its thermal and electrical behavior; MOSAIC has proof-of-principle hardware, but the researchers have yet to show it scales to commercial DRAM capacity, yield, cost, and reliability.

Still, any viable solution to the multifaceted AI memory problem is a welcome development. SoftBank and Intel’s Z-Angle Memory (ZAM) and NEO Semiconductor’s 3D X-DRAM — both still in development — aim to solve the constraints of conventional memory. Meanwhile, the overall market is already feeling the squeeze on price and availability, even as memory makers divert capacity toward the more lucrative AI HBM and server products, driving consumer RAM prices even higher.

Micron has said it will invest up to $3 billion in the U.S. semiconductor supply chain, with $500 million of that going to GlobalWafers as strategic financing for its 300 mm raw silicon wafer plant in Sherman, Texas, alongside a 10-year agreement giving Micron access to that plant's wafer output. In a separate announcement, the memory maker raised its planned U.S. spending to more than $250 billion through 2035, up from $200 billion, and confirmed the first concrete pour at its Clay, New York campus more than a quarter ahead of schedule.

Sherman is the sole operating facility in the U.S. capable of producing advanced 300 mm raw silicon wafers, the substrate on which every leading-edge DRAM, NAND, and logic die is built. GlobalWafers opened the plant in May last year and holds a $406 million CHIPS Act award covering the site and a silicon-on-insulator facility in St. Peters, Missouri. The 142-acre campus is designed for up to six phases, one of which is running. Micron's other American sites draw their wafers from Japan, Taiwan, Germany, and South Korea; Shin-Etsu, SUMCO, GlobalWafers, Siltronic, and SK Siltron together control the overwhelming majority of global 300 mm supply, making raw silicon the most concentrated layer in the chip space.

Doris Hsu, chairperson and CEO of GlobalWafers, set out her terms for building phase two at Sherman during the plant's opening last year, telling Reuters the company needed profitability at the first two phases, customers willing to sign long-term contracts, reasonable pricing, prepayments, and government support. Thursday's announcement supplies most of that list in a single transaction.

Wafer suppliers spent the 2023-2024 downcycle protecting margins rather than adding capacity, and SUMCO is winding down 200mm production at Miyazaki this year while holding the line on new 300mm expansion. Customers, not suppliers, are therefore now underwriting the capacity. The last time the industry did this, during the 2017-2018 megacycle, chipmakers signed prepaid long-term agreements that turned into liabilities when pricing rolled over.

Silicon wafer shipments reached 3,275 million square inches in Q1 2026, up 13.1% year over year, with SEMI.org attributing the growth to AI data center demand across advanced logic, memory, and power devices. Micron's first new Idaho fab, ID1, is expected to begin wafer output in mid-2027, and production at Clay isn't expected until around 2030. The company began making 1-alpha DRAM at its Manassas, Virginia fab in May.

Micron told investors last December that it can serve only half to two-thirds of customer demand, and nothing announced Thursday changes the supply position of DRAM this year or next.

Rapidus is bidding Japan's entire return to leading-edge logic on one fab in Chitose, Hokkaido, and the schedule now turns on a 2027 mass-production target for a 2nm process that no high-volume customer has yet committed to.

Since opening the IIM-1 pilot line in April last year, the company has run wafers through Japan's first mass-production-grade EUV scanner, produced a 2nm gate-all-around prototype that reached its expected electrical characteristics in July, and closed a ¥267.6 billion funding round in February that made the Japanese government its largest shareholder. CEO Atsuyoshi Koike said the same month that more than 60 companies are in talks over 2nm capacity, but not one has yet signed a volume agreement. Given that its entire production base is the single IIM-1 facility, this leaves Rapidus with no diversification and no fallback site if the node doesn’t go ahead as planned.

However, the fab has the hopes of an entire nation pinned on it, and its plans are promising. Here's the breakdown.

A ticking clock on IIM-1

IIM-1, short for Innovative Integration for Manufacturing, broke ground in September 2023 at Bibi in Chitose, with the cleanroom completed in 2024. ASML delivered a TWINSCAN NXE:3800E in December 2024, the first mass-production-grade EUV system installed in Japan, and the tool completed its first exposure on April 1st last year. The pilot line also began operating that month.

Rapidus is currently targeting 2027 for mass production, but the company has given that date without any further qualification, with its business plan simply pointing to production beginning in the second half of fiscal 2027 and scaling to full volume in 2028. The same plan sets out a capacity ramp from roughly 6,000 wafer starts per month at the outset to around 25,000 within the first year, a fourfold increase that Rapidus is counting on to bring per-wafer costs down.

IIM-1’s siting in Chitose offers the abundant water that wafer cleaning demands, a cool climate that eases cooling loads, and some of Japan's strongest renewable-energy potential across wind, solar, and hydro. Local and prefectural authorities have organized around the project under a “Hokkaido Valley” initiative that aims to build a semiconductor cluster spanning Tomakomai, Chitose, and Ishikari.

The 2nm process

Rapidus’s 2nm node is a gate-all-around nanosheet design derived from the IBM 2nm process announced in 2021, the product of a partnership signed in December 2022. Rapidus engineers worked alongside IBM at the Albany NanoTech Complex in New York to learn the node before transferring it to Chitose. More than 150 Rapidus engineers were dispatched to Albany across 2023 and 2024 to learn the node, with roughly 80 later returning to Chitose to transfer and tune the process for production, according to IBM.

The differentiator the company is leaning on is manufacturing flow, with IIM-1 running single-wafer front-end processing throughout, branded as Rapid and Unified Manufacturing Service, with per-wafer data fed into AI models that Rapidus says will accelerate yield learning and shorten turnaround compared with the batch processing used by TSMC and Samsung. It’s understood that the 2nm Process Design Kit (PDK) reached early customers in Q1 this year. Still, Rapidus hasn’t yet published a yield figure, and its public claims extend only to the prototype attaining expected electrical characteristics.

The program extends beyond the wafer, with Japan’s New Energy and Industrial Technology Development Organization (NEDO) approved fiscal 2026 budget for Rapidus providing funds for chiplet and package design and manufacturing technology for 2nm-generation semiconductors, alongside front-end work. The company has also floated panel-level glass-substrate packaging as part of its longer-term roadmap. Building that back-end capability in Chitose rather than outsourcing it would mirror the integrated approach Intel and Samsung take.

Japan’s government as a shareholder

Rapidus’s February funding round closed at ¥267.6 billion, or about $1.7 billion, split between ¥100 billion from the government through the Information-technology Promotion Agency and ¥167.6 billion from 32 private companies. The state investment, the first made possible by a 2025 revision to Japan's subsidy law permitting government equity in Rapidus, made Tokyo the largest single shareholder, with a golden share giving it veto power over major decisions, including share transfers and technology partnerships.

That round sits on top of a much larger commitment from November, when Japan's Ministry of Trade and Industry added approximately ¥1 trillion in support across fiscal 2026 and 2027, lifting total planned government backing to about ¥2.9 trillion. The government added a further ¥150 billion in equity in early June, taking Rapidus’s combined capital and capital reserves to around ¥425 billion. The shares the state holds are structured as largely non-voting, keeping its formal voting position near 11.5%, but they convert to a controlling stake of roughly 60% if performance deteriorates, a clause that pairs with the golden share to give Tokyo both upside alignment and a downside lever.

Rapidus’s buildings and equipment are also currently owned by Japan's New Energy and Industrial Technology Development Organization (NEDO) and leased back, with the company previously obligated to buy them by fiscal 2027. The government now plans to construct fab buildings and tools with public money across fiscal 2027 and 2028 and transfer them to Rapidus as in-kind contributions in exchange for shares, removing that purchase obligation and converting what had been grant funding into direct ownership.

The customer conundrum

Koike said in February that Rapidus was in discussions with more than 60 companies and had issued preliminary price quotations to around 10 of them. The names attached to those talks in reporting by TrendForce are IBM and the Canadian RISC-V accelerator startup Tenstorrent, with Fujitsu, a founding investor, separately weighing whether to outsource a 1.4nm CPU for a successor to its Fugaku supercomputer around 2029.

The design partnerships Rapidus has actually signed, however, are with smaller players building energy-efficient AI silicon. Tenstorrent, the firm led by chip architect Jim Keller and currently being considered for takeover by Qualcomm, agreed back in 2023 to co-develop an edge-AI accelerator on the 2nm node under a NEDO and Leading-edge Semiconductor Technology Center (LSTC) project, with Tenstorrent handling the CPU and Rapidus' AI Chip Design Center building the accelerator. Rapidus signed a separate memorandum of cooperation with RISC-V inference designer Esperanto Technologies in May 2024.

Neither of these amounts to the committed high-volume order that Rapidus needs, and fast. Koike has described interest as growing “like a runaway steam engine,” but interest is not the same as allocation. Rapidus’s cost model depends on filling the 25,000-wafer ramp, and a fab running well below capacity carries the same fixed depreciation as a full one. The company has said its homegrown 2nm chips could cost around 10 times more than Japan’s current mainstream parts, a premium that’ll only narrow with volume.

Japan's two-track strategy

Rapidus is the leading half of a national plan that’s running on two tracks. The other is TSMC's JASM venture in Kumamoto, on the southern island of Kyushu, where a first fab backed by Sony, Denso, and Toyota began mass production in December 2024 on mature 12nm, 16nm, 22nm, and 28nm nodes aimed at automotive and industrial chips.

A second Kumamoto fab broke ground in 2025, and its planned node was upgraded twice, first to 4nm and then to 3nm, with production targeted around 2028. Tokyo is therefore funding mature and specialty capacity through a proven foreign operator in the south while betting on a domestic startup to reach the leading edge in the north. The Kumamoto plants carry far less technical risk and are already shipping; Rapidus carries nearly all of the program's execution risk and none of its proven output.

Even as the 2nm line ramps, Rapidus plans to begin 1.4nm process development in 2026, start construction on a 1.4nm fab in 2027, and reach mass production around 2029. The node leans on the company's research ties: Rapidus joined imec's core partner program in April 2023, giving it access to the Belgian institute's pilot line, and imec's position is that 1.4nm single-patterned layers require High-NA EUV, the 0.55 numerical-aperture tool that resolves the most critical metal layers in one exposure rather than several. Total lifetime investment is expected to exceed ¥7 trillion, with roughly ¥5 trillion needed just to reach stable 2nm production, according to figures cited by both TrendForce and Nikkei.

Rapidus’ financials highlight how far all this is from being self-sustainable, having posted a ¥375 million loss for fiscal 2025 with total assets of ¥749.5 billion, while still aiming to raise around ¥1 trillion from private investors. The ASEAN+3 Macroeconomic Research Office has been cited as estimating that committed funding still falls short of the roughly ¥5 trillion needed for stable production, dependent on a private investor base that has yet to materialize at scale.

Meanwhile, TSMC moved its N2 node to volume production in late 2025, and Samsung began first-generation SF2 mass production the same year, which puts Rapidus roughly two years behind both on a node that customers can already buy elsewhere with proven yields. There’s also something of an adversarial backdrop developing: TSMC employees were reportedly accused last year of taking 2nm trade secrets said to be destined for Rapidus, an allegation a Japanese government official later characterized as non-critical. The episode has no confirmed bearing on Rapidus’s process, which is built on IBM IP.

Rapidus is targeting operating profitability around fiscal 2030 and an IPO in fiscal 2031, a timeline that assumes the 2027 ramp lands on schedule and that the customers now in talks convert into committed volume. Three milestones will indicate whether those targets are possible: a named customer with a committed volume order, evidence that 2nm yields are tracking toward the levels TSMC and Samsung already run at, and confirmation that the capacity ramp is hitting its 25,000-wafer target.

An Intel patent application published on July 2, 2026, surfaced by Underfox, has revealed the company's plans for a new high-bandwidth memory (HBM) architecture that aims to ease the packaging and cost bottleneck of today's interposer-based HBM. The patent application — filed back on December 26, 2024 — describes what Intel calls cross-batch memory (XBM), an "ultra-high-bandwidth memory with backend transistors" built with the goal of matching HBM4's footprint while swapping conventional DRAM and its ultra-wide interface for back-end-of-line (BEOL) transistors and serial Universal Chiplet Interconnect Express (UCIe) links.

Intel's proposed design is a memory stack that addresses the assembly costs that make conventional HBM expensive by dropping the costly silicon interposer and shrinking the package, while building in its own defect repair.

The filing lays out a stack of memory dies, each holding one-transistor one-capacitor (1T1C) DRAM fabricated in the back-end-of-line, wired together with through-silicon via (TSV) "gutters" and both-sided high-bandwidth interconnect (HBI) connections. Intel describes dies of roughly 1.5 gigabytes (GB) apiece — 768 "datablocks" arranged in a 32-by-24 grid, grouped into eight channels of eight sub-channels each — stacked eight high and scaling to 16. Data then leaves the stack over UCIe I/O bundles running at 32 gigatransfers per second (GT/s), funneled out through a base die.

To understand what Intel is changing, it helps to recall what standard high-bandwidth memory does. HBM stacks DRAM dies vertically on a base logic die, threads them together with TSVs, and communicates with the processor across a silicon interposer using an extremely wide parallel interface — on the order of 1,024 bits per stack. This width is how HBM delivers its bandwidth, but it is also what makes it expensive to package and hard to scale, as every one of those wires has to be routed through an interposer sitting between the memory and the compute die. As AI accelerators have outrun the rate at which memory can feed them, this "memory wall" has become the dominant constraint on performance, which is why nearly every large chipmaker is now attacking the interface and the stack rather than the logic.

XBM's first major change is structural. Conventional DRAM cells are built in the front-end-of-line, the base silicon layer where transistors are normally fabricated. XBM instead moves the 1T1C cell into the back-end-of-line, the metal-and-via stack above the transistor layer, using thin-film transistors. Building memory in the BEOL is what lets Intel pack the die into many small, independently addressable memory blocks, and it is the same backend-transistor direction Intel has pursued for placing memory directly over logic.

The second change is the interface. Rather than HBM's wide parallel PHY, XBM serializes data onto UCIe bundles at 32 GT/s, with the base die handling the serialize/deserialize step and routing all I/O to the compute die. Moving to a standard chiplet interconnect is what makes the design "chiplet-native" and, Intel argues, simpler and cheaper to package than an interposer-bound HBM stack. The tradeoff is that 32 GT/s is UCIe's current top data rate, so the interface is already running at the spec ceiling rather than leaving obvious headroom.

Intel also leans heavily on repairability. The base die carries dedicated spare channels, built-in self-repair (BISR), decode and debug logic, and four sub-channels of redundant memory arrays that act as fungible spares for defects in the dies above — post-assembly repair designed to claw back yield on a very tall stack.

A large portion of the patent application focuses not on the memory cell at all but on how to mount it. Intel details memory-on-package (MoP) and "reversed overhang" structures aimed at cutting the stack's Z-height — conventional MoP can add 300 to 350 micrometers (µm) — while removing the stiffener normally needed to control warpage and feeding DRAM power directly from the voltage regulator. This is the concrete basis for the "smaller, cheaper package" claim.

XBM should not be confused with ZAM (Z-Angle Memory), the architecture Intel is co-developing with SoftBank subsidiary SAIMEMORY and set to present at the VLSI Symposium 2026. ZAM's innovation is on the bonding side — a fusion-bonded, nine-layer stack of largely conventional DRAM with roughly 3-µm-thin silicon between tiers — and it reportedly targets around twice HBM4's bandwidth density, with commercialization aimed at 2029. XBM, by contrast, is an Intel-only filing that changes the DRAM transistor itself and the interface. Read together, they suggest Intel is running at least two parallel HBM alternatives, a fitting move for a company that began in 1968 as a memory maker.

The caveats on Intel’s proposed HBM architecture are the usual ones for a patent. The patent was filed 18 months ago, and there’s currently no product or roadmap, signaling potential intent rather than a shipping part. The UCIe interface is already at its rate ceiling, backend-transistor DRAM remains unproven at manufacturing scale, and the whole thing still has to justify itself against HBM4E and Intel's own ZAM timeline.