Huawei has announced a new chip design framework aimed at closing the technology gap with global semiconductor leaders like TSMC and Nvidia, targeting '1.4nm-class' transistors and a 55% increase in transistor density. The firm also unveiled a new 'Tau Scaling Law' that's designed to replace Moore's Law for future chip scaling. Unveiled at the IEEE International Symposium on Circuits and Systems (ISCAS 2026) in Shanghai on Monday, this new design method is intended to circumvent strict US trade sanctions. It allows the company to develop high-performance smartphones and AI processors without relying on restricted Western manufacturing equipment like extreme ultraviolet (EUV) lithography machines.

Delivering a keynote address at the symposium, He Tingbo — a Huawei board member and President of its semiconductor division, HiSilicon — unveiled the company's new, proprietary “LogicFolding” architecture. The cutting-edge design blueprint is built directly upon the newly introduced Tau Scaling Law.

He revealed that Huawei has spent the last six years quietly refining the methodology, secretly designing and mass-producing 381 chips based on the principle. The company will debut the LogicFolding architecture in flagship Kirin smartphone processors this autumn.

Traditional chipmaking relies on Moore's Law (geometric scaling), which involves shrinking physical transistor sizes. However, as US sanctions blocked China's access to the extreme ultraviolet lithography machines required to implement this approach, HiSilicon has pivoted to a completely different methodology: the Tau scaling law.

Tau Law is a "temporal scaling" framework that prioritizes signal speed, optimizing how fast data moves across a system rather than how small the components are. To execute this theory on a commercial level, Huawei engineered the LogicFolding architecture, a blueprint that physically folds and stacks logic circuits into a dual-layer framework. By drastically shortening internal wiring to eliminate signal delay, the resulting hardware achieves a 55% increase in transistor density and a 41% boost in power efficiency, enabling Huawei to build cutting-edge processors that rival foreign counterparts without Western equipment.

The company’s upcoming Kirin smartphone chips — highly anticipated for the flagship Huawei Mate 90 series — will be the first commercial processors to feature the LogicFolding architecture. The company aims to scale this architecture to its Ascend AI processors and high-capacity data center clusters by 2030. This will provide local alternatives to restricted Nvidia hardware. By 2031, Huawei confidently projects it can design high-end chips with a transistor density equivalent to a 1.4-nanometer (nm) process.

Huawei's announcement comes as China continues its push to end dependence on foreign semiconductor players — amid sanctions and concerns about over-reliance — by aggressively investing in domestic companies and alternative technologies.

Following the announcement, shares for China's largest contract chipmaker, SMIC, surged by 7.6%. The breakthrough is a major symbolic and practical win for Beijing’s push toward complete technological self-sufficiency. While global foundry leader TSMC expects to mass-produce true 1.4nm chips by 2028, Huawei's alternative path means China can dramatically close the performance gap by packaging and structuring chips differently — significantly mitigating the impact of the US clampdown.

For decades, semiconductor design has been driven by humans coming up with bright ideas that unlock new innovations. But the benefits of better chip design have been reaped, including the rise of AI, which now means there could be another party involved in making chip designs smarter: AI itself.

‘Chip designer’ isn’t one of the roles on the chopping block as AI automation upends the job market. But in the narrow pockets of the design flow where problems are structured, and evaluators are robust, it is starting to be adopted — with benefits.

Google DeepMind's AlphaChip reinforcement-learning system has produced designs for three generations of the company's Tensor Processing Units (TPUs), with DeepMind claiming "superhuman" layouts compared with those produced by human designers. They’re not alone: Synopsys has passed 100 production tape-outs with its DSO.ai design-space-optimization tool, reporting productivity boosts of more than three times and power reductions of up to 25% for customers including STMicroelectronics and SK hynix.

"Like every new technology, AI may have multiple uses," said Borivoje Nikolić, professor of electrical engineering and computer sciences at the University of California, Berkeley, in an interview with Tom’s Hardware Premium. Nikolić drew a parallel with Moore's Law, which has historically been exploited in two ways: to reduce the cost of an existing product by porting it to cheaper processes, or to add features that were previously impossible. "I think AI will be used in both ways," he says. "At the moment, the industry seems to be focused on the first item — how to make things cheaper, how to automate things in a better way than they were in the past."

By contrast, academics are more interested in using AI to discover things humans haven't yet thought of, an approach that mirrors breakthroughs in areas such as drug discovery and protein folding with the likes of AlphaFold.

Nikolić and his colleague Sagar Karandikar have been exploring that territory in their own research on cache replacement policies, a subject deep in the weeds of processor microarchitecture. Their ArchAgent system, built on Google DeepMind's AlphaEvolve framework, generated a cache replacement policy in two days that beat the prior state-of-the-art by 5.3% in IPC speedup on Google's multi-core workload traces. On the heavily worked-over single-core SPEC06 benchmarks, it took 18 days to eke out another 0.9%. That’s a "first sign of life" for Karandikar that large language models can design genuinely new logic, rather than just tinkering with existing parameters.

"There is still a lot of human guidance, and it kind of up-levels the kind of thinking humans have to do," said Karandikar, a computer architecture researcher at Berkeley, in an interview with Tom’s Hardware Premium. "The humans involved in that project are doing more of the high-level thinking — coming up with new ideas and guiding the LLM — and the LLM does a lot of the finer policy development around that."

Where AI is making breakthroughs

For Igor Markov, a chip design researcher who has spent years at the frontline of electronic design automation, the places where AI is adding real value are specific and often mundane. Some of the biggest wins, he says, come at the low end of the flow, such as tasks that previously required engineers to interpret informal specifications written in natural language and convert them into formal descriptions that a tool can act on.

Take power and ground networks, the intricate webs of metal that feed electricity across a chip. "They’re sometimes designed just with descriptions in natural language," Markov said in an interview with Tom’s Hardware Premium. "People explain the geometry, and then it's implemented, and at some point, you need to formalize it. This is a step that was done manually, and it's pretty straightforward to automate using AI." The productivity dividend isn’t massive; “it took a couple of days, now it's a couple of hours,” he explained. But is still better than nothing, even if the output still needs to be checked.

Where Markov is most bullish is on what he calls the agentic space: the high-level orchestration of chip design flows, including deciding whether a run is doomed or whether a flow needs to be restarted entirely. “If you take a zero multiplied by something, you get a zero,” he said. “But if you already have something decent, then this high-level control can be very, very enabling.”

The most stubborn corners of the industry are starting to think about adopting AI. Analog design has long been seen as the last redoubt of human craft, but researchers have begun producing generative AI systems such as AnalogGenie, which uses a GPT-style model to discover new circuit topologies, and Princeton's AI-enabled design-space discovery for millimeter-wave and sub-terahertz power amplifiers operating between 30 and 120 GHz.

It’s in these areas that what’s often seen as AI’s failing, that it doesn’t have an inherent knowledge or muscle memory of its own, becomes a strength. Humans have a tendency when porting a design from one process node to another to assume the old topology must be close to optimal for the new one. “AI may not have those kinds of barriers,” said Nikolić.

Testing versus real life

However, some caution is needed. AI can be trained to ace demos, but can flunk the messier problems engineers face in practice. "Whether something that works in five cases works in general, and allows you to innovate, that's the key," says Markov.

There is also the problem of what it is you are asking AI to do in the first place. Ask a model to design the best chip for AI, and without a formal, unambiguous specification of what best means, the model will produce something — or anything. "You will play whack-a-mole," Markov said when it comes to making it work in practice.

He added that every previous jump in design automation has provoked similar debates about whether machines can really think. Shortest-path algorithms for wire routing, once seen as a distinctly human capability, became undergraduate coursework. Placement algorithms now routinely outperform human designers. Logic synthesis, once considered too abstract to automate, is handled by for loops and conditionals. “EDA has always been a type of AI, because it automated what people did,” Markov said. “We are just moving along the straight line, and there's no stopping.”

For now, AI is acting as a force multiplier, Markov said, squeezing more output from teams rather than shrinking them. Who’s in those teams and what they bring is also shifting: engineers who are fluent with AI coding assistants are now in demand where they weren't six months ago.

Jevons’ paradox also looms large over the potential of AI in the chip design process. As AI makes certain parts of the process dramatically cheaper and faster, Nikolić expects engineers to use that freed-up capacity to explore territory they wouldn't otherwise have dared tackle, including the design of the AI chips driving the whole cycle in the first place.

After all, if any class of silicon is ripe for the kind of optimization that hasn't yet been systematically studied, Markov argues, it is the highly structured, performance-critical accelerators powering the current boom. “There’s plenty of opportunity for humans to be improving other parts of the design flow to make it more amenable to these AI-based systems,” said Karandikar. As models become more advanced, so too might their capacities to assist in chip design and development.

Memory and storage chip prices have been steadily increasing worldwide due to the AI-driven shortage. Still, Japanese consumers are being hit the hardest, with crazy prices appearing in multiple shops across Tokyo. According to Akiba PC Hotline! [machine translated], Samsung SSDs have increased by up to 300% in various computer retailers, with the top-of-the-line 8TB Samsung 9100 Pro going up as high as $3,470.97 USD (547,980 JPY) at the current exchange rate. By comparison, you can get the same drive on Amazon for just $1,960.69 — around 43% cheaper than what is offered among Japanese retailers.

The egregious pricing isn’t limited to large-capacity PCIe 5.0 drives either. The 4TB 9100 Pro with heat sink is priced at $1,773 USD (279,980 JPY), which is a 31.5% increase from its previous price, while the 2TB version is now at $893 (140,980 JPY). These examples are around double the prices of the same models here in the U.S., as you can see on our SSD price tracker. More “affordable” Samsung models like the 990 Pro, 990 Evo Plus, and even the 870 Evo didn’t escape the pricing hikes, either. When compared to their prices from January of this year, those drives have jumped in price by as much as 384.7%.

The Japanese news outlet also reports that the price hikes aren’t limited to Samsung drives, as Kioxia SSDs also saw recent price jumps of between 39.8% to 59.4%. This meant that the 2TB Exceria Pro G2 costs $594.68 USD (93,880 JPY), while the 1TB Exceria Basic now costs $208 (32,980 JPY). It’s not all bad news for Japanese buyers, though, as Western Digital (SanDisk), Lexar, and other less popular brands saw a price drop for Gen 4 and Gen 5 M.2 NVMe SSDs and SATA SSDs.

These are some of the highest price hikes we’ve seen across the world, so far. Unfortunately, it’s unlikely that this will be the last price hike we’ll see, as PC makers are reporting increasing component costs that go beyond RAM and storage. Even other storage products like memory cards and flash drives saw price increases between 124% and 261%. Unfortunately, the average buyer cannot do anything but either pay the higher prices or wait until they go down in the distant future (if they ever do).

Samsung’s production numbers reportedly plummeted significantly during a one-day strike by the company’s labor union. According to Seoul Economic Daily, citing union officials, the company’s memory fab output fell by 18% while its contract chip foundry plunged by 58.1%. These numbers only affected the night shift after the April 23 strike, with the “Joint Struggle Headquarters” saying that it will conduct a larger 18-day labor action if company management fails to reach a deal with workers. The union estimates that the walkout, which would last for more than two weeks, will cost the company KRW 30 trillion, or over $20 billion. It also threatened to mobilize personnel assigned to the fabs’ “safety protection facilities.”

The disagreement between the union and management stems from the company’s refusal to allocate 15% of its operating profit as a bonus for its workers, amounting to around $27 billion (about KRW 40 trillion), and would net chip fab workers around $400,000 each. Aside from this, the union also demanded a 7% increase in pay and a removal of the 50% bonus cap. The management made a counteroffer of a 10% operating profit bonus, 6.2% wage increase, and preferential mortgage loans, among other benefits, but it seems that this wasn’t enough for the union.

The group pointed out that SK hynix, Samsung’s biggest domestic chip rival, has given its workers a performance bonus amounting to 10% of its annual operating profit and removed the cap on the amount. This means that their bonuses only equate to less than 30% of what the SK hynix workers get, even though Samsung is 60% larger in terms of market capitalization.

Initial estimates from the police suggested that over 30,000 people attended the strike, but the union said that about 40,000 of its members were present in yesterday's action. This is a massive number and is said to represent nearly a third of the company’s semiconductor fab workforce. A general strike lasting several days would cripple operations, reducing Samsung’s advantage of being the first company to mass-produce and deliver HBM4 memory to its customers. It could also potentially exacerbate the global memory chip shortage, resulting in longer delivery times and potentially higher prices for everyone.

Unless the two sides come to an agreement, the union will kick off the general strike on May 21. Supra-Company Union Samsung Electronics chapter head Choi Seung-ho also said that he has submitted a notice to the Seoul Yongsan Police Station that the group will hold a rally in front of the residence of Samsung Chairman Lee Jae-yong on the same day.

Pat Gelsinger has shared the story of how his initials remained on the Intel 386 silicon die, despite them being spotted by the top brass during a pre-production design review session. Creating such inscriptions "was not done," during this era at Intel, remembers Gelsinger. Nevertheless, the legendary true-blue Intel man says he uttered “some complete nonsense about substrate tap configuration experiments” to swerve a comment on the 'PG' silicon markings by the gruff (then-CEO) Andy Grove. The end result is that Pat Gelsinger's initials are etched directly into the silicon of every 386 processor ever made.

The story goes that Gelsinger and his team of fellow architects and engineers were gathered in a conference room poring over “a huge 25x25 foot printout of the [i386] chip, magnified so we could see every little detail.” This was a part of the design review stage of a chip at the time.

During the review session, the team was excited by the arrival of Grove, invited by the youthful (~25) rising star Gelsinger. However, they grew apprehensive as the Intel CEO du jour took some time to review the detailed printout.

Spotting the famous ‘PG’ initials etched into the chip design, Gelsinger recalls that a gruff Grove grumbled, “What’s this?” People were expecting some kind of outburst or rebuke from the company’s then-leader.

Gelsinger suggests he quickly took the bull by the horns, and “responded with some complete nonsense about substrate tap configuration experiments for the optimal leakage current collection effectiveness (said in a tense but convincing tone).”

Gelsinger admits his response was “total BS.” He got the impression that witnesses to this verbal/technical joust were thinking “Pat’s dead!” due to Grove’s then-legendary temper. However, the then-CEO apparently just said “OK,” seemingly swallowing the explanation whole.

Thus, Gelsinger’s personal stamp of chip architect authorship remained on the i386, despite being a blithe ~25-year-old at the time. Subsequently, his initials were also on the i486 silicon. On the more advanced chip, you will see PG, again, as well as JR. The latter is short for ‘Johnny Reb, ’ fellow chip architect John H. Crawford’s nickname.

Grove’s perspective, Gelsinger's travels

Did the explanation about the substrate tap configuration experiments wash with Grove, or was he simply being lenient with a clearly budding talent at Intel? We can only guess about that aspect of this story.

Holocaust survivor Andy Grove passed away in March 2016. He is remembered for being one of the first three employees of Intel, alongside founders Gordon Moore and Robert Noyce. Grove also oversaw the rise of the x86 architecture, stewarding the chipmaker from a $4 billion to a $200 billion valuation.

Pat Gelsinger retired from Intel under difficult circumstances but is currently busy helping companies such as xLight, Gloo, Playground Global, and Snowcap Compute.

The United States Senate has just passed the 'GAIN AI' legislation as part of the National Defense Authorization Act (NDAA), which would require AI chipmakers like Nvidia and AMD to prioritize chip orders for American companies over export orders, particularly to China and its allies. As reported by Bloomberg, the bipartisan bill sailed easily through the Senate and is now under consideration by the House.

“Today, the Senate acted to make sure American customers — including small businesses and startups —aren’t forced to wait in line behind China’s tech giants when purchasing the latest AI chips,” Senator Elizabeth Warren (D-MA), who co-sponsored the bill, said in a statement. Senator Jim Bank (R-IN), who is the lead co-sponsor of the bill, also noted that this bill will bolster the U.S. competitiveness in AI and other cutting-edge industries while reducing exports to U.S. rivals, particularly China.

The House of Representatives has already passed its version of the NDAA, but the provision saying that chipmakers must put exports on the back burner in favor of local sales is missing. Because of this, the two chambers must now work together to create a compromise, but whether this rule will be part of the final version that will be signed by the president is still up for debate.

Nvidia has always criticized this bill, saying that its global sales “do not deprive U.S. customers of anything” and that the logic behind it was “based on doomer science fiction.” The company also said that it was “trying to solve a problem that does not exist” and that it would “restrict competition worldwide in any industry that uses mainstream computing chips.” Nvidia has said that its H20 shipments do not affect the supply of H100, H200, and Blackwell chips, which makes sense as these all use different parts, and that the supply of one will not affect the supply of the other.

Although China is Nvidia’s largest market next to the U.S., its China sales still dwarf in comparison. The United States accounts for nearly half of Nvidia’s sales in FY 2024, while China is only at 13% of its overall revenue. Aside from that, this might go down in the near future owing to the ongoing trade war between Washington and Beijing, especially as the latter has banned its biggest tech companies from acquiring Nvidia’s latest chips.

In a quiet Cornell University lab, researchers have taken a hammer to decades of digital circuit convention. The result is a silicon chip that thinks less like a clock-driven processor and more like a living brain—only instead of neurons, it uses controlled bursts of microwave energy.

Dubbed the “microwave brain,” this experimental processor can juggle two jobs at once: crunching ultrafast data streams and talking wirelessly, all inside a footprint small enough for a smartwatch. And it does it at just 200 milliwatts, a fraction of the power a comparable digital neural network would consume.

How it works

The secret lies in abandoning the step-by-step digital approach. Traditional chips march data through binary logic gates in sync with a clock. Cornell’s design instead pushes information through tunable microwave waveguides, letting patterns emerge and be recognized in real time at tens of gigahertz with no waiting or bottlenecks.

Each waveguide acts like a “physical neuron,” where the microwave signal’s amplitude, phase, and frequency can be shaped to represent data features. These features interact and interfere with each other in the analog domain, producing a rich set of patterns before the signal is ever digitized. This physical mixing and propagation essentially perform the feature extraction and transformation that digital networks usually achieve through multiple software layers.

The chip’s design builds a type of AI framework directly into the hardware, using the natural behavior of microwaves to process incoming data streams. Instead of storing values in memory and repeatedly performing huge numbers of calculations, it lets the microwave network itself handle the heavy lifting. Small adjustable components—such as electronic tuners and signal shifters—can change the pathways inside the chip on the fly, allowing it to switch between different AI tasks without having to retrain from scratch.

What it entails

President Donald Trump has revealed that he will announce new tariffs on semiconductors next week. The U.S. President said during a CNBC interview, “We’re going to be announcing on semiconductors and chips, which is a separate category, because we want them made in the United States.”

This is likely to be bad news for Taiwan, which is home to TSMC, the world’s largest semiconductor manufacturer. The White House has currently imposed a 20% tariff on Taiwanese goods, which is down from the initial 32% Trump announced as part of his ‘Liberation Day’ tariffs. However, many tech companies, including U.S.-based HP and Dell, oppose Washington’s plans to put tariffs on chips, saying it’s bad for business.

TSMC has already started producing chips in the U.S. at its Arizona site. However, the output of this single site is likely not enough for the demand of the entire United States, with its production already sold out through late 2027. This might be good news for its competitors, though, like Samsung, which had to delay its Texas fab because of a lack of customers.

Trump has made tariff threats on Taiwan-made semiconductors as early as January 2025, saying that he will impose taxes as high as 100% on chip imports from the country. However, this was soon overshadowed by his early April tariff announcements that practically taxed every trading partner that the U.S. has. Aside from this, the White House exempted computer chips from these sweeping tariffs, but promised to put taxes on these imports at a later date.

Although the President hasn’t singled out Taiwan for his semiconductor tariffs this time, it’s still going to negatively impact the nation, as that is one of its biggest exports to the U.S. We’re also unsure if the current production capacity of American chip fabs would be enough to meet local demand.

Demand for chips and semiconductors is currently growing, as everything from cars and computers to household appliances and switches now has at least one chip. Trump wants to rebuild semiconductor manufacturing in the United States, and applying tariffs is his way of forcing companies — American or otherwise — to put up fabs within its borders. Some industry leaders, like Nvidia CEO Jensen Huang, call it an utterly visionary plan; however, it might also hurt the average consumer, especially as America’s output might not yet be enough for its needs.

GlobalFoundries: GlobalFoundries has announced a $16 billion chip production expansion in the U.S., following increased demand from domestic customers. Read more

SpaceX: Elon Musk's SpaceX is planning to build its own advanced chip packaging factory in Texas with a 700mm x 700mm substrate size, purported to be the largest in the industry. Read more

DDR4 prices: A new report indicates DDR4 prices increased by 50% in the second half of May alone, as costs continue to soar. Read more

Chips Act: U.S. Commerce Secretary Howard Lutnick says the White House is renegotiating with Chips Act recipients. Read more

Chip news: June 5

Intel could be planning an Arrow Lake refresh

A new manual for an upcoming Intel W880 motherboard hints at a refresh for Arrow Lake dubbed 'Arrow Lake S'. Previous rumors have swirled that there could be another generation of Arrow Lake architecture chips on the way, with K and KF variants only. Details from the leak are scant aside from 125W TDP.

Sales of Arrow Lake generally have been stagnant, and even in the burgeoning AI industry, customers are turning towards older Raptor Lake CPUs.

Read more

TSMC targets UAE for chip manufacturing

Taiwan Semiconductor Manufacturing Corporation (TSMC) is reportedly in talks with White House officials to discuss the possibility of opening a chip fab in the UAE. According to reports, TSMC has met with U.S. special envoy to the Middle East Steve Witkoff and state-owned investment firm MGX. The plant would reportedly be of a similar size to its Arizona project.

Read more

Nvidia allegedly working on new China alternative to H20

Reports indicate that Nvidia is working on a new Blackwell-based alternative to its banned H20 chip for China. This B30 will reportedly feature multi-GPU scaling capabilities, which some have suggested could be NVLink. More likely, Nvidia may use its ConnectX-8 SuperNICs, currently deployed in its new RTX Pro Blackwell servers, which can feature up to eight RTX Pro 6000 GPUs.

Read more

Intel and SoftBank collaboration

Intel and SoftBank have reportedly set up a company called Saimemory, with a view to building a stacked DRAM substitute for HBM. The new prototype is reportedly based on Intel technology and Japanese patents, with a commercialization target before the end of the decade. SoftBank is also gunning for priority supply of these chips.

Read more

Qualcomm Snapdragon X2 Elite variant coming?

A seasoned tech tipster has stated that Qualcomm is testing Snapdragon X2 Elite devices in configurations with 64GB of RAM. Roland Quandt, best known for his exclusives published via Germany’s WinFuture.de, stated this weekend that the “SC8480XP aka SD X2 Elite [is] in testing with 64GB RAM.” He is also more convinced than ever that this upcoming Oryon V3 CPU architecture chip will come packing 18 cores.

Read more

U.S. extends Section 301 exemptions

The U.S. has extended tariff exemptions under Section 301, which include graphics cards and motherboards, until at least August 31. It means a 25% price rise that has been lurking behind some PC components for years. While a court recently struck down some of President Trump's tariffs, they'll remain in effect as the court battle over their legality plays out. Section 301 tariffs aren't included in the ruling, however.

Read more

TSMC quashes UAE expansion rumors

TSMC has quashed rumors that it is planning to build a chip fab in the UAE, telling members at its annual shareholders meeting that this wouldn't align with its strategy.

CEO C.C. Wei also addressed tariffs, saying slowing demand caused by increased prices could impact its business, but stated that AI demand was consistently strong and continued to outpace supply.

Read more.

EDA ban to hit Chinese companies

A new ban on the sale and licensing of semiconductor design software to China could hit the company hard when it comes to building more advanced semiconductors in the future. While a new FT report notes that China has usable software for the 7nm process and up, more advanced processes for use in AI aren't supported yet, possibly hampering China's longer-term semiconductor development.

Less advanced chips for smartphones and the like are not expected to be affected.

Read more

Next-gen Intel CPUs break cover

In Intel news, an official document from the company points to new Nova Lake-S/U and LGA 1700 Bartlett Lake-S chips in the company's pipeline. Not a confirmation by any means, it shows what Intel is at least targeting over the next year or two. High volume manufacturing is expected later this year for launch and sales in 2026.

Read more

Zen 6 spotted in AIDA64

Not to be outdone, fresh AMD leaks also point to the Zen 6 Ryzen range in AIDA64. The latest update for the software adds "preliminary support for next-generation AMD desktop, server and mobile processors," pointing to future support for the Ryzen 10000 series, if that's the branding AMD chooses to adopt.

Read more

TSMC's rising wafer costs

A new report indicates that TSMC is gearing up to begin its N2 (2nm-class) chips later this year, with prices of $30,000 per wafer. That would be a hefty increase on the $18,000-$20,000 for N3 currently used. Furthermore, the A16 node (1.6nm-class) could cost 50% more than that, an eye-watering $45,000 per wafer.

Read more

Nova Lake GPU overhaul

According to a new leak, Intel's upcoming Nova Lake chips will feature both Xe3 and Xe4 chiplets, going one step beyond the media and display unit chip separation found in recent chips.

Read more

GlobalFoundries announces mammoth U.S. spend

GlobalFoundries has announced a $16 billion U.S. chip production spend following a surge in domestic demand. $13 billion of the funds will go to its existing New York and Vermont fabs, a further $3 billion to advanced package research and other new technology.

Read more

Nvidia and MediaTek's hotly anticipated AI PC chips may not see meaningful shipment volumes to the mass market until the second half of 2026, according to a new report. As reported by Digitimes, Nvidia and MediaTek are expected to jointly unveil their new 'N1' Arm chips for Windows PCs at Computex.

According to Digitimes, the joint chip will "likely debut under the Nvidia brand," with both N1X and N1 models planned, echoing previous reports. Digitimes says that both companies are well into production ramp-up, however, states "insiders believe meaningful shipment volumes won't emerge until the second half of 2026."

The chips, which we expect to be unveiled at Computex next week, will mark the second offspring of Nvidia and MediaTek's newly-minted partnership, inaugurated by the GB10 chip unveiled earlier this year.

As the report notes, early N1X benchmarks hint at performance that lags behind some Arm-based PC chips, noting "the results have raised industry concerns."

Perhaps more worryingly, Digitimes reiterates reports that there are "unresolved integration issues with endpoint devices." These manufacturing headaches have previously been reported elsewhere, and could explain the hefty lead time of 2H26 on these chips.

Snag Intel's Core i5-13600KF this Black Friday at Amazon for a steal, slashed to its lowest price of $175, down from $304 (MSRP). If you're unwilling to update to Team Blue's latest Arrow Lake processors but still need a capable enough CPU to meet your needs, the Core i5-13600KF is a fantastic choice in terms of performance and affordability.

The Core i5-13600KF is based on Intel's Raptor Lake architecture - employing six Raptor Cove P-cores alongside eight Gracemont E-cores and twenty threads. The CPU operates at a base clock of 3.5 GHz but can turbo up to an impressive 5.1 GHz. With a TDP of 125W, going up to 181W (PL1/PL2), ensure you have beefy cooling on hand to tame this beast.

The CPU hosts 24MB of L3 cache and 20MB of L2 cache and does not ship with built-in graphics (iGPU), so you'll need a dedicated graphics solution. Regarding memory speeds, the 13600KF can support DDR4-3200 and DDR5-5600 RAM - but those are baseline JEDEC specifications and can go higher with XMP kits.

While the Core i5-13600KF is a great contender for productivity workloads, it certainly is no slop in gaming. Based on our extensive testing, the 13600KF trades blows with the Ryzen 7 7700X and ironically ends up faster than its direct successor - the Core Ultra 5 245K - even with optimized memory. Combined with cheap and readily available LGA1700 motherboards, the Core i5-13600KF should be a top pick if you want to upgrade.

Admittedly, the recent CPU degradation fiasco doesn't paint a good picture of the entire 13th and 14th Generation lineups. However, you can mitigate all these issues by simply updating to the latest 0x12B microcode. In any case, the Core i5-13600KF is an excellent option for various applications, offering support for technologies like PCIe 5.0, Thunderbolt 4, and Wi-Fi 6E (motherboard dependent).

We are working hard to find the best deals for you this Black Friday. If you're looking for other products, check out our Black Friday Computer Hardware Deals Live blog for a range of products, or dive deeper into our specialized SSD and Storage Deals Live blog, Monitor Deals Live, Graphics Card Deals, or CPU Deals pages.

Forget Lunar Lake; If there's one chip family that serves as the lynchpin for Intel’s entire turnaround plan, it's this – the company’s coming Clearwater Forest Xeon that it revealed for the first time at a recent event. The chip family is exceedingly important because this is the first high-volume chip to be fabbed on the Intel 18A process node, a node so critical that Intel CEO Pat Gelsinger has said he has bet the entire company on it.

Yes, Intel will make plenty of other processors with the 18A node, but mass-producing the Clearwater Forest chips on time is paramount to building confidence in Intel Foundry for potential customers, and that's the key to Gelsinger's entire turnaround plan. It also marks the culmination of Gelsinger’s audacious and now seemingly last-ditch effort to develop five nodes in four years to spark a resurgence at the ailing chipmaker.

Intel displayed the Clearwater Forest chip for the first time at its Enterprise Tech Tour event in Portland, Oregon, last week, but it won't come to market until sometime next year. Intel also unveiled its latest Xeon 6 Granite Rapids data center chips at the event, and they do look promising — Intel has now matched the number of cores in AMD’s competing data center processors for the first time since EPYC launched in 2017. 

However, while Xeon 6 comes with Intel’s competitive ‘Intel 3’ node, the underlying process node still lags behind the 3nm TSMC transistors that will power AMD’s EPYC Turin processors that launch next month.

Intel plans for its 18A process to overtake TSMC's tech in several key criteria, thus propelling the company to a historic comeback – and that’s where Clearwater Forest comes in. As the first chip fabbed on 18A in high volumes, it is critical not only for Intel’s products but also for its Intel Foundry subsidiary to become a contract chipmaker for other companies, much like its rival TSMC. Gelsinger’s plan ultimately hinges on making chips for other companies to sustain the cash flow it needs to continue developing leading-edge chip technology.

Earlier this month, Intel canceled its interim 20A process, a pipe cleaner node that provided a development platform for the disruptive tech that powers the 18A node. That means the onus is now even more so on 18A to save the company than when Gelsinger conceded, “I’ve bet the whole company on 18A,” in an interview with TechTechPotato six months ago.

In fact, things have only gotten worse for the company since that interview. Intel has now dropped to a mere $96 billion market cap while companies it routinely thrashed in the past have risen to new heights, like AMD, which now sports a $250 billion market cap, and Nvidia, which hovers just shy of a stunning $3 trillion cap. Amazingly, even memory vendors like SK hynix and Samsung are apparently poised to generate more revenue in the third quarter than Intel, a once unthinkable feat as the memory vendors capitalize on the AI boom.

Intel continues to axe programs – so many that I’ve lost count – and has recently begun to lay off more than 15% of its workforce, or 15,000 workers. It even issued yet another emergency restructuring plan last week that includes pausing several fab projects.

The situation has apparently become so dire that what was completely inconceivable a few short years ago, Qualcomm reportedly making overtures to take over Intel, now dominates the headlines.

Looking back over history finds Intel missing out on several revolutions, like rejecting the chance to make smartphone chips for Apple, killing its nascent Larabee GPU project that could have placed it in the driver’s seat for the AI revolution, and even reportedly passing on the chance to own a piece of OpenAI before the current AI hype cycle took hold, among others.   

But while those missed opportunities were painful, they obviously weren’t fatal blows. Perhaps none of the company’s missteps have brought it to its knees like its then-uncharacteristic delays to its 10nm node. That shortcoming weakened the company as AMD barreled in with suddenly class-leading TSMC nodes powering its chips in both the consumer and enterprise markets, slowly eating away at Intel’s market share and, perhaps more importantly, taking massive chomps out of Intel’s margins.

Intel’s stagnation has even helped breathe new life into Arm, which is arguably now becoming even more of a threat than AMD in the data center and consumer market. Meanwhile, TSMC has risen to be the predominant leading-edge chipmaker for the entire world, leaving the US in the tenuous position of not having a domestic supply chain capable of taking over if the long-simmering China and Taiwan tensions boil over into a war.

Because 18A is Intel Foundry’s premiere node, now the entire plan hinges on the transistors found in the Clearwater Forest chips that the company revealed at the event for the first time. If Intel can produce those chips economically, they will be the key to its turnaround, particularly because it will entice other companies to have their own chips made at Intel Foundry. If Intel can’t punch these chips out in high volume or is beset by delays, we could be looking at a future that doesn’t include an Intel that resembles its current form. 

As unbelievable as it sounds, it really has come down to this. For now, all signs point to Intel’s 18A being a healthy, production-class node. That's a good sign, as the market needs competition to keep prices in check, not to mention that, for the sake of national security, the US needs a domestic chipmaker punching out leading-edge nodes. 

Leading chipmakers TSMC and Samsung are discussing mega factory building projects in the United Arab Emirates (UAE), according to a report published by the Wall Street Journal (WSJ). The source report indicates that investments of $100 billion or more are being considered. However, neither TSMC nor Samsung confirmed any specific plans. To be clear, we aren’t talking about a TSMC and Samsung hook-up in the UAE; they would invest and build in the region separately.

According to the WSJ, top TSMC executives have recently visited the UAE to explore expanding manufacturing operations. Interestingly, the world’s largest contract chipmaker “talked about a plant complex on par with some of the company’s largest and most advanced facilities in Taiwan,” said unnamed people familiar with the matter. Projects like this could be worth more than $100 billion, according to the WSJ.

The claims regarding TSMC are undoubtedly eye-catching, if not a little superstitious, as the Taiwanese firm is well known for limiting its crown jewels, its top technologies, to facilities on its home turf. Even its newest factories in old and trusted allied nations like the USA, Germany, and Japan don’t tread on the toes of its most advanced facilities in Taiwan – so why should the UAE be any different?

In response to Reuters questions about its purported UAE plans, TSMC said it hasn’t announced anything new. Instead of addressing the question directly, it repeated a boilerplate statement about always being open to constructive discussion on the development of the semiconductor industry. 

Moving on to the Samsung rumors, we have a similar story. The source report says senior Samsung execs have also recently visited the UAE to discuss factory building plans. According to the WSJ, such projects may be funded by the UAE, with a central role for the Abu Dhabi-based Mubadala investment company. Plans are still in the early stages, and various technical and other hurdles must be navigated. According to Reuters, Samsung declined to comment when quizzed about its UAE plans.

The UAE has featured quite frequently in tech news in recent months. We have reported on the oil-rich Middle East being at the center of an influence war between the U.S. and China. Over recent months, the pull of the U.S. and the more advanced technologies of its allies seems to have won the region over. This was likely inevitable as countries like the UAE seek to diversify from fossil fuels into technology, with eyes on AI in particular.

Megafactories run by TSMC and Samsung inside the UAE would help cement the country’s position as a Middle East tech hub, delivering enviable access to the latest semiconductors. 

The best-known AI /tech company based in the UAE is probably G42. Earlier this month, this pivotal firm got the green light from the U.S. to import advanced Nvidia GPUs for its data centers, joining its Cerebras-powered systems. G42 is also known to have divested from China-backed projects and purged its data centers of Chinese parts. Perhaps the timing of these decisions signals that the doors to mega factory investments from companies like TSMC and Samsung are now open.

Unisoc, formerly Spreadtrum, a well-known developer of various application processors from China, has introduced what it claims to be the world's first security chip based on RISC-V, an open instruction set architecture (ISA). The E450R is said to offer significant performance for security algorithms, reports IT Home.

The key feature of the E450R is its asymmetric cryptographic PKE algorithm engine, which is said to work 50% faster than on the predecessor. Even when the key length is increased, the chip maintains its performance, ensuring robust security without sacrificing speed, according to the source report. The E450R also offers a 50% improvement in typical transaction-based applications, which makes the chip particularly suitable for high-demand environments where quick processing is crucial, IT Home claims.

In addition to its cryptographic enhancements, the E450R offers a 15% boost in the erase/write speed of its non-volatile memory (NVM). This improvement allows the chip to store more data within the same time frame, making it more efficient for applications requiring frequent memory operations. 

The thing that makes Unisoc particularly proud is the use of the open-source RISC-V ISA. This allowed the company to tailor its processing cores for its workloads, offering more performance at lower power compared to the traditional approach when using off-the-shelf Arm R that leverages a lot of legacy features to maintain ISA integrity, to ensure compatibility with a range of software. This takes die space and power, unlike the case of RISC-V. Given the fact that Unisoc provides an API and maintains the software/firmware stack, there is no problem with software (in)compatibility of the E450R, at least IT Home and Unisoc do not mention this factor specifically. 

Speaking of software, Unisoc claims that the chip's tailor-made microarchitecture reduces application code size by 30% and speeds up application load time by 120%, which is crucial for real-time applications.  

The E450R has already achieved several important certifications, including National Cryptography Level 2, UnionPay chip and embedded software security certifications, and the CCRC IT EAL4+ certification. 

The only question we now have is can China's secret services bypass the level of security that the Unisoc E450R offers? 

Nvidia dominates the market of AI processors controlling over 80% of sales, according to some recent estimates. But Jim Keller, a legendary designer of processors and current Chief Executive Officer of Tenstorrent, believes that there are markets not served well by Nvidia. As a result, there are opportunities for Tenstorrent and other developers of AI processors. 

"There are lots of markets that are not well served by Nvidia," said Jim Keller in an interview with Nikkei. 

Jim Keller has a truly distinguished computer industry history, making waves at AMD, Intel, and Tesla, and is now heading up the development of AI processors at Tenstorrent. His leadership at Tenstorrent aims to deliver affordable alternatives to Nvidia's expensive GPUs that cost $20,000 - $30,000 or more each. Tenstorrent’s business approach aims to serve markets that Nvidia does not adequately address, particularly in the edge space. According to Tenstorrent, its Galaxy systems are three times more efficient and 33% less expensive than Nvidia's DGX, which are perhaps the world's most popular AI servers.

Tenstorrent is on track to release its second-generation multipurpose AI processor by the end of the year, Nikkei reports, without disclosing the name of the processor. Based on Tenstorrent's latest roadmap from last fall, the company intends to release its Black Hole standalone AI processor and Quasar low-power low-cost chiplets for multi-chiplet AI solutions. 

The company claims its upcoming processors provide performance efficiency comparable to Nvidia's AI GPUs. The touted efficiency and lower cost is partly achieved by avoiding usage of high-bandwidth memory (HBM) and using GDDR6 instead, which is a logical thing for entry-level and mainstream AI processors mostly designed for AI inference. Meanwhile, Tenstorrent says that its architecture is less memory bandwidth-hungry than competitors, which is a key reason for its higher efficiency and lower cost. 

While Tenstorrent yet has to grab a significant part of the AI processor market, the company for now focuses on cost-effective yet scalable AI solutions that can address a wide range of applications currently not served properly by Nvidia. However, these markets are not going to be blue oceans as many companies will try to address them with their products in the coming quarters battling against the entrenched Nvidia. Rather than compete with Nvidia head-on, it will be much easier for new entrants to address niche markets that are indeed not directly catered for by the green team.

Researchers at the U.S Department of Energy’s Princeton Plasma Physics Laboratory (PPPL) are developing the next generation of chips that will be smaller, thinner, and more efficient. The research team, led by Shoaib Khalid, is investigating the use of transition-metal dichalcogenide (TMD) to replace silicon in processors.

In an interview with Tom’s Hardware, Shoaib said, “The chips are getting smaller and smaller, such that their limit is almost reached in terms of functionality and size.” Moore’s Law states that the count of transistors in semiconductors will double every two years. However, as the process nodes become smaller, Moore’s Law is expected to slow down to three years, instead of the usual two. After that, we will soon reach the physical limits of semiconductors as we start hitting the two-nanometer scale or smaller.

Shoaib and his team are investigating the use of TMD as a 2D material to replace the current 3D chip design. TMDs can be as thin as three atoms and act like a metal sandwich — the ‘bread’ uses chalcogen elements, also known as elements from the oxygen family. This could either by oxygen, sulfur, selenium, or tellurium. Any transition metal can then serve as filling for the ‘sandwich’.

Since TMDs are so thin, tiny alterations, such as a missing or extra atom in any layer, would impact the material’s performance. Although these alterations are called ‘defects’, they are not necessarily an unwanted phenomenon. For example, the presence of hydrogen during the TMD’s manufacturing process causes an excess of electrons. This gives the TMD a negative charge.

Shoaib said, “Depending on the type and nature of defects, they behave differently in a material and can change the properties of the material. For example, they can create excess electrons in the material, making it n-type (material with more electrons) or create more holes, making it p-type (material with more holes or positive charge).”

Computer chips use a combination of negatively charged (n-type) and positively charge (p-type) material to deliver better electrical conductivity. Current semiconductor technology uses doping to get these properties on purpose. By knowing how defects on TMDs are induced, we can then create these n-type and p-type material, but on the atomic scale.

The principles behind silicon semiconductors and TMDs are the same, but the latter have several advantages. Shoaib says that TMDs have tunable bandgaps controlled by changing the number of layers; they can be as thin as a monolayer, which is just three atoms high. You can use different materials to create them and they are flexible but durable. “In the end, the goal is to have smarter and cheaper chips,” say Shoaib.

The problems that Shoaib and his team are tackling today are likely the same ones that the researchers working on the first semiconductors were facing 50 years ago. However, they were working at a bigger scale, size-wise. The PPPL team is solving the same issue, but at an atomic level.

So, how soon will we see TMDs inside our phones and PCs? TSMC is already looking at 1nm chips, with other companies introducing a roadmap to sub-nanometer transistors by 2036. “Companies like Intel are already working on making transistors from these TMDs,” says Shoaib. “I think, by 2030, we might have an actual TMD transistor that can be used in devices.”

Former ASML boss Peter Wennick has shared his belief that the trade war over semiconductors between the U.S. and China may go on for the next few decades. Nicknamed the "Chip War," Wennick believes that the offensive exists exclusively for ideological reasons, rather than any basis in fact. 

TrendForce first reported on the comments made by the former ASML boss on Dutch radio outlet BNR. Wennick is something of an expert on the Chip War thanks to his time as CEO of ASML, the world's largest vendor of chipmaking tools and equipment. Retiring in April of this year, Wennick served as CEO for 10 years, during which time ASML grew to become the third-largest European company by market cap.

ASML's two largest customers are Taiwan and China, with ASML providing TSMC and Samsung with EUV lithography supplies for both major companies' 7nm and below processes. While EUV lithography tools are prohibited for sale to China, the Chinese semiconductor industry remains ASML's second-largest customer, with ASML having operated in the country for 30 years. Only a small portion of ASML's customer base in China is heavily affected by the ever-tighter U.S. export bans, with the most recent round of export restrictions in April stopping ASML from being able to service high-end products already in use in China.

As CEO of ASML, Wennick claims he maintained a line that served primarily to benefit his stockholders and executive leadership, favoring neither the U.S. nor China in their skirmish, but advocating with both sides to play nice and follow the law. Wennick highlighted conversations he had with both Washington and Beijing to lessen U.S. export bans and raising China's awareness of intellectual property violations. This may have caused both governments to suspect him of allegiance to the other side, he says. 

Wennick's fears that the Chip War, due to the geopolitical mess wrapped up in the potential for profit in the semiconductor industry, may continue for decades. ASML has had an outsized impact on global discussions around the Chip War, with its home country, the Netherlands, becoming China's line of communication to the West regarding ever-increasing sanctions. 

Tan Jian, China's ambassador to the Netherlands, shared with NRC in January, "If the Americans treat us in a hegemonic way, we will of course react, but our relationship with the EU should not be affected." President Xi Jinping went on to meet with Dutch Prime Minister Mark Rutte to request that Netherlands stop following U.S. restrictions on ASML hardware, saying "The Chinese people also have the right to legitimate development, and no force can stop the pace of China's scientific and technological development and progress." How Chinese companies like SMIC and Huawei will continue to advance their chip processes without access to ASML's EUV tools remains to be seen. 

An engineering sample of the upcoming 16-core Zen 5 AMD Ryzen 9 9950X 'Granite Ridge' processor appears to have been benchmarked in AIDA64. The results were shared by Anandtech Forum user igor_kavinsky, who cites another user as the owner of the CPU, and credited himself as simply sharing the info in his original Zen 5 speculation thread. 

In the original AIDA64 interface screenshots, the test notes the score of "16x Granite Ridge HT" came just under "32x Ryzen Threadripper" in both FP32 and FP64 results. The engineering sample was also tested with a Dual-Channel DDR5 RAM configuration running at 8,000 MT/s, with timings given of 34-45-40-42. 

AMD Ryzen 9 9950X Leaked AIDA64 Benchmark Scores

In the original thread posting, these results are noted as being a 39-45% improvement over Ryzen 9 7950X, a 55-60% improvement over Intel Core i9-13900K, and a mixed improvement over the Ryzen Threadripper 7975WX. The mainstream results are a straightforward victory for AMD's upcoming Ryzen 9 9950X if they prove accurate, but some other factors come into play when compared to core-dense HEDT rivals.

Ryzen 9 7950X's sole win over Threadripper in this set of tests is in AES encryption scores, which are boosted by 11% over the 32-core Threadripper 7975WX, but FP32 performance drops by 13% and FP64 by 16%. The FP16 and FP32 results are marking some loss in floating point performance, but we are still seeing a mainstream Ryzen 9 CPU compared to a platform-rich (PCIe lanes, RAM, etc) Threadripper.

Another interesting discussion about these benchmark results in the original thread is that the testing used AVX-512, which could be attributed to these ~40% range gains. AVX-512 is something of a miracle performance solution when it can be properly applied to a workload. For example, AVX512 is noted by RPCS3 developer Whatcookie as one of many factors making extreme Sony PlayStation 3 emulation optimizations possible on modern PCs, after its original Cell processor co-developed by IBM was compared to supercomputers of the time.

For more recent Ryzen 9000 benchmark leaks, please check out our Ryzen 5 9600X CPU-Z test score leak analysis. We will see the first of AMD's new generation Ryzens become official in July.

Huaxia Chip was a rising star among China’s original computer processor intellectual property (IP) designers. However, despite a voracious appetite for all things IT, the company now appears to be bankrupt and set to be liquidated. Obviously, something must have gone wrong, and machine translations of various Chinese news reports hint that the courts will find “dishonest persons subject to execution.” Don’t worry though, as we looked more carefully at the original wording and translations, and the people involved will more likely face consequences such as equity and trading freezes – nothing fatal.

It is highly likely that China’s Huaxia Chip was an important part of the country’s homegrown computer chip industry. It claimed to have completely independent intellectual property rights to CPU, DSP, GPU, and AI processor IPs. That is a bold and expansive claim, but its specialism seemed to be in robotics, edge, IoT, and cloud computing fields. Huaxia was placed in the top five Chinese innovative IC design companies in the 2020 China IC Design Achievement Award and it won the 2020 Hard Core China Chip Most Potential IC Design Enterprise Award.

The firm maintained its headquarters in Beijing and R&D and sales centers in Shanghai and many other China tech hub locations. However, looking at its English-language website, the company became relatively quiet after 2018.

With all the above in its favor and China thirsty for homegrown tech in the face of ratcheting U.S. sanctions, it is hard to understand how Huaxia Chip fell from grace. The source reports and court statements aren’t very enlightening. Huaxia seems to have folded after being challenged by Xi'an Jiubufang Enterprise Management Partnership's lawsuit. We know that, for years, an alleged graft scandal has plagued Chinese business development, and juicy government-backed investment funds targeting the semiconductor industry are ripe targets. However, we don’t have enough information to say for sure if such underhand business practices were to blame for the bankruptcy of Huaxia.

Reading the various reports, people involved in Huaxia’s business seem to be under close scrutiny, have had their assets and equity frozen, and will not be able to trade stocks. We guess this could change subject to the outcome of the court case. Huaxia Chip looks set for liquidation, and the government will probably be keen to get this process over with ASAP, as the firm holds more than 100 patents which could be important to the country’s semiconductor and chip development hopes.

Microsoft is putting all of its weight behind Qualcomm's latest Snapdragon X Elite, and expects it to outperform Apple's competing M3 silicon — at least in certain workloads. The Verge reports that the trillion-dollar software and hardware behemoth is extremely confident in Qualcomm's new CPU and is planning several demos for its upcoming AI PC event in Seattle that will show the Snapdragon chip outperforming Apple's M3-equipped MacBook Air laptops.

These demos Microsoft is putting together supposedly won't just show the Snapdragon Elite X beating the Apple M3 in a single metric, but will show superiority across several workloads — including AI acceleration and app emulation. In fact, Microsoft claims that Windows devices sporting the X Elite chip will have faster app emulation than Rosetta 2, the compatibility layer Apple uses on its ARM-based CPUs to run x64 apps that don't have an ARM flavor.

Unsurprisingly, Microsoft is also reportedly planning to use the Snapdragon X Elite in its upcoming Surface Pro 10 and Surface Laptop 6 rather than Intel's competing Core Ultra Meteor Lake processors. However, these chips will be for consumer versions of the Surface devices. We already know that Microsoft's business counterparts will be sticking with Intel Core Ultra CPUs.

There's plenty of evidence to back up Microsoft's confidence. The Snapdragon X Elite is proving to be one of the most capable ARM-based consumer CPUs we've ever seen outside of Apple silicon. Several leaked Geekbench results already indicate the Snapdragon X Elite is on par with and even outperforming Intel's previous generation Core i7 and Core i9 Raptor Lake mobile CPUs, as well as AMD's latest Ryzen 9 8945HS flagship.

Qualcomm has also unveiled impressive performance figures for its new ARM CPU, even going so far as to show it outperforming the Apple M3 by a noticeable margin in Geekbench. Qualcomm senior VP and GM of compute and gaming at Qualcomm Kedar Kondap told us that the company believes its work on the X Elite is a pivotal moment, going on to say that the performance of the Snapdragon X Elite is something we haven't seen before in terms of leadership, technology, and power efficiency.

The Snapdragon X Elite is built using TSMC's 4nm process and is equipped with 12 Oryon performance cores that turbo up to 4.3GHz. There are no efficiency cores in the chip, and it's rated at up to 4.6TFLOPs of GPU performance and 45 TOPS of AI performance from its neural processing unit (NPU). Qualcomm claims the chip has 2X faster multi-threaded performance than Intel's i7-1360P and Apple M2, while consuming just one-third the power.

We'll have to wait and see if Microsoft's demos truly deliver what the company wants to accomplish. It's a safe bet Microsoft's will be able to show superior performance to the M3, but the question will be whether the demonstrated performance will extend beyond the specific use cases and apply to standard workloads. Microsoft's AI PC event reportedly starts next month in Seattle.

Phillip Tennen, developer of the open-source axleOS, has recently decided to use what he learned from that project to create an IRC client that runs entirely within the UEFI pre-boot environment, with no operating system required. This "UEFIRC" is nearly fully functional, with a graphical interface and a TrueType font renderer, and it's all written in the Rust programming language.

For those unfamiliar, "UEFI" is just the modern term for "BIOS" since UEFIs replaced those long ago. You'll often see both referred to as BIOS for simplicity's sake. They both serve the same purpose as the first piece of software any PC runs before it can boot into an Operating System, where virtually all your work will be done. All the extra features the UEFI interface offers, including complete mouse control and network access, lend themselves to ease of use and even high-end (if questionable) applications like these.

However, according to Tennen's write-up, UEFIRC is entirely unnecessary. It's even an elaborate joke! Maybe it was meant for April Fool's, though that might have led people to believe it wasn't real, so it may be for the best that he waited.

A note says, "I told a friend I was making a joke project, then explained it. She said she wasn't sure when to laugh. I'm not sure either." While UEFIRC isn't recommended for mainstream use (and is referred to as not "that usable"), it does function well enough for test footage to exist. The header is Phillip entering an IRC room for UEFI development to announce his discovery (to the immediate confusion of the first replier), and below is video test footage with some basic IRC commands.

More exhaustive documentation of achieving this is available at the source for those who want to learn the intricacies of forcing the scrolling text to function inside a UEFI. There are also some other unexpected UEFI applications out there in the wild, like Mitnal, a UEFI Twitter client.

While the ongoing Chip War between the United States and China has seen plenty of negative outcomes for Chinese and US customers alike, the struggle is also contributing to a massive boost in Malaysia's semiconductor manufacturing, reports Financial Times. 

Since United States regulations are pushing many semiconductor manufacturers away from China — especially high-end US tech companies that rely on that Chinese manufacturing power — they are being forced to look elsewhere on the globe for land and labor. Some of the manufacturers pouring into Penang, Malaysia are familiar faces, including the likes of Intel (which has been present in a Penang rice field since 1972 according to FT), Micron, and more. The turn in the market has reached the point where Malaysia now leads the world in US chip imports, as of February 2023.

The rush of foreign investments into Malaysia has come in the tunes of multi-billion dollar ambition from major industry players, like Intel. By itself, Intel will spend a whopping $7 billion on new, Malaysian chip assembly and testing facilities. The overall total of foreign Malaysian investment in 2023 was $12.8 billion, and that exceeded its seven-year combined total from 2013 to 2020.

As one may expect, the Prime Minister of Malaysia, Anwar Ibrahim considers the development of Malaysia's semiconductor industry and workforce into higher-value manufacturing a "critical goal", according to Feb. 24 FT interview. 

Interestingly, the "foreign investments" certainly aren't limited to US companies: other major nations are getting their tech built in Malaysia too, with China also being one of them. This seems to make Malaysia an interesting meeting ground for buyers and sellers at the highest level of the international semiconductor and electronics market, though there is some speculation in even the Financial Times piece that US scrutiny could eventually put a damper on this.

What Malaysia seems to need more than anything else to maintain its major player status is a front-end semiconductor manufacturing plant. Trade Minister of Malaysia, Zafrul Aziz, asserts as much in the original Financial Times piece, concluding "I am optimistic we will attract more than one. All it takes is one to kick-start a wave."

Penang, Malaysia is a north-western, two-part state of Malaysia composed of the city Seberang Perai and the separate Penang Island, which is separated from the other half by the Penang Strait. Penang Island also holds the capital, George Town, which is Penang's second-largest city by population.

Meanwhile, Seberang Perai is the third-largest city in Malaysia and the site of many shipping ports, rice fields, and so on. Comparatively, the Penang Island is a little more urbanized, but still a major economic hub in its own right. Based on its geography and available resources, Penang seems like the ideal place for many foreigners to start doing business in Malaysia.

Qualcomm's upcoming Snapdragon X Elite came out ahead of Intel's Meteor Lake in a head-to-head comparison running Windows, shared by YouTuber Erdi Özüağ (via TechPowerUp). The Snapdragon X Elite X1E80100 beat the Core 7 Ultra 155H in all four benchmarks that were run, with two wider victories and narrower wins. The results are a good omen for the Snapdragon X Elite, though it's important to note the tests were in a controlled environment and definitely warrant at least a few grains of salt.

The performance demo was conducted at an official Qualcomm event during Mobile World Congress last month, but Özüağ's video seems to be the first time we've seen these benchmarks. Based on the language in the video and the fact that nobody else has disclosed these figures in the two weeks since MWC, we get the sense that maybe Özüağ wasn't supposed to show this stuff publicly — or perhaps Özüağ was just working on the video for the past several days and nobody else found the benchmarks interesting enough to talk about.

Both the Snapdragon X Elite and Meteor Lake laptops shown in the comparison were made by Qualcomm, and are the same models we've seen used for every other official comparison, such as this NPU performance test. The Snapdragon X Elite laptop was equipped with the 12-core X1E80100 CPU running at 28 watts and 4.3 GHz, and was paired with 64GB of RAM. The Intel laptop used a Core Ultra 7 155H, presumably (though not certainly) with the same RAM and power limit.

Qualcomm's reference laptops were set up to demo performance in 7-Zip file compression, Visual Studio code compilation, and 3DMark's Wild Life Extreme GPU test. The X1E80100 barely edged out the 155H in 7-Zip and was about 46% faster in Visual Studio, which are CPU-bound benchmarks. The two chips were tied in Wild Life Extreme, showing that Qualcomm can match Intel in integrated graphics.

Özüağ also showed performance data from UL Procyon, which tests NPU performance, and there the X1E80100 beat the 155H by a whopping 460%. That's an even bigger performance gap than Qualcomm showed in Stable Diffusion in a different MWC demo, though of course it also raises the question of whether the Intel system was properly optimized for the benchmark.

This all fits in with what we've seen from the Snapdragon X Elite in previous performance previews against both Intel and AMD processors. As with all pre-launch figures, some skepticism should be applied, especially when Qualcomm is making the reference laptop used to represent the competition and is showing just a handful of benchmarks in a controlled environment. It's clear that the Snapdragon X Elite is fast, but we don't have the full picture quite yet. We'll find out how it performs in a wider selection of benchmarks when it launches later this year.

Quantum computing specialist EeroQ recently announced the successful tape-out of its Quantum Processing Unit (QPU) chip. Codenamed “Wonder Lake” (which makes it sound like someone’s been paying attention to Intel’s codenames), Eeroq’s QPU was taped-out at a US semiconductor manufacturing foundry. Due to using a CMOS (Complementary Metal-Oxide Semiconductor) manufacturing approach that drinks deeply from standard chip manufacturing knowledge, the company expects its helium-based qubits to turn out far more scalable (and thus sustainable) than other qubit manufacturing approaches.

Qubits are units of computation in the quantum realm which are expected to unlock orders of magnitude more processing power in specific tasks, such as optimization problems, material physics, chemistry, and others.

“This scaling architecture has passed the rigorous design checks required for compatibility with today’s standard chip manufacturing process (CMOS),” said Nick Farina, EeroQ CEO, in a blog post.

Crucially, Wonder Lake’s qubit count is within the higher range we’ve seen yet; at 2,432 helium-electron qubits, it’s one of the most densely populated QPU designs. Initially proposed in 1999, EeroQ’s quantum technology is based on quantizing isolated electron spins suspended above pools of liquid helium (eHe).

Quantizing here refers to "turning into qubits," - meaning that a particle or pre-existing material has been harnessed and has become an available computation unit. In this case, the technology takes advantage of an effect known as "Rydberg states," which translate the motion of the suspended electron (a property known as spin in physics) into computable representations of the 0s, 1s, and everything in between that's allowed by quantum computing.

The way EeroQ's Quantum Processing Unit (QPU) is fabricated is promising as well: it taps into the decades of expertise the semiconductor industry has had with CMOS technology. In the vein of what Intel is trying to do with its Tunnel Falls QPU, this approach allows the company to tap into a well-understood technology that serves as the basis for its technology. The fabrication process also sounds deceptively simple: wafers etched to EeroQ's specifications pass through the company's labs, where a layer of liquid helium is applied, and electrons are deposited onto purpose-etched reservoirs. With a small magnetic bump, the electrons floating above the helium layer (held in place by the CMOS reservoirs) can initialize their spin states. After that, it's just a matter of firing up whatever quantum workload can fit within the chip's circuits. According to EeroQ, using CMOS technology will eventually lead to fabrication-related quantum gate errors to a mere 0,01%. Let's call it quantum yield.

Of course, not all qubits are the same, so these can’t be compared to IBM and others’ superconducting qubits. For its part, Eeroq says its helium-electron qubits provide an extremely high 10+ second qubit coherence timeout with a high order of qubit connectivity - meaning that more complex qubit circuits can be built to accelerate pre-existing workloads or process new ones. Further, EeroQ says its qubits’ mobility across the helium layer gives them a 50% reduction in overhead for error correction mechanisms to be applied.

Error correction is considered the holy grail of quantum computing at the moment, and serious work is being done in the field of error mitigation that we hope leads towards error correction - that EeroQ uses this specific wording is relevant. That said, the company is still looking to extract actual utility from its CMOS-based QPUs; at the moment, they still haven’t demonstrated their two-qubit gate design, a necessary stepping-stone for a post-NISQ (Noisy Intermediate-Scale Quantum) future.

As we’re still in quantum computing’s infancy, it’s not a discussion on the quality of qubits and which technology is best; it’s instead a recognition that there are vast differences between qubit approaches. But it seems Wonder Lake’s call to fame doesn’t just sit there; the company called attention to how efficient their chip is: a design choice carved in stone as early as qubit choice and engineering design.

“There are two particularly challenging parts to making a useful quantum computer: high-quality quantum gates and a path to scale,” Farina writes in EeroQ’s blog. “With our latest work, we are proud to join the leadership ranks on scalability. Together with recent advances in error mitigation and more efficient algorithms, we can see the commercial quantum future coming together sooner than expected – led by the ability to leverage our architectural advantage to scale rapidly.”

As EeroQ puts it, their main advantage is that they thought about their quantum computing tech in reverse, focusing on achieving a many-qubit interaction before attempting to extract utility from the limited computing resources (usually one or two-qubit gates) they could physically explore. That focus on scaling allowed the company to build a quantum solution that requires only 30 control lines per chip - a marked reduction in the control complexity needed, for instance, on superconducting qubit systems. That, in turn, will allow for cost savings in computing area cost and how expensive the control system is.

GlobalFoundries, the semiconductor manufacturing specialist spun out of AMD's now-defunct chip manufacturing unit, has voiced worries regarding the German government's distribution of a $22 billion ($20 billion) subsidy that's part of the (now approved) $47.4 billion European Chips Act. According to available information, Intel is the number one recipient of the grant by a large margin, guzzling around 50% of the contended $20 million to be allocated to Intel's new Magdeburg Mega Fab complex. Intel, by the way, already tried to expand its slice of the subsidy pie, but was denied.

GlobalFoundries' argument is that the subsidy amount is big enough that it can distort other companies' ability to compete. It's relatively easy to see the argument: Intel's Fiscal Year 2022 revenue stood at $63.1 billion; GlobalFoundries' stood at just $8 billion. And with another $6 billion of the subsidy's $20 billion tranche being reserved for Taiwan's TSMC and Germany's own Infineon, that means all other semiconductor players will have to share a $4 billion budget.

As GlobalFoundries CEO Tom Caulfield told the Financial Times, "there is a real risk of dependence on a single supplier, market foreclosure and less resilient supply chains as a consequence."

Policy decisions such as these are always hard to execute correctly; there's no arguing against GlobalFoundries' assessment that subsidies of this magnitude can distort competition. But then again, there are reasons for a company such as Intel to receive a bigger slice of the subsidy pie: Intel's more cutting-edge tech naturally has higher R&D  and installation costs, and its facilities operate at a scale that's different from what GlobalFoundries could possibly achieve with its current financial results and product portfolio (despite the fact that the company has a relevant and unique European positioning in fields such as optical and RF technologies). That said, it's important to get these injections right.

It is, however, also of note that GlobalFoundries hasn't been completely shunned by these semiconductor subsidies; the company (alongside European company STMicroelectronics) recently stood at the receiving end of an $8.2 billion injection courtesy of the French government. That's almost half of the company's yearly revenue being delivered to open hands. This funding was also approved under the umbrella of the European Chips Act. 

According to the Financial Times, Saxony's Economy minister, Martin Durlig, waved GlobalFoundries' criticism with a simple "there will be enough room for all providers to thrive." Hopefully, $47.4 billion will be enough.

In a report published by the Semiconductor Industry Association (SIA) today, it was estimated that there will be a shortfall of 67,000 workers in the industry, by 2030. Key roles that will be difficult for US-based chipmakers to fill include technicians, computer scientists, engineers. Thus, SIA has devised a three point action plan to address the talent gap.

The US has big ambitions for its domestic semiconductor industry, even going so far as encouraging foreign investment from the likes of Samsung and TSMC. The enactment of the landmark CHIPS and Science Act of 2022 means that the US will be responsible for a significant share of global semiconductor output, going forward. However, the problems we are already seeing, regarding the lack of qualified and interested staff, look set to get worse.

SIA expects the semiconductor industry’s workforce will grow by nearly 115,000 jobs by 2030. That means the total workforce of roughly 345,000 people today should expand to 460,000 by the end of the decade to keep pace. The labor market is going to struggle to keep up, though. Making clear the core issue, SIA says that current degree completion rates mean that roughly 67,000 - or 58% of projected new jobs - will go unfilled.

This insufficient qualified labor problem isn't isolated to the semiconductor industry in the US. Looking at the wider jobs market for a moment, SIA says that by the end of the decade an estimated 3.85 million additional US jobs requiring proficiency in technical fields will be created. Of these about 1.4 million are at risk of being unfilled. These are huge numbers, and for industries like semiconductors, which have been judged to be one of strategic importance to the country, the issue is most concerning.

So, what is to be done to address the gap between the number of skilled chip industry workers expected to be needed, and the number that are expected to become available? SIA has a three point plan, and we have quoted the bullet points verbatim, below.

• Strengthen support for regional partnerships and programs aimed at growing the pipeline for skilled technicians for semiconductor manufacturing and other advanced manufacturing sectors.
• Grow the domestic STEM pipeline for engineers and computer scientists vital to the semiconductor industry and other sectors that are critical to the future economy.
• Retain and attract more international advanced degree students within the US economy.

Hopefully, the above will be heeded by national and local government, and key industry players, and prove to be enough to help fill the talent gap - or if not it at least make things more manageable.

Some of the news coming out of TSMC's big venture into the US, propelled by the CHIPS Act, isn't very encouraging for those who might be interested in working in the industry. For example, in June, we reported on TSMC chairman Mark Liu saying workers in the industry should expect long shifts, and have their passion for the industry tested by the work environment. This came after news that TSMC was struggling to hire local workers in Arizona, so was seeking more Taiwanese migrants to fill the gaps.

Late last month, the Dutch government restricted sales of ASML's advanced immersion lithography tools to China. But it looks like ASML is facing further high-tech export constraints from the Dutch and United States governments. The new rules may restrict ASML's ability to service existing production tools in China, according to a Bloomberg report that cites sources familiar with the matter. 

The upcoming Dutch export control rules will require ASML to obtain government authorization to service, repair, and supply parts for its dry and immersion lithography tools that are already used by its customers in China. Services, repairs, and spare parts represent an important source of ASML's income. The company has previously earned some $5.743 billion — 27% of its revenue — servicing its tools in the field.

Without proper servicing and repairs, fab equipment will get out of order, and this could hurt companies like Semiconductor Manufacturing International Corp. (SMIC) even more than the lack of new tools. It's possible they will not be able to fulfill existing orders at all or on time.

In addition, the U.S. is also expected to impose restrictions preventing ASML from selling older versions of its wafer fab equipment to six Chinese fabs without an export license from the U.S. One of the targeted fabs includes a plant run by SMIC, according to Bloomberg. The Biden administration is expected to enforce the new restrictions through the Foreign Direct Product Rule, limiting sales from foreign suppliers if their products contain American components.

These dual restrictions add to the extensive export controls previously imposed by the U.S. government in October 2022, to undermine Chinese semiconductor firms' competencies over time. Later on, the governments of Japan and the Netherlands imposed similar restrictions. Naturally, these curbs also hit makers of wafer fab equipment from the U.S., Japan, and the Netherlands.

The upcoming existing Dutch export regulations, due to take effect on September 1, require acquiring an export license to sell the Twinscan NXT:2000i and more sophisticated scanners to Chinese firms. These scanners utilize deep ultraviolet (DUV) lithography and are capable of fabricating chips using 7nm and 5nm class process technologies. The Chinese government was outraged by the new rules, and for a reason.

"The Dutch regulations are an official stake in the ground that the Netherlands does not want China to be able to produce the most advanced semiconductors," said Dylan Patel, founder of the research group SemiAnalysis. "Without access to these machines, Chinese chip companies will be at a significant disadvantage in the global market for advanced logic and memory chips. 

Patel predicts that the new restrictions will significantly affect the future plans of Chinese chip companies, potentially leading to delays or cancellation of certain future leading-edge process technologies.

TrendForce, the high-tech market intelligence firm, has published an update on the NAND flash market. Its latest forecasts see NAND flash's average selling price (ASP) falling between 3 and 8% in Q3 2023. However, for PC enthusiasts and DIYers, the news is even better. Client SSD pricing is expected to fall between 8% and 13% during the same period. In other words, those in the market for PC SSD storage could perhaps save an extra 10% or more on today's prices if they wait a few weeks.

NAND Flash chips are a major component in PC SSDs, products like smart devices (UFS and eMMC) and memory cards. The pricing of this essential storage device component will continue to fall significantly in the current quarter, according to the analysts at TrendForce.

As usual with its market intelligence reports, TrendForce provides some background and reasoning for its predictions. Currently, NAND makers are "making concerted efforts to scale back production," notes the analyst firm. Manufacturers thus hope to create artificial scarcity and pump up NAND pricing. However, TrendForce reckons that these efforts will be in vain and aren't enough to address the state of oversupply in Q3 2023 – even with the traditional seasonal demand surge expected in the coming months. TrendForce predicts pricing of NAND and NAND-based products to slide.

Looking closer at the segmented figures TrendForce provides, we see that Client SSD pricing looks likely to improve – for customers, not NAND producers. Last quarter was a great period for PC enthusiasts, DIYers and upgraders, with SSD ASPs sinking between 15% and 20%, according to the tabulated data. The continued drop of between 8 to 13% is also well worth waiting for if you are looking to do a new build or for an upgrade and you aren’t in a mad rush to complete it.

If you follow our news and deals, you will know there have been plenty of SSD bargains in recent weeks. We can be pretty confident that the deals are set to continue, probably improving, based on TrendForce data.

With NAND ASPs dropping, it isn’t only PC SSDs (AKA Client SSDs) that will be more keenly priced. TrendForce outlines how products like Enterprise SSDs, mobile storage (eMMC and UFS), and other devices have been impacted and are predicted to react. So, for Tom’s Hardware readers, the expected drops in smartphone and tablet components might be welcome to reduce upward price pressures on such products.

Last but not least, TrendForce doesn’t predict NAND ASPs will continue to drop steeply after Q3. It expects to see a rebound in high-end 3D NAND pricing first but doesn’t give any firm prediction on whether NAND ASPs will continue to fall or rise in Q4 2023.

We already knew that Nvidia Founder and CEO Jensen Huang and AMD President and CEO Dr. Lisa Su were related somehow. It is also a matter of record that they were both born in Tainan, Taiwan. Now we have a much clearer picture of how closely related the pair of PC tech titans are, thanks to the work of senior media person and family heritage researcher Wu Jia Jin, host of Jin Chat.

Nvidia founder and CEO Jensen Huang's fame has skyrocketed in recent months. His newfound 'rock star' like fame, as witnessed by thronging crowds at the recent Computex 2023 show, is most apparent in his country of birth, Taiwan. Is it because of the success story of Nvidia, the business focus of ordinary Taiwanese news broadcasts, or that Nvidia burst through the $1 trillion valuation milestone? It is probably a mix of these things, and now it means Huang is getting a lot more media coverage, both business and personal.

Circling back to Huang's family connection with AMD boss Lisa Su, back in 2020, we thought that they weren't very close relatives. Reinforcing this belief, Lisa Su told CTA boss Gary Shapiro in a Masters of Leadership webinar that she and Huang were "distant relatives, so some complex second cousin type of thing." The truth may be they are quite a lot closer than that. Imagine if one of your cousins has a child — that is basically the relationship between Huang and Su, if the published family tree is accurate. We would thus say they are pretty close relatives, and it isn't that complex to understand.

Su may have felt the relationship was more distant and complex due to how close, or distant, certain segments of the family were. Remember that Jensen Huang's grandfather (Su's great-grandfather) had at least 12 children, which would imply lots and lots of cousins not added to the family tree. Moreover, we know that Huang's family emigrated to Oregon, US, when he was just four years old. Similarly, Su's family emigrated to the US when she was just three years old. There isn't just the likelihood of a great mass of relatives which are hard to keep track of, the family was also scattered geographically wide when the CEOs were at a very young age.

So, how would we describe the relationship between Nvidia's Huang and AMD's Su? In English family relationship terminology, they could be described as first cousins once removed. Su could also say that Huang is her second uncle. However, the naming of relatives in Mandarin Chinese and Taiwanese Hokkien is much more complex, depending on the male/female branch of the family tree, gender, birth order, and so on.

Other than the pinning down of the precise family relationship between Huang and Su, the precipitated media reports recount the respective CEO's life stories. They aren't quite rags-to-riches tales, but they apparently worked very hard to get where they are today; they didn't inherit these tech empires.

Lastly, it is amusing to see how the Chinese language media frames this family tree story. Some reports claim to have now discovered the true relationship between "the godfather of AI and …the queen of semiconductors."

Intel's upcoming site near Magdeburg, Germany, promises to become Europe's largest and most advanced semiconductor manufacturing facility, but its costs are increasing, and Intel recently expressed its intention to get more subsidies from the country's government. But the authorities are unwilling to provide the company more money as they do not have it in the budget, reports Financial Times.

"There is no more money available in the budget," Germany's finance minister Christian Lindner told Financial Times. "We are trying to consolidate the budget right now, not expand it."

Intel's proposed fab near Magdeburg, Germany, was expected to initially cost $18.7 billion (€30 billion), and the federal government agreed to provide the chip giant some $7.2 billion in state aid. But the fab construction was delayed, and due to rising energy prices, material costs, and inflation, now the cost of the production facility is expected to be about $31.675 billion (€30 billion), possibly because Intel also decided to install more advanced tools in a bid to produce chips on more sophisticated production nodes in Germany. To bridge this funding gap, the company is seeking additional subsidies of $4.223 billion - $5.279 billion (€4 billion - €5 billion) from the German government. 

The dispute over subsidies for Intel has caused a split among the authorities. Chancellor Olaf Scholz and Economy Minister Robert Habeck are believed to be open to increasing financial support to Intel as they are encouraged by the prospect of Intel increasing its total investment from €17 billion to €30 billion. However, Christian Lindner is against such subsidies and insists that the chancellery and economy ministry need to identify where additional funding would come from.

Some economists in Germany argue that subsidies are a waste of taxpayers' money, while others fear that reducing dependence on Asian suppliers in the chip industry is challenging given the complexity of semiconductor supply chains.

Intel declined to comment on the minister's comments but said, "There is a cost gap and we are working with the government on how to close it."

Lindner's opposition to subsidies is not limited to Intel. In addition, Lindner opposed Habeck's plan to subsidize the cost of electricity for energy-intensive industries, arguing against the legality of such state aid under EU rules. Meanwhile, lower electricity costs could also support Intel's upcoming fab near Magdeburg. Habeck reportedly suggested that the funding for these subsidies could come from the Economic Stabilization Fund, a suggestion Lindner dismissed as a violation of coalition agreements.

There may be some previously "under the rug" issues with the United State's CHIPS Act, a $280 billion subsidy package aiming to strengthen the country's semiconductor manufacturing infrastructure. Namely, the recent gridlock regarding raising the debt ceiling limit has already sparked a reduction in part of the Act's funding allocation, which could spell more challenging times for high-tech funding allocation further down the road.

The CHIPS Act outlines a plan to inject $280 billion in funding to increase domestic semiconductor manufacturing capability in the US. This investment aims chiefly to insulate the United States' supply of chips in case of increased geopolitical tension surrounding Taiwan, home to all-powerful TSMC - whose research and manufacturing expertise make it a prime target for absorption.

Luckily, the $52 billion injection in direct subsidies to chip manufacturers (such as Intel) is already funded, so initial work on laying the foundations for US-based foundries can start. This is the quickest way for the US to reduce its dependence on Taiwan's manufacturing capabilities.

But out of the $280 billion, a large portion of it ($170 billion) requires yearly appropriation by Congress - meaning its allocation depends on whether more funding can be borrowed through the raise in the debt ceiling limit. That leaves it at will of the negotiations between Republicans and Democrats. 

This $170 billion is split between the National Science Foundation and the Department of Energy and is intended to fund workforce development, STEM education, and research and development over the next three years. And there've already been cuts in this year's planned injections: the National Science Foundation (NSF) received $9.87 billion out of a maximum of $11.9 billion, and the Department of Energy (DoE) received $8.1 billion out of a maximum $8.9 billion.

We've also seen a turning of the tides where tech companies have been triggering layoffs - if companies are cutting workforce costs, there's less need to introduce new workers into the field. But research and development are fundamental in a competitive scene where the leading edge, the ability to churn out chips with the highest number of performant transistors, attracts the most significant returns. Just ask TSMC.

Yes, the first round of funding for foundries "Made in the US" has been secured. But as Intel, TSMC, AMD, and any other company that's meddled with semiconductor design and manufacturing knows, this field isn't a 100-meter sprint. It's the 800-meter tactical racing equivalent to a multi-year fab construction. All while adapting design processes and tooling so that everything - including supply chains and human resources - fit so right that the well-oiled machine of semiconductor manufacturing can output adequate amounts of wafers - never mind yields - come the day of the opening ceremony.

Unfortunately, neither Democrats nor Republicans function like a semiconductor factory; their plans are fickler and more prone to error. It remains to be seen whether there'll be enough motivation to keep funding flowing to where it's best applied.

TSMC’s uncompromising attitude toward employment has resulted in a crop of unfavorable reviews on sites such as Glassdoor in the U.S. TSMC's U.S. operations have a paltry 27% approval rating on the anonymous employer review site, according to a report by Fortune. This is pretty dismal compared to firms such as Intel — which has an 85% approval rating.

TSMC chairman Mark Liu responded to the startling contrast in approval ratings in an interview with local Taiwanese press by suggesting people who aren't passionate about semiconductors and prepared to work long shifts shouldn't look for a career in the industry. Liu also said that TSMC's U.S. employees were not expected conform to the same work culture as its Taiwanese employees, according to a report by Focus Taiwan.

Regardless, disgruntled U.S.-based TSMC employees have anonymously shared scathing reviews of the company's work culture on Glassdoor. Nearly 100 reviews for TSMC in the U.S. have been published on the site, and common complaints mention:

• Heavy workloads
• Poor training
• Long hours
• High expectations from bosses / supervisors
• Limited breaks
• Work-life balance issues
• ‘Military style’ controlling atmosphere
• Stress

Some of the most concerning comments highlighted in the Fortune report included complaints about the "brutal" corporate culture, about 12-hour days being standard practice, and about employees sleeping in the office for a month. 

The average review of TSMC North America (not specifically Arizona) on Glassdoor appears to have improved since the Fortune report dropped. At the time of writing the company’s overall approval rating on the site is around 58% — which still isn't great compared to Intel.

TSMC Executive Comments

In response to the report, chairman Mark Liu said, "Those who are unwilling to take shifts should not enter the industry, since this field isn't just about lucrative wages but rather a passion for [the semiconductor industry]." 

Liu also suggested that U.S. employees had it relatively easy compared to Taiwanese employees, but said work culture was open to discussion so long as TSMC's core company values were adhered to. 

TSMC has plans for two wafer fabs in Arizona, which will output chips on 3nm and 4nm process nodes. The first plant is scheduled to start mass production in 2024, while the second will come up to speed in 2026. 

TSMC has already had trouble recruiting for its new Arizona fabs due to its requiring a six- to 12-month training period in Taiwan, but perhaps its corporate culture is also to blame. A recent report of criminal behavior disturbing residents of the TSMC Village communities in Arizona will also be of concern to potential recruits.

A report citing sources at various IC design houses suggests TSMC is planning price rises for its chip fabs as soon as January 2024. Taiwan’s DigiTimes says that TSMC has been preparing customers for price hikes for quite some time and, depending on factors like order volume, it intends to push prices of advanced process output up between 3 and 6%. These rises would be in addition to the 10 to 20% increase in quotes for 2022, according to the source.

The semiconductor market is very much a rollercoaster, as reflected in headlines over the last year spanning shortages and surplus, plans for closures and expansions, and stock price collapses and surges. TSMC is so pivotal to the tech industry at large that it has been buffered somewhat from the worst of the recent business cycle. There were some indications of lowered confidence in recent months, with talk of recession in many parts of the globe,  between the crypto collapse and the AI boom, but TSMC hasn't fared too badly. Now, with customers like Nvidia and Apple pushing for more, newer, and better chips, it is understandable that TSMC should again feel confident enough to dig out its price increase email template.

DigiTimes' sources assert that numerous large clients have been in these price discussions with TSMC. Apparently, TSMC is softening the potential pain from price rises by saying that those that accept the increases will be better placed to reserve production capacity in advance. This tactic should appeal to the most successful customers for whom time is of the essence.

Highly successful tech businesses like Nvidia and Apple will not stray from close TSMC partnerships for the foreseeable future. The question is whether they took these price rises on the chin, or whether they were big enough customers to negotiate special rates better than the reported 3-6% increase on the wider customer base. Other major TSMC advanced process technology customers include AMD, Broadcom, MediaTek and Qualcomm.

Many readers will no doubt be wondering whether a 6% increase in the price TSMC charges to companies like AMD and Nvidia will impact the price of the CPUs, GPUs, or PC systems they buy. Such questions are always hard to answer as there are many other unknown costs moving up (usually) and down, and a product’s bill of materials (BOM) and its relation to the retail price can vary significantly between company and product category. Retail business sellers also have to deal with customer expectations and psychological price targets, which is why we so often see products priced at $199, $349, $499, $999, etc.

According to the Economic Daily (UDN) in Taiwan, TSMC is putting its process node pedal to the metal to satisfy customers like Nvidia and Apple. The report suggests the contract chipmaker has started pre-production work to prepare for 2nm trial production, and that 2nm mass production is on track for 2025. TSMC’s plans, if successful, would also likely maintain its competitive edge against rivals like Intel and Samsung.

Sources speaking to UDN say that engineers and support workers have been dispatched to the Zhuke R&D plant in Baoshan, Taiwan to prepare for 2nm trial production. A small-scale production line aims for 1,000 wafers this year, says the source report, with trial production penciled in for 2024, and mass production in 2025. For mass production, expanded facilities at Zhuke and a new plant in Taichung could be utilized.

It may be important for TSMC to get an early start with its 2nm family (which it refers to as the N2 family). This will be the first processor generation from the company to adopt a new gate-all-around (GAA) transistor architecture, so there may be unforeseen hurdles to deal with.

The source report includes some interesting insight into 2nm development, which indicates that things are trending positive for the company so far. Firstly, UDN says that customer competition is more intense than ever, indicating key partners are rightly investing in their customized solutions to ensure success.

Secondly, TSMC is said to be applying AI to its manufacturing process for efficiency gains, and this move is expected to save energy and thereby reduce carbon emissions. With these new AI-enhanced production methods, dubbed AutoDMP, Nvidia-powered AI is being used to optimize chip designs 30x faster than previous methods / technologies. The Nvidia-powered technique is touted to make processors cheaper, offer higher performance, and be more power efficient.

For further technical information on TSMC’s N2 family, including N2P and N2X, check out our previous coverage from the 2023 North American Technology Symposium in April.

TSMC clean room workers are routinely tasked with finding 'bombs' in machinery, according to Taiwan's United News Network (UDN). We have put 'bombs' in inverted commas, as the staff undertaking equipment testing and maintenance will find sticky notes with 'bomb' written on them located on various components instead of potentially explosive devices.

UDN based its report on anonymous TSMC employee chitchat on social media, so we must add a pinch of salt to this eyebrow-raising work practice. However, we understand that the 'bomb-finding' appears to work as follows:

• A supervisor will hide sticky notes inside machines which are part of a regular service/inspection rota
• Workers maintaining and inspecting machines will spot and collect the notes as they work through their daily tasks.
• If any notes are missed, workers will have 'points' deducted. However, we don't know how important positive or negative points are to workers.

The 'bomb' notes are designed to minimize workers signing checklists as completed without actually performing the work. However, you might sympathize with one of the supposed TSMC employees responding to the 'bomb-finding' task by asking, "What kind of punishment is this?" This method of inspiring workers appears to have occurred or been most common at TSMC's 14B factory in the Southern Taiwan Science Park, Tainan.

Various equipment inspections are essential to TSMC employee operations as they seek to minimize downtime for a host of preventable issues. Careful examination is used to check for anything abnormal, with problems like water or acid leakage among the most significant issues noticeable by the eye.

Considering the above, it doesn't seem like the TSMC's 'bomb-finding' activity is reckless. However, potential issues may be that it causes employee stress or that some may find the use of the bomb word tasteless.

UDN reckons the source of the 'bomb' sticky notes idea might be Taiwan's military. The report says that the army routinely does sabotage exercises where critical strategic facilities are infiltrated. Operatives leave spray-painted 'bomb' and 'poison gas' tags in their wake. Obviously, a 'defending' team would seek to minimize the spray paint antics or attend to them rapidly.

As a side note, UDN says that bomb threats are an increasingly popular way for criminals to extort money from organizations and businesses - suggesting many schools, theme parks, and public transport hubs often get targeted with such shenanigans. Thus, using this language in the workplace might not be a great idea.

An official statement published by the Cyberspace Administration of China (CAC) says that operators of critical information infrastructure "should stop purchasing Micron products." The decree, spotted by Wen-Yee Lee, comes after a network security review of Micron's products sold in China.

Apparently, the Chinese government security review found that "Micron's products have relatively serious potential network security issues, which pose a major security risk to my country's critical information infrastructure supply chain." However, no detail of any specific security risks presented by the usage of Micron products is given. As a matter of "national security," we assume the Micron products ban will be implemented swiftly and thoroughly.

At the end of its statement, the CAC asserts that all companies from all countries are welcome to enter the Chinese market, as long as they abide by local laws and regulations. Again, there is no hint given about how the US-based Micron has broken any regulations.

Without some detail behind its claims about Micron, it is natural to wonder whether China has a political and/or economic motivation for its ban on the firm's products. It is, therefore, worth looking at some contemporary Micron / China news to provide some context to the decision.

Firstly, let us consider the political background. The additional US trade restrictions implemented last October meant that emerging Chinese DRAM and NAND giants (Yangtze Memory Technologies Co YMTC and ChangXin Memory Technologies CXMT, respectively) have far narrower markets to address. The biggest visible fallout from the US government decision last year was when, after a period of close cooperation, Apple had to abandon plans to use YMTC NAND in its iDevices. If US sanction activity motivated China's new Micron ban, then it would be a pretty clear tit-for-tat exchange.

Another possibility is that China's industrial efforts and investments have already started to pay dividends in DRAM and NAND production. Thus the country now feels alternative supplies are adequate enough that it can shun US-based Micron's products without issues.

We also must think about how Micron may be affected by this China ban. Micron is one of the big three memory makers, but DRAM and NAND are commodity products where brands hold little cachet. Recent financials suggest that 25% of Micron's 2022 revenue was generated by sales in China. However, in a commodity market, the flow of DRAM and NAND ICs should be able to adjust quite quickly.   

There were signs that these semiconductor commodities were primed for further steep declines in 2023, but there has been a surge in positivity more recently, off the back of the interest in companies that enable/serve the AI technology sector. Thus Micron shares have seen some ups and downs in recent months, but are still up nearly 20% in the last six months, and the last five days have accounted for half of that gain. Some of the positivity behind Micron also stems from its recent substantial investment in Japan, with local government support.

At the time of writing, Micron shares have not reacted at all to the China ban in after-hours trading. As this is fresh breaking news, we can't rule out more of a reaction on Monday.

Samsung says it has reaffirmed its DRAM leadership by kicking off mass production of 16-gigabit (Gb) DDR5 DRAM at 12nm. The South Korean electronics giant asserts that the memory ICs resulting from this new process reduce power consumption by about a quarter vs. the previous generation and will enhance wafer productivity by as much as a fifth. Moreover, these leading-edge memory chips will boast a maximum pin speed of 7.2 Gbps.

Discussing the manufacturing milestone, Jooyoung Lee, Executive Vice President of DRAM Products & Technology at Samsung Electronics, said that "Using differentiated process technology, Samsung’s industry-leading 12nm-class DDR5 DRAM delivers outstanding performance and power efficiency." However, PC users will have to wait for the trickle-down, as the first use of these 12nm DDR5 ICs will be in applications like data centers, artificial intelligence, and next-generation computing.

Samsung says that the development of 12nm-class DRAM was enabled by "the use of a new high-K material." In further detail, it explains that the transistor gate material used in these ICs has a higher capacitance, making its state easier to accurately distinguish. Furthermore, Samsung's efforts to lower operating voltage and reduce noise have also helped deliver this optimized solution.

These are still 16 Gb ICs, so Samsung hasn't traveled further along its density roadmap with these DRAM chips. Instead, the heralded benefits concern power efficiency, speeds, and wafer economy. If you want some numbers, Samsung specified that the 12nm DDR5 ICs reduce power consumption by 23% compared to the previous gen, and enhance wafer productivity by up to 20%.

Will This Mean Faster DDR5 Modules?

We already mentioned the new DRAM ICs offer a pin speed of 7.2 Gbps. Samsung says that, in theory, this means that a DRAM-to-DRAM transfer of two 30GB UHD movies could take place in about a second. However, it neglects to highlight that this is the same pin speed as that touted for the previous gen 14nm DRAM ICs. This doesn't rule out the potential of the newer DDR5 DRAM to perhaps overclock better; for this information, we will have to wait and see.

Samsung says it has already confirmed the new (12nm) 16Gb DDR5 ICs have been through compatibility testing for use with AMD systems. It is also working closely with other unnamed companies.

Let's face it: While Moore's Law may not be completely dead, it's been on life support for a while. Long gone are the days of "easy" performance uplifts from improving manufacturing processes and simple transistor size reduction. These days, R&D teams looking to improve semiconductor performance year upon year have to branch out into not only the transistor architecture itself, but also into materials engineering. Exploring elements present in the periodic table might give semiconductors a more performant, smaller, and energy-efficient design than silicon can provide. Now, a research team with the School of Engineering at Stanford has turned to one of the most precious metals on Earth - palladium - as a potential venue for faster, cooler, and more energy-efficient memory architectures.

The research team's approach is based on the assumption that we've transitioned from the Internet era and are already knee-deep in the AI era. According to Shan Wang, the Leland T. Edwards Professor in the School of Engineering at Stanford University, the idea behind the research was to "enable AI on edge – training locally on your home computer, phone, or smartwatch – for things like heart attack detection or speech recognition." 

Yet to do that, he added, our current technological crop won't cut it; we need a type of memory that's been on again, off again in the press without a single product being brought to market: memristors. Memristors are a type of non-volatile memory that doesn't require a constant stream of electricity to keep data in its memory banks (check here for a more detailed breakdown of memristor tech). The first author of the paper, Mahendra DC, explained it best: “We are hitting a wall with the current technology,” DC says. “So we have to figure out what other options we have.”

To break that wall, the researchers singled out a compound of palladium, manganese palladium three, as having the necessary properties to finally (they say) enable a memristor design to see the light of day. Part of the checklist pertains to our current semiconductor manufacturing tech: any material being considered for novel semiconductor designs has to deal with the fact that tooling and supply lines are entrenched in the silicon realm. So any alternative compound will ideally be able to be slotted into current manufacturing processes without the need for extensive (and expensive) process changes.

Manganese palladium three fits that bill, yet the most important property of the compound is the way in which its particles are manipulated to become memory storage devices. Information here is derived not from a voltage state, as is the case with standard memory devices such as NAND flash, but by manipulating the spin orientation of the compound's electrons. Essentially, researchers are able to manipulate an electron's magnetic field, bouncing its center between the north and south poles. Then one of these spin directions (north-south or south-north) is made to represent a 1 or a 0 - unlocking the binary system behind computing as we know it. Named "spin orbit torque magnetoresistive random access memory", or SOT-MRAM for a necessary shortening, this spin data storage method has the potential to store data faster and more efficiently than current technology allows - and at higher densities as well.

Of course, there's one big caveat in all this research: Palladium is currently almost as expensive as gold, but is also a much rarer commodity. To make matters worse, 40% of the world's palladium is currently mined in South Africa, with 44% more being mined in Russia. So palladium availability might itself be a problem from the logistics and geopolitics side of the equation. Beyond that, increased demand for palladium for SOT-MRAM would certainly drive its price up even higher. All off this will need to be considered when pursuing novel transistor designs.

All in all, the research is promising, but like all memristor news in the last decade (and longer), it seems that we're still standing at the beginning of the proverbial yellow brick road. Whether or when we see the castle at the end, well, that's for the future to tell us.

On Monday, Taiwan’s Minister for National Defense, Chiu Kuo-cheng (邱國正), made a statement about the nation’s territorial integrity. According to the Taiwan News, the Taiwanese minister said that the island’s armed forces would not tolerate any U.S. attempts to destroy TSMC in the event of a war with China. Normally the war of words is heated where Taiwan and China are concerned, but Taiwan and the U.S. are besties, so what is going on?

The geopolitical tensions between the U.S., Taiwan, and China are fundamentally confusing with various principles, understandings, ambiguities, and policies meaning different things to the different parties. We are accustomed to seeing the U.S and Taiwan as allies, but perhaps a recent assertion by U.S. Congressman Seth Moulton has overstepped the mark.

On the topic of U.S. chip policy and China, Moulton recently told political conference-goers that "the U.S. should make it very clear to the Chinese that if you invade Taiwan, we're going to blow up TSMC.” Openly talking about these policies seems to be frowned upon in Taiwan, as made clear by the Defense Minister Chiu’s statement in response to Taiwan media questioning on Monday.

The Taiwanese minister made it clear to reporters that the armed forces are responsible for defending Taiwan (and its people, materials, and strategic resources) from any aggression, regardless of the source. “If they want to bomb this or that," the armed forces will not tolerate it, asserted Chiu.

U.S. Congressman Seth Moulton is a former Marine Corps officer, and a member of the Democratic Party, and it is important to remember his comments about destroying TSMC’s chip making fabs in Taiwan are not without precedent.

Last year we reported on TSMC Chairman Mark Liu, who told interviewers that, “Nobody can control TSMC by force.” He reasoned that, “If you take a military force or invasion, you will render TSMC factories inoperable.” But, explosives wouldn’t be needed, according to Liu, as TSMC's fine-tuned operations would simply crumble as its real-time connections with the outside world evaporated. However, we should remember that China’s motives should not be measured against business rationality. Instead, its policies may be colored by a leader seeking glory, destiny, a legacy, and so on.

Similarly, in October last year, we reported on a statement by the director general of Taiwan’s National Security Bureau, Chen Ming-Tong. According to reports, Chen told Taiwanese lawmakers that TSMC would have to essentially close without the contributions of the likes of ASML and other foreign suppliers. “Even if China got a hold of the golden hen, it won’t be able to lay golden eggs,” summed up the National Security boss.

Hopefully the above situations will never unfold, and the actual policies regarding the destruction (or defense) of strategically important facilities like TSMC will therefore never be enacted.

TSMC is said to be in wide ranging talks about the significant investment required to open a new chip fab in Saxony, Germany. Private investment may be as high as €10 billion (~$11 billion), according to "people familiar with the matter" talking to Bloomberg reporters. Public funds might end up matching that amount, to reel in this strategic investment, but the European Commission will have to greenlight any state aid. It is understood that TSMC's first fab in Europe will concentrate on 28nm production.

If the plans reported upon by Bloomberg are correct, TSMC will work in partnership with NXP Semiconductors, Robert Bosch, and Infineon Technologies to provide a wide base for the venture. The partnership will spread the €10 billion (~$11 billion) investment risk. TSMC partners' local business knowledge will help in both planning and the raising of state aid. Public funds won't quite meet the private investment level, at least initially. Bloomberg's report says that state subsidy levels will start at around the €7 billion mark ($7.75 billion), but could well rise to match the private investment capital.

If the EU Chips Act was designed to catch the biggest fish in semiconductors, TSMC and its partners' plans will be hard to resist. According to Bloomberg, it is typical for similar projects to gain 40% funding through EU subsidies, as the region strides to double its global semiconductor production share by 2030. Approval for these state subsidies will have to come from the European Commission, and negotiations over the size of subsidies will understandably be intense.

TSMC's 28nm Node

If the negotiations run smoothly, Bloomberg says that the Saxony chip fab project could be approved by TSMC by August. It won't be a leading edge facility, says the source, instead it will be tasked with churning out 28nm chips. A report shared by AnandTech last summer says that TSMC is "strongly encouraging its customers," still using its oldest nodes to migrate their mature designs to 28nm, which will become a new base level semiconductor component fab choice. While PC enthusiasts might turn their noses up at 28nm, the output will be welcomed by manufacturers who fared badly during the chip drought of the early 2020s.

In the wake of Bloomberg's report, the Taiwanese chip-making giant has confirmed to Reuters that it is still evaluating the possibility of building a fab in Europe. However, that doesn't really confirm nor deny anything of substance. Also remember, at this stage the plans could still change or fall through, even if the unnamed sources speaking to Bloomberg are highly credible.

Will Intel be a Neighbor?

We reported last month that Intel was still in negotiations with the German government over the scale of subsidies on offer. It was looking for a further $5 billion in subsidies, which will be used to establish a new chip fab near Magdeburg.

The warning signs regarding Samsung's chip business have been readily apparent for quite some time, but the severity of the company's troubles was realized yesterday. Samsung revealed its earnings for the first quarter of 2023, and it wasn't pretty. The company said revenue came in at 63.75 trillion Korean won ($47.6 billion), which was down 18 percent compared to Q1 2022, but in line with Samsung's previous guidance. 

However, the South Korean chip and electronics giant said that operating profit during the quarter plummeted from KRW 14.12 million ($10.5 billion) during Q1 2022 to just KRW 640 billion ($476 million). To put that figure into perspective, Samsung hasn't reported an operating profit that low since Q1 2009.

So, what went wrong during the preceding quarter? Well, Samsung's semiconductor business has traditionally been the company's significant revenue and profit generator. Samsung is a dominant player in this field, providing memory chips and NAND flash that are found in everything from computers to smartphones to tablets to IoT devices. And as we reported earlier this month, Samsung dialed back memory production due to low customer demand. It was speculated at the time that Samsung witnessed a 30 percent drop in orders during Q1 2023 and that inventory exceeded demand by 21 weeks.

Further compounding matters is that the oversupply has led to contract pricing for DRAM and NAND falling by as much as 24 percent and 16 percent, respectively, according to KB Securities. When you factor in that Samsung commands 45.1 percent of the DRAM market and 33.8 percent of the NAND market, the first quarter became a perfect storm of negativity for the company's financials resulting in a KRW 4.58 trillion ($3.4 billion) loss for its semiconductor business.

Samsung's LSI Business also fared poorly during the quarter, with the company noting that "earnings fell sharply in the first quarter due to a drop in demand for major products such as SOCs, sensors and DDIs." 

Looking forward, Samsung says it will accelerate its shift from DDR4 production to DDR5 and LPDDR5x while also addressing the demand for HBM3. In addition, it will ramp production of QLC NAND to address the needs of customers in need of high-density storage products.

The Visual Display (TVs), Digital Appliances, and SDC (display panels for smartphones, laptops, automotive, etc.) businesses were also down year-over-year. However, the one bright spot for the company was its mobile business, which posted revenue of KRW 31.82 trillion ($23.6 billion) and operating profit of KRW 3.94 trillion ($2.93 billion). Samsung indicated that the successful launch of the Galaxy S23 series (including the premium-priced Galaxy S23 Ultra) bolstered the mobile business.

It should be noted that Samsung largely expects to see some recovery during the second half of the year for its businesses that were down in Q1 2023. The company also reiterated that it is already mass-producing first-generation 3-nanometer silicon and that its second-generation 3 nm products will enter mass production next year.

A fire broke out at a new TSMC manufacturing facility on Tuesday night. Reports say that the fire was detected at 7.30pm local time, and firefighters had managed to extinguish it by 9.10pm. No people were injured. The impact upon TSMC’s manufacturing and production plans, and which customers may be impacted by this unfortunate event, isn’t clear at the time of writing.

The facility was still under construction. The source of the fire is being investigated by the county's Fire Bureau, and a safety investigation is underway. Located in Zhunan, yet part of the expansive Hsinchu Science Park, TSMC broke ground on the construction last year. So far, TSMC had completed the structure of the main building and installation of some key factory infrastructure.

Reports suggest TSMC were at the stage of building clean room facilities and piping systems at the site. Over the course of the blaze, which covered 300 sq.m (3,230 sq.ft), damage was done to some machinery that had already been installed plus the chlorinated polyvinyl chloride (CPVC) piping systems.

Miaoli Fire Department, who successfully extinguished the blaze and shared the image above, are investigating the cause of the fire. TSMC will also work with its contractors to investigate what happened on Wednesday evening, and ensure it can avoid any mistakes which may have sparked this fire.

TSMC’s plans for the Zhunan facility are that it will become an additional advanced IC packaging and testing site. This is an important business for TSMC, which would like to continue to vertically integrate, to become a one-stop-shop.

Note that this is an additional advanced packaging and testing plant, meant to join four existing facilities in Taoyuan, Hsinchu, Taichung and Tainan. A different report says that the new Zhunan facility will be TSMC’s sixth plant dedicated to advanced packaging and testing. Whatever the case, it seems that TSMC will have coverage for packaging and testing work, but the new capacity at Zhunan will just take a little longer to come online.

Earlier today we reported on TSMC’s sharing of more information about its upcoming 2nm class processes. It expects to be expanding its N2 plans, with both N2P (performance) and N2X (eXtreme performance, high voltage) nodes. You can read more in our extensive article.

As weak demand persists, market researchers at TrendForce expect the average selling price (ASP) of NAND Flash to decline a further 5-10% in 2Q23. There appears to be a pivot point in the second half of the year where supply and demand could be balanced and a turnaround could happen. A lot rests on manufacturers cutting back on production, plus shipments of NAND Flash packing products like servers, smartphones, and computers.

Even if things fall favorably for the NAND Flash industry, TrendForce’s best estimate is that 3Q23 will be a stable period, with the chance of a rebound in 4Q23. The report from TrendForce comes hot on the heels of the terrible financial report shared by memory giant Micron. In case you missed that, Micron’s YoY revenue dropped nearly 53% on reduced take-up of its NAND and DRAM products. Micron’s forecast was similar to TrendForce’s in at least one respect, with a gloomy prediction for 2Q23.

TrendForce doesn't offer a lot of positivity with regard to client SSDs – consumer devices sold as PC components or through PC-makers. Inventories of PCIe Gen 3 SSDs are being actively whittled down, while take-up of PCIe Gen 4 drives are slower than some expected. And the decline in QLC NAND pricing has dragged down TLC NAND as well. 

It will be difficult for NAND makers to cut prices further, but they are still expected to drop 5-10% in the coming quarter. There remains some hope that at some point in the second half of 2023, the industry will see positive momentum on dependent markets like laptops and smartphones – with new generation devices helping shift more NAND.

Enterprise SSDs have a rosier outlook this year. TrendForce says that Chinese demand will be strong due to government initiatives that have recently been approved. Moreover, the launch of AMD Genoa is a positive for server makers and thus NAND suppliers. Nevertheless, NAND Flash makers are in a weak negotiating position, so TrendForce is also predicting an enterprise SSD price decline in the order of 8-13% in 2Q23.

There are similar stories to be told in the eMMC (smartphone / laptop), and UFS NAND (smartphone) markets. However, UFS demand is looking more positive, with the introduction of UFS 4.0 for flagship smartphones, and expectations of a healthy mid-year and peak season for new / refreshed devices in 2023.

TrendForce’s report doesn’t specifically mention any underlying causes of the supply and demand issues we are now seeing, although some are pretty obvious. Memory is dealing with the same issues that the wider PC and tech industries are tackling, like inflation and its impact on disposable income, the Ukraine war, and the pandemic tech sales boom in preceding years. While that's not great for business, in the short term it should mean increasingly enticing SSD deals for those looking to increase storage in their PCs and game consoles. 

Nvidia is one of the leading designers of chips used for artificial intelligence (AI) and machine learning (ML) acceleration. Therefore it is apt that it looks set to be one of the pioneers in applying AI to chip design. Today, it published a paper and blog post revealing how its AutoDMP system can accelerate modern chip floor-planning using GPU-accelerated AI/ML optimization, resulting in a 30X speedup over previous methods. 

AutoDMP is short for Automated DREAMPlace-based Macro Placement. It is designed to plug into an Electronic Design Automation (EDA) system used by chip designers, to accelerate and optimize the time-consuming process of finding optimal placements for the building blocks of processors. In one of Nvidia’s examples of AutoDMP at work, the tool leveraged its AI on the problem of determining an optimal layout of 256 RSIC-V cores, accounting for 2.7 million standard cells and 320 memory macros. AutoDMP took 3.5 hours to come up with an optimal layout on a single Nvidia DGX Station A100. 

Macro placement has a significant impact on the landscape of the chip, “directly affecting many design metrics, such as area and power consumption,” notes Nvidia. Optimizing placements is a key design task in optimizing the chip performance and efficiency, which directly affects the customer.

On the topic of how AutoDMP works, Nvidia says that its analytical placer “formulates the placement problem as a wire length optimization problem under a placement density constraint and solves it numerically.” GPU-accelerated algorithms deliver up to 30x speedup compared to previous methods of placement. Moreover, AutoDMP supports mixed-sized cells. In the top animation, you can see AutoDMP placing macros (red) and standard cells (gray) to minimize wire length in a constrained area.

We have talked about the design speed benefits of using AutoDMP, but have not yet touched upon qualitative benefits. In the figure above, you can see that compared to seven alternative existing designs for a test chip, the AutoDMP-optimized chip offers clear benefits in wire length, power, worst negative slack (WNS), and total negative slack (TNS). Results above the line are a win by AutoDMP vs the various rival designs.

AutoDMP is open source, with the code published on GitHub.

Nvidia isn’t the first chip designer to leverage AI for optimal layouts; back in February we reported on Synopsys and its DSO.ai automation tool, which has already been used for 100 commercial tape-outs. Synopsys described its solution as an “expert engineer in a box.” It added that DSO.ai was a great fit for on-trend multi-die silicon designs, and its use would free engineers from dull iterative work, so they could bend their talents towards more innovation.

Dell aims to remove any reliance upon Chinese manufacturing by 2027, at least regarding US-sold products. According to Taiwan's Commercial Times, Dell's actions are part of a phased plan inspired by geopolitical influences.

The source claims the above news to come from internal Dell documents. Geopolitics has apparently sparked the creation of these plans, and we think this must be a reference to the growing US-China technology trade tensions. Also, China has been ramping up threats of an armed invasion of Taiwan, a crucial center for semiconductor technology, components, and computers. Earlier signs of Dell's plans to extricate itself from business dealings with Chinese entities were seen in January when three insiders told Nikkei Asia that Dell wanted to stop using chips made in China by 2024.

Dell's plans would seem to follow the following timeline:

• 2025: Dell will begin to execute its plans for reducing the reliance on Chinese components across its desktops, laptops, and peripherals
• 2026: All chips used in Dell products should be made by non-Chinese companies outside of China. It is also said that there is a target for 60% of Dell’s US-sold devices to be made outside of China by this time.
• 2027: By this time, Dell hopes to have 100% completed its transition, with no products sold in the US coming from factories in China.

It isn’t entirely clear from the text and diagram translations, but Dell appears to be aiming to cut out Chinese-made semiconductors as a priority. The next component on the hit list will be ICs made in China by foreign companies, followed by a wave of reducing any other Chinese components and replacing Chinese manufacturing of its devices.

We must also remember that the Commercial Times report indicates that the timeline primarily concerns laptop and desktop PCs intended for Dell’s US (home) market. The report said that even after 2027, more than half of Dell products sold worldwide will still be at least partly made in China. Dell ships around 50 million PCs a year, so changing production locations isn’t a trivial or quick task. However, with these plans, Dell is moving at its own pace rather than being forced by unfortunate events.

While the outlined geopolitical influenced strategy might win some favor in the US, particularly among China hawks, it would be understandable if Dell doesn’t come out and explain things the way the Nikkei and Commercial Times sources have done. Dell still sells millions of PCs to Chinese consumers, with recent reports suggesting it shifted 1.5 million Dell and Alienware branded PCs in Q3 2022 and 1.1 million in Q4 2022. Dell’s home market remains its biggest, with US customers buying 4.7 million Dell PCs in Q3 2022, as a recent example.

Leading-edge semiconductor production facilities already cost north of $10 billion, and the more advanced process technologies they use, the more expensive they get. But in Europe, inflation, rising energy pricing, and the increasing costs of materials have driven Intel's Germany fab costs so significantly that the company is now seeking an additional $5 billion in subsidies from the country's government, reports Bloomberg. 

When Intel announced plans to build its fab near Magdeburg, Germany, it said that the production facility would cost $18.7 billion and it would get $7.2 billion in state aid to complete the project. The site near Magdeburg is set to become Intel's next major fab campus, which was meant to cost as much as $88 billion. However, German carmakers couldn't get all the chips they needed in 2020 ~ 2022, so the country was more than willing to help Intel to establish its new manufacturing site near Magdeburg.  

But then the company had to postpone construction of the chip plant, and now with soaring energy prices, costs of materials, and high inflation, it believes that it would need to spend $31.675 billion (€30 billion) on the project. This is why Intel would need an additional $4.223 billion – $5.279 billion (€4 billion – €5 billion) in subsidies from Germany, the report says. 

The chip giant did not directly confirm the sum, but it was mentioned in talks with the German government, who confirmed that it was renegotiating the deal. 

"Disruptions in the global economy have resulted in increased costs, from construction materials to energy," a statement by Intel reads. "We appreciate the constructive dialogue with the federal government to address the cost gap that exists with building in other locations and make this project globally competitive."

The German economic ministry declined to comment on its talks with the CPU giant but reaffirmed the EU's goal of producing 20% of the global chip output by 2030 is still in place.

"With this goal in mind, the federal government is prepared to support the semiconductor industry in Germany with several billion euros and to enable new factories to be set up," the ministry said, adding that extra funding would need approval from the European Commission.

In addition to its fab in Germany, the company intends to build an R&D center in France. This project is on track, according to Bloomberg's sources. 

AMD approved hefty six-figure bonuses for several top execs earlier this month — at least four AMD VPs and the CEO, Dr. Lisa Su, will pocket substantial annual cash bonuses. In contrast, Intel’s diligent team faces painful salary cuts and suspended bonuses despite the bullish statements from Intel CEO Pat Gelsinger in 2022 when he talked about AMD being in the rear-view mirror.

Dr. Su received the biggest performance bonus — nearly a million dollars ($918,800) — for steering the red team’s ship through choppy waters to enjoy the success we continue to see. The four VPs named in the SEC filing will receive bonuses between $300,000 and $375,000. One of the VPs most readers will be familiar with is Mark Papermaster, the CTO, and ELVP of Technology and Engineering.

As for Intel, the company faces mounting difficulties highlighted during its most recent catastrophic quarter. In brief, mid-level and higher-ranked execs face pay cuts between 5 and 15%, with Pat Gelsinger taking a 25% cut. Wage raises and bonuses are out of the picture for these same execs, but hourly employees will still receive bonuses.

AMD and Intel target many of the same markets, but some key differences exist in the company operations and structures. First, Intel’s foundry business weighs heavily on its finances and fortunes, and it is busy restructuring and making massive investments around the globe to boost its manufacturing reach and prowess. For instance, Intel's planned $40 billion in investments for the new Arizona and Ohio fabs come as the company's first-quarter revenue dropped 40%, meaning that extraordinary actions must be taken.

Recessionary forces are also in the driver's seat (war, Covid hangover, etc.), affecting the entire industry. Nevertheless, several computer business leaders and industry insiders see a turn for the better in the second half of this year.

Compared with Intel, AMD doesn’t have as much of a semiconductor manufacturing burden. It is more nimble and light after divesting manufacturing operations and the massive investments needed to build and maintain a competitive foundry business. On the flip side, when the boom comes, AMD chipmaking partners will make hay and demand higher prices for their in-demand services.

For Intel employees currently enduring pay cuts, subsidies from multiple governments worldwide should help make the largest Intel investments successful in their respective locations. Intel also cut its long-standing dividend to free more cash for investment amidst its belt-tightening exercises, and also recently told employees that it will reverse its pay cuts later this year. That means that, as the cycle turns, the lean times could be followed by times of plenty, especially if Intel's drastic turnaround plan, which includes a pivot to the IDM 2.0 model, comes to fruition. 

It is no secret that the PC business and tech industry at large is going through a bit of a rough patch. However, a new report penned by Dean McCarron of Mercury Research paints an incredibly bleak picture of the state of things. Probably the biggest bombshell is that figures show the x86 processor market has just endured “the largest on-quarter and on-year declines in our 30-year history.” Based on previously published third-party data, McCarron is also reasonably sure that the 2022 Q4 and full-year numbers represent the worst downturn in PC processor history.

The x86 processor downturn observed has been precipitated by the terrible twosome of lower demand and an inventory correction. This menacing pincer movement has resulted in 2022 unit shipments of 374 million processors (excluding ARM), a figure 21% lower than in 2021. Revenues were $65 billion, down 19 percent YoY. McCarron shines a glimmer of light in the wake of this gloom, reminding us that overall processor revenue was still higher in 2022 than any year before the 2020s began.

Another ray of light shone on AMD, with its gains in server CPU share, one of the only segments which saw some growth in Q4 2022. Also, AMD gained market share in the shrinking desktop and laptop markets. For more AMD-specific financials and sales performance data, please refer to our coverage of its Q4 and FY 2022 results. Also, Mercury shared some charts for those interested in poring over x86 CPU market shares overall and per segment. A key observation here is AMD’s overall market share growth spurt from around 23% of the x86 market in 2021 to nearly 30% in 2022.

Inventory Adjustments May Be Having a Bigger Negative Impact Than Reduced Sales

McCarron was keen to emphasize that Mercury's gloomy stats about x86 shipments through 2022 do not necessarily directly correlate with x86 PC (processors) shipments to end users. Earlier, we mentioned that the two downward driving forces were inventory adjustments and a slowing of sales – but which played the most significant part in this x86 record slump?

The Mercury Research analyst explained, "Most of the downturn in shipments is blamed on excess inventory  shipping in prior quarters impacting current sales." A perfect storm is thus brewing as "CPU suppliers are also deliberately limiting shipments to help increase the rate of inventory consumption… [and] PC demand for processors is lower, and weakening macroeconomic concerns are driving PC OEMs to reduce their inventory as well."

Mercury also asserted that the trend is likely to continue through H1 2023. Its thoughts about the underlying inventory shenanigans should also be evidenced by upcoming financials from the major players in the next few months.

We have seen several big tech firms and analysts seemingly indicate that H2 2023 is going to be a turning point, an inflection point where downtrends will be broken. Again this seems to be the case with Mercury's report, and of course, we hope that there will be enough highly inspiring CPU, GPU and related technology improvements coming through in H2 2023 to ignite the PC market yet again (in addition to world peace).

Synopsys' AI solution for semiconductor design has achieved the milestone of 100 commercial chip tape-outs. The company said in a press release that customers such as STMicroelectronics and SK hynix have enjoyed up to 3x productivity increases, seeing chips with up to 25% lower total power and significant reduction in die size thanks to Synopsys DSO.ai (Design Space Optimization AI).

To be clear, AI isn't stealing chip designers' jobs: Synopsys prefers to say that, thanks to its software, human chip designers and hardware engineers are freed from iterative work, and, thanks to AI augmentation, are able to instead work on innovation.

“With reduced design and verification cycles and effort, design teams can spend more of their time innovating on their core ideas,” Synopsys said. It hopes the engineering talent shortage can at least be somewhat relieved by AI. 

The milestone achieved by Synopsys shows that AI use in electronic design automation is rapidly becoming mainstream. Moreover, this AI can be particularly useful to industrial sectors increasingly looking to enter the chip design business — auto makers, for example. Synopsys even goes so far as describing DSO.ai as an “expert engineer in a box.”

So, what exactly does Synopsys DSO.ai do? The biggest clue is in the acronym: Design Space Optimization — by AI. The tool takes care of the floor-planning for a new chip (or iterations of it). Synopsys says DSO.ai is a great fit for the trending multi-die silicon designs, which would involve a high volume of repetitive tasks for humans to plan. 

To complete its task, the AI software optimizes power, performance, and area (PPA) for any given chip design space. Working on PPA is a proven route to doing better with fewer resources, and has been a very popular target for optimizing during recent years with shortages of key materials due to cryptocurrencies and the pandemic.

Synopsys customers have been reaping the benefits of DSO.ai with impressive results claimed.  Synopsys asserts that its customers have seen productivity boosts of more than 3x, power reductions of up to 15%, substantial die size reductions in finished designs, and reduced use of resources. It also suggests an ideal task for AI is facilitating multi-foundry strategies to mitigate the impact of supply chain vulnerabilities and to lower costs.

Synopsys is already looking at broadening the use of AI in other chip design and verification workflows. It seems we may be seeing a breakthrough moment in AI-based chip design, and it's interesting to see this news as consumer-facing AIs from Google, Microsoft, and OpenAI are also making headlines.