Microsoft has unveiled Project Solara, a chip-to-cloud platform designed to power a new generation of "agent-first" enterprise devices — hardware designed to run AI agents instead of traditional apps. Announced at the Microsoft Build 2026 Developer Conference on the 2nd of June, the platform, developed by Microsoft's Applied Sciences Group, features a lightweight edge OS called the Microsoft Device Ecosystem Platform (MDEP). Interestingly, the OS is built on the Android Open Source Project (AOSP) rather than Windows.

MDEP is paired with Azure-hosted agent services and persistent cloud-based state, meaning devices act as interfaces to AI agents running across Microsoft's cloud infrastructure rather than as fully self-contained computers. Together, the software stack forms what Microsoft describes as a chip-to-cloud architecture for enterprise AI devices, combining cloud-hosted agents with centralized security, management, and orchestration capabilities.

"The 'operating system' is liminal, transcending the device and the cloud. The system brings a lightweight window to the edge, where the agent manifests and where the state, via Azure, can encompass a constellation of specialized devices,” explained Steven Bathiche, Microsoft's Corporate Vice President and Technical Fellow in the Applied Sciences Group.

To populate that ecosystem with hardware, Microsoft has partnered with Qualcomm and MediaTek as its first silicon partners — Qualcomm for portable and wearable form factors and MediaTek for stationary devices. The company has no plans to manufacture end products itself. Instead, the company is releasing reference designs for OEMs to build from, alongside an "approved chipsets" requirement that gives Microsoft certification-level control over which hardware qualifies for the platform, similar to Google's GMS certification model for Android.

To demonstrate Solara, Microsoft unveiled two concept reference designs built on the platform. A stationary desk-mounted AI hub built around MediaTek IoT silicon and a wearable AI badge powered by Qualcomm hardware. The desktop companion features a display, a camera, a UWB (ultra-wideband) presence sensor that handles automatic login and lock, dual far-field mics, and two USB-C ports. Connected to an external display, the device can double as a Windows 365 cloud PC client.

Meanwhile, the wearable badge is equipped with a touchscreen, Hello for Business fingerprint sensor, far-field high-SNR microphone array, side-facing camera, and 5G, WiFi, Bluetooth, and GNSS connectivity — targeting front-line workers such as nurses, retail staff, and field workers. Microsoft confirmed that both devices are intended as reference designs for OEM partners rather than retail products.

Central to the platform is what Microsoft calls just-in-time UI — an adaptive interface layer that allows a single agent to render appropriately across different screen sizes and input modalities without requiring developers to rebuild the experience for each device. Microsoft positions this on a spectrum between conventional responsive design and fully generative UI, where AI constructs interfaces dynamically with no predefined structure; Solara currently targets the middle ground, prioritizing consistency while avoiding per-device redesign overhead.

“The same agent can render a custom experience on multiple screen sizes and modalities with little or no additional work from the developer. For us, that is the first proof point: a path to specialized devices without requiring developers to rebuild the experience from scratch each time,” said Bathiche.

Another notable aspect of Solara is Microsoft's decision to build MDEP on Android rather than Windows. AOSP scales naturally to the lightweight, constrained hardware that wearables and embedded devices run on — something Windows, with its memory and processing overhead, was never designed to do. It also sidesteps the application compatibility expectations that come with Windows. Because Solara devices are built around cloud-hosted agents rather than traditional software, Microsoft can optimize the platform for dedicated AI hardware without carrying decades of legacy PC baggage.

To manage multiple agents running simultaneously, Microsoft is also working on an agent dispatcher and agent task manager — components that automatically surface or activate the right agent based on context, rather than requiring users to launch each one manually. Neither component is shipping yet. Early agent integrations include Dragon Copilot for healthcare workflows and GitHub Copilot for developer task tracking — both exploring how persistent, context-aware agents behave differently on dedicated hardware than they do inside a browser or IDE.

The platform appears to be aimed at enterprise buyers in retail, healthcare, and field service sectors, where dedicated agent hardware makes more sense than repurposing a smartphone. Microsoft has already lined up pilots with Best Buy, CVS Health, Levi's, Target, and AccuWeather, with broader OEM deployment targeted across healthcare, hospitality, financial services, legal, and industrial verticals.

When Intel introduced its IDM 2.0 strategy in 2021, a substantial part of the plan was to build custom x86 processors for clients and then produce them at Intel Foundry. However, the company has never assigned an executive to manage its custom silicon business, and the only major design wins were for custom Xeon CPUs used by Amazon Web Services. However, this month, Intel made two important steps that make it a de facto contract chip designer: it appointed an executive to lead its custom silicon business and signed a multi-year contract to build custom Xeon CPUs for Nvidia's AI platforms.

The current state of Intel's custom CPU business

Intel has been offering semi-custom Xeon processors to various customers for over a decade. These CPUs typically serve hyperscalers or large AI/data center customers who require performance or efficiency enhancements optimized for their specific workloads. These Xeon processors are tweaked from standard off‑the‑shelf Xeon SKUs with different frequency bins, power envelopes, packaging, microcode, feature sets, or even special-purpose accelerators designed for certain workloads.

When Intel discussed custom x86 processors in 2021, it mentioned customizable cores, custom IP, and customized Intel IP, which is far more impressive than its semi-custom offerings. However, the only custom products that Intel developed for a large client and disclosed publicly are the aforementioned Xeon CPUs for Amazon Web Services.

Intel has never revealed the degree of customization on the IP level with these products, though we know from AWS that its custom Xeon 6 CPU has an unknown number of cores, a 3.90 GHz all-core turbo frequency (up from 3.20 GHz for the off-the-shelf Xeon 6952P model), and faster DDR5-7200 memory support (up from DDR5-6400). However, such a level of customization is not something we usually expect from a bespoke CPU in a world where hyperscalers run dozens of highly customized models for their in-house processors inside data centers.

This is something that must change if Intel truly plans to serve a crowded market, with names like Alchip, Alphawave, AMD, Andes, Broadcom, GUC, Marvell, MediaTek, and Sondrel. This month, Intel appointed Srini Iyengar to lead its Central Engineering Group, enabling the company to build a custom silicon business serving a broad range of external customers. That job is not going to be easy, but Iyengar has the appropriate experience to do it.

Srini Iyengar has spent over two decades at Intel, with the latter half of his career focused on custom silicon architecture for infrastructure platforms. As a Principal Engineer, he has played a key role in architecting Arm-based Infrastructure Processing Unit (IPU) SoCs, defining product features to optimize performance, power, and area (PPA), and collaborating across IP vendors, verification, firmware, and manufacturing teams to deliver tailored solutions. Previously, he led the architectural development of special-purpose accelerator subsystems for server CPUs.

In addition, Intel this month disclosed its custom silicon unit's biggest win so far: a multi-year contract with Nvidia, under which it will develop and build bespoke Xeon CPUs for Nvidia's AI infrastructure. Given that Nvidia controls the lion's share of the AI hardware market, this is a significant contract both in terms of volume and in terms of Intel's public image.

Custom silicon is on the rise

When mentioning custom silicon in the context of Intel, we primarily refer to consumer and data center processors, as these are the areas where Intel excels. However, the semiconductor industry is witnessing a sharp rise in demand for bespoke application-specific processors across virtually all verticals, including AI, automotive, cloud, consumer, data centers, and consumer electronics.

Around a decade ago, only large companies could afford to develop their own custom chips, but with maturing contract chip development services, IP ecosystems, foundry yields, and a changing competitive landscape, the interest in bespoke chips is stronger than ever.

Apple's early lead in custom processors for smartphones set the tone for the consumer electronics industry, showing how proprietary silicon can deliver performance, efficiency, and product differentiation. In 2025, Google, Huawei, and Xiaomi have also developed their own smartphone SoCs.

In the data center, hyperscalers like Amazon and Google have become major drivers of custom silicon. AWS has its own Trainium accelerators for AI training, Inferentia accelerators for AI inference, and Graviton CPUs for general-purpose compute. Google produces its own chips for AI (TPUs), video (VCUs), and its own application processors for smartphones. These companies benefit from integrating hardware and software stacks, enabling better efficiency and lower costs at scale. The trend extends to other hyperscalers, including Alibaba, Baidu, Meta, Microsoft, and OpenAI.

Automotive manufacturers are also investing in their own processors (motivated by Tesla's early lead) as they shift to software-defined vehicles (SDVs). These companies are set to use multiple high-end SoCs across a vehicle, with the main Advanced Driver-Assistance System (ADAS) SoC likely using multiple chiplets. These companies are not only interested in reliability, performance, and features, but also in the long-term availability.

Advanced AI-assisted EDA tools from Cadence, Synopsys, and Siemens AI as well as simulation tools from Ansys (now part of Synopsys), greatly streamline the development of custom processors, which lowers the barrier for companies that intend to establish their own chip design division. If the return on investment looks compelling, unit costs are reduced significantly, or the total cost of ownership is lowered several times, or performance improvements are dramatic, then companies will at least consider starting an in-house chip design initiative.

However, development of chips is always a risk, both in terms of money and time-to-market. This is especially true for companies starting from scratch or lacking in-house expertise. Also, far not all companies that can benefit from custom silicon can afford an internal chip division. Finally, there are companies not willing to afford an internal chip design department for many reasons. This is where contract design-to-order service companies come into play.

What do customers want?

Companies that want the benefits of custom silicon, without the complexity of managing specifics like IP licensing, verification, or tape-out, have a very specific set of requirements for a contract chip designer, as it essentially becomes a strategic partner, not just a service provider.

One of the biggest priorities for interested companies is the experience of their partner. Customers seek design firms with a proven track record of delivering complex SoCs or ASICs, ideally in their industry domain. This includes not only silicon delivery but also successful tape-outs on advanced nodes, system-level architectural understanding, and familiarity with key verticals such as automotive, AI/ML, client PCs, and networking.

Secondly, customers expect access to a comprehensive IP portfolio. Most firms do not want to source and license every IP block individually; therefore, contract designers must provide or integrate essential IP blocks, such as PCIe, DDR, SerDes, Ethernet, USB, and security. Many customers also require the ability to integrate their own custom IP or differentiate through co-developed blocks. Therefore, the chip designer must have licensing flexibility, reuse rights, and expertise in deep IP integration.

Thirdly, clients demand a mature and automated design flow. Faster tape-outs and fewer silicon bugs are critical to stay on schedule and within budget. This also means the designer should handle validation, testbench creation, simulation, and signoff with minimal supervision.

Next, clients also value pre-existing relationships with foundries and Outsourced Semiconductor Assembly Tests (OSATs), which ensure that the designed processor will enter mass production and will ramp up to target volumes at predictable costs. Long-term support — including yield increase, silicon validation, firmware tuning, and product lifecycle management — is often a crucial factor.

Lastly, some clients prefer a full turnkey model, while others opt for joint development with an eventual handoff. Normally, a contract chip designer should support both and ensure strong IP protection, data security, and clear ownership terms.

What Intel can and cannot offer

Intel today can offer several — but not all — of the elements that customers expect from a contract chip designer. While the company is taking serious steps toward building a competitive custom silicon business, it still lags behind established ASIC players in a number of key areas.

Intel has decades of experience designing some of the world's most complex processors, including consumer CPUs for a wide range of PCs, Xeon processors for data centers, GPUs, FPGAs (via Altera), and even AI accelerators. The company clearly understands power, performance, and area (PPA) tradeoffs at scale.

For customers that require high-performance x86-based compute or custom server-class silicon, Intel's architectural know-how is likely second to none. Additionally, in recent years, Intel has gained experience in integrating chiplets made on different nodes from various foundries, a feat no one else in the industry has yet achieved in volume.

However, unlike Alchip, Andes, GUC, Marvell, or MediaTek, Intel does not have a proven track record of integrating Arm, RISC-V, or 3rd-party cores into customer designs.

Intel owns a wide range of IP, including x86 cores, GPUs (and supporting hardware like media engines, display engines, display controllers, audio codecs, etc.), AI accelerators, security engines, special-purpose accelerators (primarily for data centers), and high-speed I/O controllers and PHY (e.g., DDR, HBM, Ethernet, PCIe, Thunderbolt, UPI, USB, etc.). For customers looking to build a product around x86 and reuse trusted Intel IP blocks, the company can provide a strong starting point — particularly for data center and perhaps even for AI accelerators.

However, Intel's IP for its own 18A process technologies is relatively limited (for now), so it will have to license IPs from third parties like Synopsys, which is not a big problem, but adds complexity. While Intel has proven IPs for TSMC's process technologies, these are mostly focused on consumer solutions, not on data center solutions, which will again mean reliance on third-party IPs.

With Intel Foundry and packaging capabilities like EMIB, Foveros, and 3D chiplets, Intel can offer not only custom chip design but also manufacturing and advanced integration options. This is a strong differentiator versus design-only houses that rely on external foundries and OSATs. For chiplet-based SoCs or heterogeneous designs, Intel has a compelling packaging roadmap. Intel also has a good relationship with TSMC.

There might be a perception issue, though. Intel may favor its own IP, packaging, or node choices, which could limit design freedom. Some customers may prefer a neutral, foundry-agnostic partner that will deliver their GDSII file to TSMC.

Finally, Intel's traditional business model is focused on low-mix/high-volume chip development and production. A contract manufacturer is focused on agile development, tape-out, and low-volume production, and we have no idea how ready Intel's teams are for such work.

The first step

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.

Nvidia and MediaTek are expected to introduce their jointly developed Arm-based processors for PCs at Computex 2025, according to ComputerBase.

The upcoming chips — N1X and N1 — are aimed at desktops and laptops and mark Nvidia's deeper entry into the Windows-on-Arm ecosystem. But retail availability may be delayed until 2026 due to unresolved technical hurdles, reports German news site Heise, citing SemiAccurate.

CEOs of both companies — Jensen Huang from Nvidia and Rick Tsai from MediaTek — are scheduled to deliver back-to-back presentations at the Computex trade show in Taipei, Taiwan, on May 19 and May 20, respectively. One likely announcement is a new family of processors combining MediaTek's Arm-based CPU with Nvidia's Blackwell GPU. The collaborative products will rely on the companies' technologies to build the GB10 platform for compact AI workstations.

Through its partnership with MediaTek, Nvidia hopes to address markets that are currently served by AMD's APUs with high-performance Radeon graphics as well as Arm-based Snapdragon X processors. Nvidia's discrete Blackwell GPUs promise to offer higher performance and better compatibility with games than AMD's Radeon and Qualcomm's Adreno, so without any doubt, it will attract the attention of gamers.

According to early reports, the N1X and N1 processors are expected to feature up to 10 Cortex-X925 high-performance cores and up to 10 Cortex-A725 cores, though less powerful CPU configurations will likely be introduced to address markets currently underserved by Qualcomm and its Snapdragon X processors as well as AMD's Ryzen APUs with built-in Radeon graphics processors.

To support the production of an unknown high-volume product, MediaTek has reportedly secured a large amount of packaging capacity for flip chip ball grid array (FCBGA) chips, which implies their use in PCs rather than mobile devices. According to DigiTimes, this capacity reservation occurred around the end of 2024 and was unusually large. However, it remains unclear if the packaging timeline aligns with the initial production of processors jointly designed by MediaTek and Nvidia.

Despite the enthusiasm surrounding processors by MediaTek and Nvidia, the actual launch timeline remains uncertain. Multiple sources reported that development issues could significantly postpone the commercial rollout of systems based on the new processors. Some projections suggest that these delays could push the release into 2026, but for now, there is no confirmation on those rumors.

The global semiconductor industry saw explosive growth in 2024, mainly driven by sales of processors for AI applications, according to TrendForce. The Top 10 largest fabless chip developers earned nearly a quarter of a trillion dollars last year; roughly half of that sum came from Nvidia.

The largest fabless chip designers generated $249.8 billion in revenue, up 49% from the previous year. The growth boomed due to skyrocketing demand for AI GPUs, ASICs, adjacent chips (e.g., network processors, DPUs), datacenter CPUs, and recovered demand for client PCs. Market consolidation also intensified, with the Top 5 firms now accounting for over 90% of the revenue among the Top 10.

Nvidia was at the forefront of the industry and extended its dominance, posting $124.3 billion in revenue (a 125% increase from 2023) and capturing 50% of the revenue share. Hopper-based H100, H20, and H200 GPUs drove the company’s revenue increase, as Blackwell-based B200/GB200/B100 only emerged in the fourth quarter. As demand for Blackwell parts — which are believed to be more expensive than Hopper GPUs — increases this year, they will likely enable an even higher revenue for the green company this year.

Qualcomm ranked second, earning $34.86 billion, a 13% year-over-year increase. The company’s growth came from smartphones, the automotive sector, and PCs, a new source of the company’s revenue. The company secured a legal victory against Arm, and there are no risks that the latter will withdraw its licenses. The company also confirmed its interest in datacenter CPUs, though its entry into this market is likely a few years down the road.

Broadcom held third place, with its semiconductor unit bringing in $30.64 billion, up 8% from the previous year. AI-related products accounted for more than 30% of its semiconductor revenue. Despite a mid-year slump, demand for wireless communication, broadband, and server storage will drive the company’s growth in 2025.

AMD followed in fourth, increasing revenue by 14% to $25.79 billion. Its server business surged by 94%, boosting its position in datacenters and the cloud. Strategic partnerships with Dell, Google, and Microsoft are expected to help sustain its momentum, according to TrendForce.

MediaTek secured the fifth spot, with $16.52 billion in revenue, marking a 19% annual increase. The company’s success was driven by mainstream 5G smartphones, power management chips, and AI-related products. Its collaboration with Nvidia on Project Digits positions it for further expansion in 2025 as AI integration in mobile devices increases.

Marvell was the sixth largest fabless chip designer, with 5.637 billion in revenue, up 2% from the previous year. Realtek moved to seventh place with $3.53 billion (16% YoY growth), benefiting from a recovery in PC and automotive-related sales. Meanwhile, Novatek dropped to eighth, with revenue declining 10% to $3.2 billion.

Will Semiconductor and MPS closed the top ten, each showing 21% revenue growth, reaching $3.05 billion and $2.2 billion, respectively. Will Semiconductor benefited from high-end CMOS image sensors in Android phones and autonomous vehicles, while MPS saw success as its power management chips entered the AI server supply chain.

TrendForce expects AI to drive growth in various sectors, from datacenters to personal devices, in 2025.

Leading Taiwanese chip maker MediaTek is reportedly conducting simulations to assess the potential impact of U.S. tariffs. The move comes amid growing tensions and trade policy shifts that could reshape the global semiconductor industry. According to Reuters, MediaTek’s CEO, Rick Tsai, acknowledged the uncertainty surrounding the possible tariffs, emphasizing that while the situation remains fluid, the company is preparing for the worst.

"We are making our own assumptions, doing some simulations," Tsai stated to Reuters, refraining from providing specific details about the scenarios the company is analyzing. He expressed confidence that any potential impact in 2025 would be "manageable" but noted the situation's complexity, making precise estimates difficult.

If the U.S. imposes new tariffs, MediaTek could face increased costs when exporting its chips to American markets, potentially affecting pricing, supply chains, and profitability. The company relies heavily on manufacturing partnerships in Taiwan and China, which could expose it to trade restrictions depending on how the U.S. structures the tariffs. MediaTek will likely explore alternative supply chain adjustments and pricing strategies to mitigate risks.

Beyond trade concerns, the semiconductor industry is also navigating new competitive pressures. The emergence of DeepSeek, the Chinese startup offering efficient and cost-effective artificial intelligence solutions, recently caused a global dip in tech stocks. This development has raised questions about the long-term investment strategies of chipmakers and data centers. However, Tsai remains optimistic about AI’s future growth. "With the recent DeepSeek phenomenon, we actually are getting more optimistic," he said. "The trend is democratizing AI. It will spread more for average users."

MediaTek’s proactive approach underscores a broader industry effort to adapt to changing economic conditions. The company aims to anticipate potential disruptions and maintain stability in a highly competitive global market by running impact simulations. While the implications of future U.S. tariffs remain uncertain, MediaTek's strategic planning highlights the importance of flexibility and resilience in the semiconductor sector.

Similarly, TSMC is considering significant price increases of up to 15% for its chip production to offset potential losses in response to recent U.S. tariffs on Chinese imports. The move is expected to lead to higher costs for consumer electronics, as companies like Samsung, which rely on TSMC's chips, may pass these expenses onto consumers. TSMC is exploring shifting production to U.S. facilities currently under construction to mitigate tariff impacts.

Instead of increasing the physical data transfer rate beyond the 23 Gbps offered by Wi-Fi 7, the next-generation Wi-Fi 8—based on IEEE's 802.11bn Ultra High Reliability (UHR) specification—will focus on improving connection reliability and user experience.

Traditionally, new Wi-Fi iterations (as specified by IEEE 802.11 standards) have focused on maximizing data transfer rates by increasing channel bandwidth and number of channels and introducing new modulation methods. With Wi-Fi 7, the maximum PHY rate is 23 Gbps, though nobody expects to hit speeds that high. Also, the reliability of high-speed Wi-Fi connections leaves much to be desired. To that end, the next-generation Wi-Fi 8 iteration will not increase theoretical speed but will introduce new features designed to improve real-world performance and boost connection reliability, reports PC World, citing a MediaTek whitepaper.

On a high level, Wi-Fi 8 (802.11bn) resembles Wi-Fi 7 (802.11be): it uses 2, 4, 5, and 6GHz bands, the same modulation (4096 QAM), eight spatial streams, MU-MIMO, multiple OFDMA, and a maximum channel bandwidth of 320MHz.

However, according to the MediaTek paper, the new spec introduces several key features designed to improve real-world performance and connection speeds: Coordinated Spatial Reuse (Co-SR), Coordinated Beamforming (Co-BF), Dynamic Sub-Channel Operation (DSO), and enhanced Modulation Coding Scheme (MCS). Remember that we are talking about the standard as MediaTek sees it. Some features could be mandatory, while others could end up being optional.

The Coordinated Spatial Reuse (Co-SR) feature addresses varying signal strength between nearby devices and distant access points in high-density office scenarios. The capability enables APs to adjust and coordinate their power levels dynamically depending on the distance between devices and other APs to maintain the right signal strength. According to MediaTek's preliminary testing, this can improve overall system efficiency by 15%—25%.

Similarly, Wi-Fi 8's Coordinated Beamforming (Co-BF) advances previous beamforming technology by coordinating signal direction among multiple access points. This technology allows the system to avoid sending signals to areas and devices that are not needed, reducing interference and concentrating the signal toward active devices. In MediaTek's tests, Co-BF has improved throughput by 20%—50% in setups involving mesh networks shared in public spaces and some homes.

Dynamic Sub-Channel Operation (DSO) capability enables the network to assign sub-channels based on device requirements and abilities, increasing efficiency and boosting throughput by up to 80% (for advanced devices) while potentially avoiding bottlenecks.

Wi-Fi 8 will also incorporate refined data rates with additional levels in the Modulation Coding Scheme (MCS) lookup table, allowing devices to make smoother transitions in connection quality as they move through different areas. By adding finer gradations, such as a 16-QAM coding rate, Wi-Fi 8's MCS promises to reduce sudden drops in datarates, enhancing overall transmission stability and improving bandwidth by 5% to 30%, depending on the exact scenario.

The final Wi-Fi 8 standard is projected to be completed in 2028, and initial products based on the draft specification are anticipated in early 2028, pending regulatory approvals.

According to reports from leakers on the Chinese social media site Weibo, Mediatek and Nvidia are collaborating on a 3nm AI CPU. IT Home shared a report from user "Mobile Chip Expert" today, claiming that the CPU is entering the tape-out phase of production this month, with mass production on track for late 2025. 

We've reported on rumors of potential MediaTek and Nvidia collaborations for a few months now, so as usual, take this new report with a serving of salt. Previous rumors of a PC chip collaboration were the companies working on an AI card processor. However, today's reports are connected to rumors of an Nvidia/MediaTek AI processor that started in May.

The MediaTek AI processor is expected to be paired with an Nvidia GPU. The post also names Lenovo, Dell, HP, and Asus as prospective customers looking to adopt the processor in OEM hardware. The chip has also been linked to a rumored $300 price tag. 

Today's report curiously only refers to the rumored chip as an "AI PC" CPU, surprising as both previous leaks about the collaboration and MediaTek's stable of products would point to a mobile chip or SoC being more likely. MediaTek specializes in mobile products and typically employs AMD as a hardware partner. AMD uses MediaTek’s Wi-Fi 6E solutions in its mobile platforms, and MediaTek and AMD have also been connected through rumors of a 5G notebook for some time. 

It would not be surprising if the MediaTek/Nvidia chip transitioned to only an AI PC solution and moved away from being a mobile phone offering. After the big waves and lessons learned from the Qualcomm Snapdragon X Elite launch this year, there is room for another Windows-on-Arm competitor from a primarily mobile company like MediaTek. One of the most prominent problems reviewers had with the performance of Snapdragon X-powered computers was their lack of graphical strength, which Nvidia can quickly solve. Nvidia's name recognition alone will do much to boost the MediaTek chip, whether it is Nvidia's return to mobile graphics or a different venture into laptop-form-factor AI PCs.

MediaTek and Nvidia's only officially announced collaboration is the Dimensity Auto Cockpit platform, an SoC for cars that license Nvidia's graphics IP and Nvidia Drive OS. The chip seeks to be a front-runner in automotive chips by offering full support for infotainment platforms, including AAA gaming with RTX graphics and safety features such as HDR multi-camera support. So far, no US-based cars seem to have adopted the platform.

Nvidia is finally addressing a perceived deficiency in implementing G-Sync in the best gaming monitors. This morning, the company announced a partnership with MediaTek that will bring G-Sync to MediaTek’s upcoming scaler chips, thus eliminating the need for a separate (and costly) Nvidia-spec G-Sync Ultimate module.

Nvidia G-Sync, like AMD FreeSync, allows a monitor’s refresh rate to sync with the rate at which a graphics card can render each individual frame in a game. Depending on the game, the frame rate can vary wildly (particularly during viewing intense, complex scenes), so technologies like FreeSync and G-Sync ensure that the monitor and graphics card are in lockstep to eliminate tearing artifacts that can occur.

While G-Sync and FreeSync offer similar functionality, and you’ll find many monitors that support both implementations, FreeSync has always held one critical advantage – it doesn’t require additional hardware support. On the other hand, if a monitor manufacturer wants to support Nvidia’s full suite of G-Sync Ultimate features, they need to add a separate module, which adds cost.

Thanks to Nvidia’s new partnership, MediaTek's onboard scaler chip includes the full complement of G-Sync Ultimate technologies. MediaTek is a popular manufacturer of scalars that go into monitors, so there is a potential for cost savings to be passed on to the customer. When G-Sync was first introduced over a decade ago, the extra hardware necessary to add full support cost upwards of $200. In addition, the FPGA for G-Sync Ultimate requires active cooling because the chip runs hot. MediaTek is using a custom ASIC for its new scaler, which will have vastly lower power demands and will not require active cooling.

However, we don’t know how much the true cost savings will be for customers, as no financial details or licensing terms were revealed in the announcement. For example, will MediaTek have to pay a licensing fee to implement FreeSync into its scalers? And if a license fee is required, how much cheaper will it be compared to opting for a separate G-Sync module?

Nvidia was quick to point out that its newly announced Pulsar technology, which provides a claimed 4x uplift in motion clarity to reduce blur further while preserving fine detail, is also supported in MediaTek’s new scalers. The first monitors to use the new G-Sync-infused MediaTek scaler will be the Acer Predator XB273U F5, AOC Agon Pro AG276QS2, and Asus ROG Swift PG27AQNR. All three are 27-inch 1440p monitors with a 360 Hz refresh rate.

The partnership between Nvidia and MediaTek isn’t new. Earlier this year, MediaTek announced that it would integrate Nvidia’s next-gen graphics IP into its Dimensity Auto Cockpit SoCs used in automobiles.

Huawei Technologies has filed a lawsuit against MediaTek for alleged patent infringement, reports Nikkei. The case was filed in a Chinese court and is believed to be one of Huawei's efforts to increase its revenue by collecting licensing fees and royalties. Although MediaTek confirmed that Huawei had sued it, it did not reveal the nature of the allegations or which patents are involved. 

MediaTek is one of the world's largest suppliers of application processors for smartphones and other types of consumer electronics. In fact, the company used to supply system-on-chips (SoCs) for some of Huawei's own handsets. MediaTek's clients include major consumer electronics brands like Amazon, Samsung, Oppo, Sony, Vivo, and Xiaomi. Many of MediaTek's customers are either based in China or manufacture their devices in China, so banning sales of its processors in Tianxia would be devastating for MediaTek's business in general. Details about the lawsuit, including the damages Huawei seeks, have not been made public.

MediaTek confirmed the lawsuit in a filing with the Taiwan Stock Exchange, noting it would not significantly impact the company, but declined further comment.  

Nikkei's sources suggest Huawei's lawsuit aims to collect royalties to fund ongoing research and development. Huawei possesses a substantial number of standard-essential patents, particularly in 5G technology, which holds about 20% of global patents. Since 2021, the company has intensified its efforts to collect royalties, forming licensing agreements with various European automakers such as BMW, Mercedes Benz, and VAG (Volkswagen Audi Group).  

In 2022, Huawei earned $560 million from patent royalties. Around 200 companies, including Amazon, Samsung, and Oppo, currently pay for the use of its technologies. 

Huawei is not new to turning to litigation to collect royalties. The company sued T-Mobile in 2014, Samsung in 2016, and Verizon in 2020 over mobile connectivity patents. More recently, in 2022, Huawei sued Amazon and Netgear over its Wi-Fi 6 and Wi-Fi 5 patents. The company also sued Stellantis the same year to collect royalty fees for using its patents in cars such as Citroen, Fiat, Peugeot, and Opel.

Qualcomm competitor MediaTek is developing an Arm-based SoC designed to run Windows, according to three insiders who talked to Reuters. Meaning the choice of chip for your next laptop just got that much more complicated.

Qualcomm has been the sole Microsoft partner for Arm-powered Windows PCs since 2016, after the two firms signed an exclusivity deal. This venture led to the mid-May launch of Snapdragon X-powered Windows Copilot+ PCs, with Microsoft confident that the Snapdragon X Elite could surpass Apple’s M3 MacBook Air. Early benchmarks seem to confirm this increase in performance and battery life, giving Windows laptops a much-needed boost against the phenomenal efficiency of Apple-silicon MacBooks.

But this partnership is coming to an end soon: Arm's CEO confirmed that the Qualcomm-Microsoft deal expires this year. This means other manufacturers, like MediaTek, Nvidia, AMD, and even Intel, could get in on the Windows-on-Arm bandwagon. Qualcomm CEO Cristiano Amon even went as far as predicting that 50% of Windows PCs will be powered by an Arm chip in five years.

MediaTek has been working on Arm-powered Chromebook designs for several years, and was planning as early as 2022 to build Windows-on-Arm processors. So, with the upcoming expiry of the Microsoft-Qualcomm partnership, it makes sense for the Taiwanese company to enter the market as soon as legally possible. After all, designing and creating a chip doesn’t happen overnight; even Arm said that it takes at least nine months to build an SoC from ready-made components, with more complicated, custom designs requiring a year or more to reach mass production.

For what it's worth, MediaTek’s processors aren’t as popular as Qualcomm’s Snapdragon chips, yet MediaTek actually has a 36% market share in smartphone SoCs (compared to Qualcomm’s 23%), especially as its entry-level and mid-range chips are found in many Android devices.

Playing the Nvidia card

One advantage that MediaTek might be hiding up its sleeve is its history of partnership with Nvidia. The latter is the biggest desktop graphics card maker in the world, with an 88% market share, and it’s poised to launch its most powerful GPU yet this year, the RTX 50-series. But what many might not know is that these two companies have collaborated before: MediaTek’s Dimensity Auto Cockpit SoC sports integrated Nvidia graphics. We’ve also had reports from last year that MediaTek plans to use Nvidia GPU tech for its smartphone and Windows PC SoCs.

Stories about MediaTek’s PC ambitions were further fueled when its CEO, Rick Tsai, was spotted having dinner with Nvidia’s Jensen Huang, TSMC’s Morris Chang, and Quanta’s Barry Lam. These four CEOs are the key to making a complete Arm-powered laptop — TSMC for chip fabrication, MediaTek for the CPU, Nvidia for the GPU, and Quanta for the final laptop assembly.

Nevertheless, neither Microsoft nor MediaTek have commented on the news yet. It could be that MediaTek is still building its prototypes, and Microsoft is yet to look at them, so both parties are still tight-lipped about what is happening. Nevertheless, we would welcome MediaTek’s entry into Windows-on-Arm SoC production. After all, a little competition wouldn’t hurt the market, especially if it gives consumers a better and more affordable computer.

The heads of some of the biggest names in the chip making business were seen together at the Ningxia night market in Taiwan. It looks like that Huang took TSMC founder Morris Chang, MediaTek CEO Rick Tsai, Quanta head Barry Lam, and renowned architect Kris Yao with him for dinner. This get together happened on the eve of Computex 2024, and the group were mobbed by fans and the media in this rare public outing.

Nvidia CEO Jensen Huang is known to go out publicly and was spotted late last year eating street food in Vietnam, after discussing potential investments in the country with government leaders and company executives in Hanoi. 

Many of us know Nvidia as the largest supplier of graphics cards, and has also recently become the third most valuable company in the world due to the recent boom in AI tech. TSMC has also recently been in the news as becoming the most valuable semiconductor company, so much so that it had added self-destruct functions to its fabs in case of a foreign invasion. Kris Yao, who joined the group for dinner, is also one of Taiwan’s leading architects, and has designed several iconic buildings globally, including three TSMC facilities and the Quanta Research and Development Center.

MediaTek was also at the dinner table. Although lesser known than Qualcomm, the company has a major presence in the smartphone space, where it has a 36% SoC market share as of 23Q4. Furthermore, MediaTek has already worked with the company to roll out processors for automobiles that use Nvidia graphics. With the rumors of Nvidia wanting to develop Arm CPUs for Windows computers, MediaTek could be the company’s way into desktop and laptop computing, especially after its failed bid to acquire Arm in 2022. Last, but not least, we also saw the Quanta head with the group. Quanta Computer is a major Taiwan-based OEM laptop manufacturer. Some of its clients include the largest laptop brands in the world, including Apple, Dell, HP, Acer, Lenovo, and Microsoft.

No statements were made at that time, although Huang said that Taiwan is “one of the most important countries in the world.” However, this informal meeting adds more fuel to the rumor mill surrounding Nvidia’s plans. For example, there has been some talk about the company opening a second AI R&D center in Taiwan. So, it makes sense to have TSMC onboard this project, especially as Nvidia needs it to make its AI chips. If it plans to put a new building, then it needs to have an architect — and who else would be better suited for the job than someone who TSMC has worked with before?

There have also been reports of Nvidia working with MediaTek and Intel to build Arm-based CPUs for mobile computing. If you add Quanta into the mix, then this news starts to sound more credible, even though we haven’t received any official confirmation yet.

This public spectacle could be a sign of things to come for Nvidia. After all, informal deals made over the dinner table between heads of big companies could be just as binding, if not more so, than official announcements. Or it could just be a quick dinner between friends and acquaintances who happen to have some free time before one of the biggest tech expos in the world.

Driven by unprecedented demand for its H100 processor used for AI and HPC applications, Nvidia more than doubled its revenue in the calendar year2023 and became the world's largest fabless chip developer by revenue last year. The company did much better than the remaining nine companies in the top 10 and was among only four developers that posted revenue increases last year, reports TrendForce.

Revenue of the top 10 fabless chip design firms rose to approximately $167.6 billion, a 12% increase from the previous year, largely due to gargantuan contributions from Nvidia, which recorded a stellar 105% growth and earned $55.268 billion and reportedly captured over 80% of the AI accelerator market. TrendForce believes that Nvidia's ongoing dominance is set to strengthen in 2024 as the company is on track to release its Hopper-based H200 GPU and Blackwell-based B100, B200, and GB200 products.

Broadcom also saw positive growth: its 2023 revenue reached $28.445 billion, a 7% increase. This was partially fueled by the AI chip sector, which now accounts for nearly 15% of its semiconductor solutions, according to TrendForce.

Conversely, AMD faced a slight setback, with revenue dropping 4% to $22.68 billion, primarily due to reduced demand in the PC sector. However, growth in its data center and embedded businesses provided some offset, spurred by the company's acquisition of Xilinx.

On the other hand, Qualcomm and MediaTek — two major developers of application processors for smartphones — experienced declines due to a slump in the handset market. Qualcomm's revenue fell by 16% to $30.913 billion, impacted by poor performance in the handheld device and IoT sectors amidst low shipment volumes in China. Similarly, MediaTek saw a 25% reduction in its revenue to $13.888 billion, suffering losses across its smartphone, power management IC, and smart edge divisions. 

Looking ahead, the IC design industry is poised for further expansion in 2024, driven by demand for AI processors. The proliferation of AI is anticipated to extend into personal computing devices, such as PCs and smartphones. This shift is expected to fuel continued revenue growth across the sector.

Notably, the rankings among the top ten companies saw some shifts. MPS made its way onto the list with a 4% increase in revenue, reaching $1.821 billion, thanks to its strong performance in automotive and enterprise data sectors. In contrast, Realtek experienced a 19% decline in revenue to $3.053 billion, leading to a drop in its ranking due to faltering PC sales and halted projects in China. However, prospects for recovery are on the horizon for Realtek with the rise of Wi-Fi7 and renewed telecom tenders expected to drive the company's sales.

When we first heard rumors that MediaTek was working to integrate Nvidia's graphics IP into its processors last year, we were cautiously optimistic, as this would be the first time the green company licensed its graphics technology to a third party. And this week MediaTek introduced its new Dimensity Auto Cockpit system-on-chips for automotive applications. These pack Nvidia's next-generation graphics IP, marking a milestone for both companies. 

MediaTek's next-generation Dimensity Auto Cockpit SoCs are based on Arm Cortex cores featuring the latest Armv9 instruction set architecture, as well as Nvidia's next-generation GPU that can run artificial intelligence workloads, such as large language models. The GPU also supports real-time raytracing graphics and DLSS 3 image upscaling. The processors are supported by Nvidia's proven Drive OS (which is, of course, CUDA-enabled), which allows automakers to use all the software and IP designed for this platform. 

Nvidia has been trying to license its graphics IP since 2013 and its Kepler architecture, but without any success (or at least any announced success). Apparently, support for the Drive OS platform, CUDA, and highly competitive performance of Nvidia's next-generation GPU (possibly Blackwell?) is what made MediaTek license Nvidia GPU IP and software IP. 

"Generative AI is transforming the automotive industry in the same way that it has revolutionized the mobile market with more personalized and intuitive computing experiences," said Jerry Yu, Corporate Senior Vice President and General Manager of MediaTek’s CCM Business Group. "The Dimensity Auto Cockpit portfolio will unleash a new wave of AI-powered entertainment in vehicles, and our unified hardware and software platform makes it easy for automakers to scale AI capabilities across their entire lineup." 

The Dimensity Auto Cockpit CX-1, CY-1, CM-1, and CV-1 processors are highly integrated monolithic SoCs (according to HardwareLuxx) that have built-in integrated multi-camera HDR ISP that supports front-facing, in-cabin, and bird's-eye-view cameras for numerous safety applications, as well as an audio DSP that supports the latest voice assistants, according to MediaTek. 

"Generative AI and accelerated compute are reshaping the automotive landscape," said Ali Kani, Vice President of Automotive at Nvidia. "The new Dimensity Auto chipsets harness NVIDIA’s graphics and AI technologies, helping deliver fresh in-vehicle user experiences, added safety and security capabilities, and enhanced connectivity services spanning across vehicle segments."

Renowned hardware leaker @kopite7kimi corrected their own prediction that Nvidia's next-generation flagship graphics processing unit based on the Blackwell architecture would feature a 512-bit memory interface. It will not. Nvidia's GPU, currently known as GB202, will apparently continue to use a 384-bit memory bus, but will adopt GDDR7 type of memory.

Nvidia's GB202 GPU is projected to feature up to 24,576 CUDA cores, a 33% increase over the number of CUDA cores packed in the AD102 GPU (18,432 CUDA cores). Rumors suggest the company will use one of TSMC's 3nm-class fabrication processes to make its GPUs based on the Blackwell architecture — though it remains to be seen whether Nvidia and TSMC will customize TSMC's 3nm-class nodes for GPUs, or if they'll stick to the default.

"Damn, we must adjust our evaluation of [GeForce] RTX 5090/5080," wrote kopite7kimi in an X (formerly Twitter) post.

"I think I probably made an empirical mistake," said kopite7kimi in another post. "I mistakenly applied the ratio of Ada Lovelace's L2 [cache] and [memory controller] to Blackwell, [which led to an incorrect assumption regarding GB202's 512-bit memory interface]." 

When asked whether they meant a 384-bit memory interface for GB202 graphics processing unit, they replied positively, clarifying that the part will also use GDDR7. While a 512-bit bus would enable Nvidia to massively increase bandwidth available to its next-generation flagship graphics cards (presumably named the Nvidia GeForce RTX 5090), using GDDR7 over a 384-bit interface will also provide tangible benefits. 

If Nvidia uses Micron's 32 GT/s 16 Gb ICs, then its RTX 5090 will get 1.536 TB/s of memory bandwidth — that's up from the the 1.008 TB/s the RTX 4090, the best graphics card available today, gets. Yet, with 16 Gb ICs, Nvidia will still have to stick to 24 GB of memory on its premium consumer board.

Well-known hardware leaker @kopite7kimi, who has proven to typically know details about Nvidia's plans, has shared some of his thoughts about possible configurations of Nvidia's next-generation codenamed Blackwell GPUs. If his assumptions, presumably based on knowledge of certain details, are correct, then Nvidia's Blackwell may get quite an upgrade in terms of CUDA core count and memory interface. However, since this information is unofficial, it should be taken with caution.

"As I mentioned before, GA100 is 8 [GPC] * 8 [TPC], and GH100 is 8 [GPC] * 9 [GTC]," @kopite7kimi wrote in an X post. "[Compute] GB100 will have a basic structure like 8 [GPC] * 10 [TPC]. [Client PC] GB202 looks like 12 [GPC] * 8 [TPCs]."

Nvidia's GPUs are organized in groups of large graphics processing clusters (GPCs) comprised of smaller texture processing clusters (TPCs) that, in turn, consist of a group of streaming multiprocessors (SMs) that house actual CUDA cores. 

Assuming that the renowned hardware leaker is correct and Nvidia is set to maintain the number of streaming multiprocessors per TPC and CUDA cores per SM, then the company's compute GB100 GPU for artificial intelligence (AI) and high-performance computing (HPC) applications will get 20,480 CUDA cores (an 11% increase over GH100) in its full configuration, whereas client PC-oriented GB202 GPU will get 24,576 CUDA cores (a 33% increase over AD102) in its full configuration. 

Given the workloads that GB100 will encounter, it is logical for Nvidia to use 'fat' GPCs with as many TPCs per GPC as possible to minimize data exchange between TPCs as much as possible. At the same time, it might be reasonable for Nvidia to make GB202's GPCs somewhat 'fatter' than they are today to simplify the internal organization of the GPU. However, an all-new organization of GPCs will likely require driver optimizations.

In addition to a new microarchitecture and increased number of CUDA cores, Nvidia's GB100 is expected to get a 8,192-bit HBM3/HBM3E memory interface, whereas GB202 is projected to get a 512-bit GDDR7 memory bus, the leaker claims.

About a week ago, @kopite7kimi assumed that Nvidia's Blackwell might be the first GPUs to adopt a multi-chiplet architecture but did not elaborate. It is unclear whether he discussed the number of GPCs and TPCs across one chiplet or across a number of them. Keeping in mind that TSMC N3 process technology (which will presumably be used to make Blackwell GPUs) has limited advantages over N4 when it comes to transistor density and Nvidia's GH100 is already close to the maximum die size possible using existing lithography equipment, it is possible that Nvidia has only managed to squeeze 20,480 CUDA cores into GB100. In this case, it might be reasonable for Nvidia to use two GB100 dies for its next-gen compute GPU.

Assuming that Blackwell GPUs are set to arrive in late 2024 – early 2025 timeframe, the target specifications of Nvidia's next-generation Blackwell processors were set well over a year ago, and by now, the highest-end Blackwell GPUs have probably been taped out. Samples of these devices are probably being tested somewhere in the company's laboratories. As a result, there are hundreds of people at Nvidia who know the specifications of the company's Blackwell GPUs and can share some details with people outside of the chip designer.

That said, it is possible that @kopite7kimi has more or less accurate information about Nvidia's next-generation Blackwell GPUs. Yet, it is also possible that all the details about Nvidia's upcoming GPUs are essentially educated guesses rather than actual solid information.

Nvidia is set to adopt TSMC's 3nm-class process technology next year, a report by DigiTimes reads. The company is expected to produce its codenamed GB100 compute graphics processor using the technology. However, it is likely that the company will produce all Blackwell GPUs using the same fabrication technology.

The report claims that Nvidia's GB100 is projected to be launched in 2024, which will align with Nvidia's typical two-year cadence of launching new GPU architectures. The B100 will likely be Nvidia's next-generation compute GPU product for artificial intelligence (AI) and high-performance computing (HPC) applications. Meanwhile, the latest rumors suggest that the GB100 product will use a multi-chiplet design to increase performance tangibly compared to the GH100 product based on the Hopper architecture. 

It remains to be seen which of TSMC's 3nm-class process technology will be adopted by Nvidia. TSMC has numerous 3nm nodes, including performance-enhanced N3P and HPC-oriented N3X. Nvidia has used customized fabrication technologies for its Ada Lovelace, Hopper, and Ampere GPUs, so it is possible that the company will use a custom node with its Blackwell graphics processors as well. 

Nvidia will, of course, not be alone with the adoption of TSMC's N3 technology next year: AMD, Intel, MediaTek, and Qualcomm are all set to adopt one of the foundry's 3nm-class nodes in 2024 – 2025. In fact, MediaTek has already taped out its first N3E design with TSMC. 

At present, only Apple uses TSMC's N3B (1st generation N3) technology to make its A17 Pro system-on-chip for smartphones. The technology is expected to be adopted for other SoCs, including M3, M3 Pro, M3 Max, and M3 Ultra for Mac personal computers.

MediaTek this week announced that it had successfully taped out its first flagship smartphone system-on-chip using TSMC's 3nm-class fabrication process, which technically means it has beaten Apple to the first 3nm SoC. Based on the information disclosed by the SoC developer, we are dealing with an application processor set to be made on TSMC's N3E manufacturing technology, which makes it one of the industry's first chip to use this node.

Our focus is generally on PC hardware and mobile SoCs are generally not under our purview. That said, there are three aspects that matter in MediaTek's announcement: the fact that this is the industry's first officially disclosed N3E SoC, the announcement of a 3nm processor ahead of Apple, and MediaTek's relationship with Intel Foundry Services. Let's start with the most obvious one.

Beating Apple to 3nm

Formally, MediaTek beats Apple to the punch with a 3nm smartphone SoC, but there is a catch. Apple's 3nm mobile application processor is reportedly in mass production and is about to hit the market later this month when the company launches its iPhone 15 series. Meanwhile, MediaTek's next-generation Dimensity flagship will be made in 2024. 

Being TSMC's primary customer, Apple uses the foundry's leading-edge process technologies ahead of its rivals and it is generally believed that TSMC has been making 3nm SoCs for Apple since late 2022 using its most advanced production node, N3B.

The First N3E SoC

TSMC has two 3nm-class fabrication processes: the baseline N3 (also known as N3B) that can feature up to 25 EUV layers and can use EUV double patterning for higher transistor density. TSMC's other, a simplified N3E that can use up to 19 EUV layers and is not supposed to use EUV double patterning. TSMC's N3 offers smaller SRAM cells compared to N5 and a bit higher logic transistor density, but N3E provides more aggressive power (-32%) and performance (+18%) improvements compared to N5. These are indeed characteristics that MediaTek mentioned about TSMC's 3nm node it uses, which clearly points to N3E.

Although Apple is taking advantage of a higher transistor density with N3B, N3E promises a wider process window and potentially better yields, which is of crucial importance for costs. TSMC plans to initiate high volume manufacturing (HVM) using N3E towards the end of 2023.

It is quite remarkable that MediaTek disclosed its N3E tape out ahead of other TSMC clients (e.g., AMD, Nvidia, Qualcomm), especially keeping in mind that developers of mobile SoCs do not tend to make such disclosures. The reasons why the company decided to do so are unknown.

What About IFS?

MediaTek signed a strategic pact with Intel to use its advanced process technologies for a range of chips for client devices in July, 2025. In fact, MediaTek was the only big fabless chip designer to ink such a deal with IFS and disclose it publicly.

MediaTek's 3nm-related announcement with TSMC comes as Intel Foundry Services intensifies its marketing push, so the announcement may be considered as a way to distract attention from IFS. Meanwhile, MediaTek has so far not made any announcements regarding its tape outs with Intel's foundry division, which may not exactly be surprising as MediaTek probably intended to use Intel's 20A and 18A production nodes in 2024 – 2025 and onwards rather than Intel's 4nm and 3nm-class technologies in 2023 – 2024. 

In a live stream today, Chinese handheld gaming PC specialist Ayaneo unveiled its first device aimed at Android gaming. The new Ayaneo Pocket AIR looks a lot like the company's previous Windows-powered handhelds, but it manages to be thinner and lighter. Also, thanks to the Arm architecture, and retaining the "Windows handheld grade" battery, it is suggested that users can enjoy games "without worrying about battery life."

Android and Windows handheld markets may not really have a lot of overlap. On Tom's Hardware, we look most closely at PCs and PC gaming. PC AAA games titles are quite different from the library available on Android, except some common casual titles (and streaming services). 

Thus we think many readers won't be interested in this type of device for Android gaming, or games streaming, especially when they have powerful smartphones already. Nevertheless this device is worth a closer look to see Ayaneo's efforts in trying to jump into the Android market and compare the device with its Windows-powered lineage.

Starting with the basics, the Ayaneo Pocket AIR has a 5.5-inch 1080p AMOLED screen. The processor used is a 6nm MediaTek Dimensity 1200, which has 1x Ultra Core (Arm Cortex-A78 up to 3.0 GHz), 3x Super Cores (Arm Cortex-A78 up to 2.6 GHz), and 4x Efficiency Cores (Arm Cortex-A55 up to 2.0 GHz). Inside MediaTek's SoC there is also an Arm Mali-G77 MC9 GPU, support for various codecs, 5G, Wi-Fi, Bluetooth, and more. Despite the flagship smartphone guts, Ayaneo says that the Pocket AIR has an "x86-level powerful active cooling system." 

Similarly, it says this compact new device uses a "Windows handheld grade" battery, at 7,350mAh. Sadly, it hasn't been forthcoming with screen-on-time gaming endurance figures, so we will have to wait and see about extended battery life claims.

We have some physical details of the Ayaneo Pocket AIR to share, too. With its 5.5-inch screen you shouldn't expect a very hefty device, but at 380 g (13.4-ounces) it is about twice as heavy as a common-or-garden smartphone. The device's 'waist' is 17 mm thick. Ayaneo asserts that the Pocket AIR "fits easily in your pocket."

As the Ayaneo Pocket AIR is a pre-release, pre-crowdfunding stage device at the time of writing, we don't seem to have a great depth of specs, so details about memory and storage options seem to be absent right now.

Some other interesting details we have access to though, concern the controls. In brief, buyers of the Ayaneo Pocket AIR should expect a very similar extensive control set. So, here we have lots of buttons and pads including hall sensing joysticks and triggers, and there is the SoundTapMagic audio vibration system, an X-axis linear motor, and six-axis gyroscope included for your gaming delights.

On top of the Android installation, Ayaneo has ported its Windows gaming UI app, AYAHome, as well as the AYASpace management app. Another software nicety is the Ayaneo Cloud Game Database for retro gaming fans.

Running emulators looks like a focus of fun on the Ayaneo Pocket AIR, as there are many slides on the Ayaneo blog and Pocket AIR product pages highlighting this aspect of the device. Even its color scheme with "classic red and white," harks back to the 8- and 16-bit eras.

We don't have pricing or availability for the upcoming Ayaneo Pocket AIR. At the time of writing the firm is asking for sign-ups for news of its Indiegogo campaign launch. Let's hope it is priced competitively, to give this device some impulse purchase appeal. We've seen a lot of pricey handhelds from this firm recently, but it has made some sub-$300 devices in the past.

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.

MediaTek plans to integrate a graphics processing unit (GPU) based on one of Nvidia’s microarchitectures into its SoCs (system-on-chip) for smartphones and Windows-on-Arm mobile PCs, reports DigiTimes. Integrating a high-performance GPU could greatly improve their competitive advantages over rivals, such as Qualcomm and Samsung — the latter of which licenses AMD's GPU designs.

While licensing GPU technology has been Nvidia's plan for years, the information about a collaboration with MediaTek comes from an unofficial source and not all the details are clear. As usual, apply some skepticism and know that plans could change.

MediaTek and Nvidia are reportedly working on mobile handset platforms with MediaTek contributing its vast SoC expertise and Nvidia contributing its GPU architecture prowess to enhance the AI and graphics capabilities of smartphones. In addition, the two companies are also building more powerful SoCs for Windows-on-Arm notebooks, according to DigiTimes.

The report comes as a big surprise as MediaTek has been collaborating with AMD for a while, and AMD uses a variety of MediaTek’s Wi-Fi 6E solutions with its mobile platforms. In addition, the two companies are rumored to be working together on a 5G-enabled notebook solution.

But licensing Nvidia’s architecture has its rationale for MediaTek. First up, Nvidia and its GeForce brands are praised by gamers, so usage of Nvidia's GPU technology will inevitably attract more attention from gamers interested in smartphones and Windows-on-Arm laptops running MediaTek's SoCs. Nvidia is the leading supplier of the best GPUs, both for desktops and laptops (based on data from Jon Peddie Research), so its technology prowess is indisputable.

Nvidia has been trying to capitalize on its vast graphics IP for about a decade now, but it does not look like any SoC developer has ever licensed Nvidia's GPU architecture. By contrast, AMD has managed to license its RDNA 2 architecture to Samsung and sells a boatload of semi-custom SoCs for Microsoft's and Sony's game consoles.

Licensing the architecture to MediaTek could bring Nvidia back into the smartphones game, the market it failed to conquer over 10 years ago. Only time will tell whether this makes sense for Nvidia financially, though. On the one hand, technology licensing is a rather high-margin business. At the same time, this doesn't leave Nvidia a lot of options to maximize its profits, something the company is very capable of doing with its client and datacenter GPUs.

Arm has formed a new 'solutions engineering' team to develop prototype chips for smartphones, laptops, and other consumer electronics and demonstrate what its technologies are capable of, reports Financial Times. There are concerns in the industry that Arm plans to either sell chips itself, or license such designs, competing with its licensees, but sources close to Arm have denied plans to sell or license the product and insist that it is only working on a prototype. 

Traditionally, Arm licenses its instruction set architecture, logical designs of its CPUs or GPUs, silicon-proven physical designs of its CPUs or GPUs, and various other IP blocks to its clients. But this time around the chip being developed by the solutions engineering team, led by industry veteran Kevork Kechichian, is said to be more advanced than ever before, according to FT's source in the industry. The complexity of the project makes some people in the industry think that Arm may either build its own-brand SoCs, or at least license reference designs instead of licensing IP. Arm declined to comment on the information.

This is not the case, according to FT's sources close to Arm. The company reportedly only wants to develop a prototype chip or chips to showcase what its IP is capable of in terms of power and performance. Meanwhile, the development of complex system-on-chips is extremely expensive. There are estimates that a fairly complex 5nm SoC design could cost as much as $540 million (with software), whereas the development cost of a complex 3nm SoC may be as high as $1.5 billion with software included.

"Working on intellectual property is one thing but really designing and working with production partners to turn those efforts into physical chips is a totally different arena. It is more capital intensive," a former Arm executive with knowledge of the effort told the Financial Times. "At some point in the future [Arm] will definitely need returns to justify that massive investment.

Since Arm itself has not commented on the matter, we can only make guesses about what its solutions engineering team is about. Considering increasing chip design costs, there may be a rationale behind investing in chip designs for Arm.

For example, the company might be developing customizable silicon-proven reference designs containing IP that is guaranteed to work perfectly when implemented on given process technologies. Very few companies can afford to pour  $500 million – $1.5 billion into a chip design, but they might want to license something that is guaranteed to work.

Another reason for Arm to develop physical implementations of its IP is because, in the coming years, many of its customers may decide to license chiplets or chiplet designs instead of IP because of costs.

But in both cases, Arm might end up competing against its own customers, such as Qualcomm, MediaTek, NXP, and others that sell their chips to device makers. This will certainly make them more inclined to adopt other instruction set architectures, such as RISC-V and this is of course a strategic threat for Arm. Meanwhile, if smaller companies are unable to license its latest technology to stay competitive with bigger ones, they will go out of business or go with open-source RISC-V designs, which is also a strategic threat.

For obvious reasons, Arm needs to address a variety of strategic challenges before conducting its IPO later this year, so forming a 'solutions engineering' team might just be one of its moves toward that end.

System-on-chips based on Arm instruction set architecture now command a 14% market share, but they are poised to at least double their share in several years, reports Counterpoint Research. Analysts note two main advantages that Arm-powered SoCs have over their x86 counterparts for now: lower overall power consumption and customization enabling richer features and higher performance. 

Arm-based SoCs will command a 15% notebook processor revenue share of this year, but their share could increase to 25% in 2027. Meanwhile, Intel's share will decrease from 68% this year to 60% in 2027, whereas AMD's share will drop from 16.7% to 14.4%. 

Nowadays, the vast majority (90%) of Arm-based notebooks come from Apple, limiting their growth as Microsoft's Windows is the world's most popular operating system. However, Counterpoint believes that once Qualcomm develops competitive Arm SoCs compatible with Windows, Arm's market share in the PC world will increase. 

Arm-based SoCs have several benefits over x86 CPUs, such as reduced power consumption and better thermal management. Designed with energy efficiency in mind, these SoCs can be tailored to specific power needs, making them perfect for mobile devices and laptops where battery life is vital. Also, Arm-based SoCs can integrate a larger number of high-performance CPU cores and tightly integrated memory, something that not all x86 CPUs can offer. In addition, custom cores in these SoCs provide advanced features that off-the-shelf x86 processor cores do not offer these days, leading to enhanced hardware and operating system integration, the analysts note. 

Also, Counterpoint mentions the built-in AI capabilities of Apple's SoCs and believes that those capabilities can be used for such workloads as image and video encoding and compression, noise cancellation, image recognition and vocal enhancement. 

Counterpoint Research believes large PC vendors can introduce custom Arm SoCs tailored to their needs and featuring differentiation. For example, lower-power Arm processors can be used to build tablet-notebook hybrids that are lighter and longer lasting compared to today's 2-in-1 and convertibles. 

As more makers with knowledge of Arm SoCs enter the market (e.g., smartphone suppliers), they contribute their know-how in Arm-based hardware and software, further enhancing the appeal of Arm-based PCs. The growing availability of native Arm-based applications will also improve user comfort and familiarity with the platform. The shift towards Arm-based PCs is anticipated to persist, and their market share is projected to rise substantially in the upcoming years. 

What is noteworthy is that for now, Counterpoint Research makes no predictions about Arm adoption by desktops. While Arm SoCs have an indisputable advantage for laptops, desktops have different requirements. For example, they take advantage of high performance, and power consumption is not always crucial for desktops. Therefore, whether Arm will gain a significant share of the desktop space remains to be seen.

As a part of Intel's strategy of devouring businesses that do not exactly fit its IDM 2.0 strategy, the company this week quietly started to exit its wireless wire area network (WWAN) business. The move has been in development for a while and while Intel is set to cease making 4G and 5G modems itself, it will continue to provide notebook solutions based on its CPUs yet equipped with MediaTek-badged modems, reports More Than Moore. 

While the world is getting more connected, connectivity is perhaps becoming a commodity business for companies like Intel that strive to be vertically integrated yet sell their products at high margins. This is perhaps why the company sold its 4G and 5G smartphone/tablet modem business  to Apple back in mid-2019 and this is why it is selling the remaining business to MediaTek. For MediaTek, Intel's PC modem technology will enable it to offer connectivity solutions for the whole stack of client devices, starting from humble Internet-of-Things gadgets and ending up with advanced ultrabooks used by road warriors. 

After Intel sold its cellular phone modem business to Apple — the key customer of this business unit so to speak — yet retained its crucial 4G and 5G patents (at least some of which it inherited from Infineon as a part of its 2011 wireless solutions acquisition), it kept offering 4G and 5G modem solutions as part of its laptop platforms. At some point, the company decided to team up with MediaTek and Fibocom for manufacturing of appropriate modems.  

Now, the company plans to transfer its 5G technology to Fibocom and MediaTek and is currently facilitating the transfer of driver code and licensing agreements to maintain customer experience. Although Intel intends to exit the 5G market by July, it will retain a small team to aid MediaTek, which adheres hardware, software and clients. The technology transfer is expected to be completed by May and is not expected to have any financial impact on Intel. Intel's OEM partners using 5G solutions can continue to collaborate with MediaTek to provide updates and upgrades to their existing product roadmaps. 

"As we continue to prioritize investments in our IDM 2.0 strategy, we have made the difficult decision to exit our WWAN client business for both LTE and 5G," said Eric McLaughlin, VP, GM Wireless Solutions, Intel, in a statement issued to More Than Moore. We are working with our partners and customers to facilitate a seamless transition to support their ongoing business and ensure our customers continue to have solutions for the connected PC segment." 

For Intel, its 4G and 5G modem business may not be the most profitable one. Furthermore, as Analyst Dr. Ian Cutress points out, the total available market of always-connected PCs (ACPCs) has not increased in the recent years. While there might be an influx of a ACPCs from Apple (which has appropriate modem technology it inherited from Intel) and Qualcomm, Intel may not be inclined to compete with these two platforms given allegedly limited profit margins in this sector.

Meanwhile, offering a whole stack of connectivity solutions makes a great sense for MediaTek, which is going to get the business unit.

"MediaTek has a strong history of delivering global wireless solutions, and continues to make investments in its wireless modem portfolio beyond the smartphone into PCs and other devices," a statement by MediaTek cited by More Than Moore reads.

The market is filled with Raspberry Pi clones that promise to be faster, cheaper or have more ports. However, most of these single-board computers use cheap processors from lower-end brands such as RockChip and Mediatek. A new player, the Thunderberry5, uses a Qualcomm Snapdragon chip, the same brand found in many major brand phones, tablets and Arm-powered laptops.

An upcoming board from French company MakeMyBoard, Thunderberry5 claims to be "the first Raspberry Pi-like SBC based on Qualcomm AI-CPU" and powering the show is a Qualcomm QCS610 Snapdragon. The Kryo 460 Octa-core CPU is powered by two Gold 2.2 GHz cores, and six Silver 1.8 GHz cores. That sounds like it would be faster than the 1.5 to 1.8-GHz CPU in the Raspberry Pi 4 B, but we wouldn't know for sure without testing.

The SoC's Qualcomm AI Engine (AI Stack and Neural Processing Engine) provides the power for general AI duties, making the board an interesting platform for machine learning and robotics. Graphics duties are carried out by an Adreno 612, clocked at up to 845 MHz. Thunderberry5 comes with 4GB of LPDDR4 RAM and 64GB of eMMC 5.1 on which the OS choices are currently Android 10 or Yocto embedded Linux. 

The curveball with this board is the embedded RP2040 microcontroller. An embedded microcontroller isn't new. We have reviewed Seeed's Odyssey and LattePanda 3 Delta, each of which had an Intel CPU and an Arduino compatible Atmel microcontroller. But, this is the first board that we have come across to feature the RP2040 as an embedded microcontroller. 

In the block diagram, we can see that the "RP4020" -- surely a typo for RP2040 -- is connected to a 2 x 20 pin header GPIO labelled "HAT_conn". We cannot ascertain if this is pin compatible with the best Raspberry Pi HATs and we have contacted the project creator for clarification. We can see in the block diagram that the RP2040 is baked into the board and uses a USB 2 to serial interface for communication. This means that it will appear as a device to the underlying Linux operating system.

Other than the specs and some images, we don't know too much about this Raspberry Pi competitor, but we have contacted the creator to learn more about Thunderberry5. For now, the price and release date are a mystery, but we will update this story once we have more information.

Although unit shipments of PCs decreased by around 15% in 2022 compared to the prior year and are expected to drop further in 2023, sales of Arm-based notebooks increased last year and will increase again this year, reports Counterpoint Research. Apple dominated the Arm-based laptop market last year with a 90% share, but as MediaTek and Qualcomm introduce their latest Arm system-on-chips for Windows PCs, such processors will gain even more share and will be in 25% of notebooks by 2027, the report says. 

Counterpoint Research predicts that Arm-powered laptops will "show a comparatively resilient demand throughout the coming quarters" due to success of Apple's MacBooks, a vanishing performance gap with x86 CPUs and strong ecosystem support. Indeed, while the whole PC market contracted by 28.1% in Q4 2022 compared to the same quarter a year before, unit shipments of Apple's PCs declined by 2.1%, according to IDC. Meanwhile, the company's PC shipments increased by 2.5% year-over-year in 2022 and commanded 9.8% of unit shipments, based on data from IDC.  

Given Apple's focus on laptops as well as steady demand for Arm-powered Chromebooks, 13% of mobile PCs sold last year used an Arm-based SoC, says Counterpoint. The share of notebooks with Arm inside is expected to increase to 15% this year, according to analysts. 

But while Apple's success with its Arm-based Macs is indisputable, most people use Windows machines, so once companies like MediaTek and Qualcomm introduce  their new Arm-powered SoCs for Windows in 2024, migration of mobile computers to Arm from x86 CPUs will accelerate. As a result, the share of Arm-based laptops will increase to 21% already in 2025 and then to 25% in 2027, predicts Counterpoint.  

Indeed, Qualcomm has very high hopes for its Snapdragon SoCs featuring Qryon general-purpose cores designed by Nuvia. These processors are now sampling with PC OEMs and the company claims that they intend to use these chips for laptops aimed at both consumer and enterprise users. 

"Our next-gen PC platform, with integrated custom Qualcomm Oryon CPUs and upgraded AI engine, has sampled on time and is exceeding our internal KPIs, delivering disruptive CPU performance per watt across tiers," a recent statement by the company reads. "We are now engaged with major PC OEMs, with multiple platform design wins across their product roadmaps for consumer and enterprise."

While Apple, MediaTek and Qualcomm will keep pushing performance and features of their Arm-based SoCs forward in the coming years, AMD and Intel will not stand still and watch how Arm is eating their lunch. Intel says that its laptop-focused Lunar Lake SoCs featuring a multi-tile design and a brand-new microarchitecture will offer performance-per-watt leadership.

One of the least-noticed highlights of Intel's earnings call this week was the announcement that its Intel Foundry Services contract chipmaking division landed an order from a major "cloud, edge, and datacenter solutions provider" for producing chips on Intel 3 production node. While this is a significant win for IFS, problems linger, as the Chinese antitrust agency has suspended its review of the Intel-Tower transaction.

A Major Win

"I am very happy that we were able to add a leading cloud, edge, and datacenter solutions provider as a leading-edge customer for Intel 3," said Pat Gelsinger, Intel CEO. "Including prior customers such as MediaTek, we now have a lifetime deal value of greater than $4 billion for IFS." 

The company has always tailored its Intel 20A and Intel 18A (2nm and 1.8nm-class, respectively) for its foundry customers. In contrast, its current and upcoming Intel 7 (10nm Enhanced SuperFin), Intel 4 (7nm with EUV), and Intel 3 (7nm+ with EUV) nodes were designed primarily with Intel in mind. So winning a major customer with Intel 3 — which is set to be production ready in 2024 — is clearly an achievement for the processor giant.  

Meanwhile, it is possible that Intel will be producing a semi-custom version of its Xeon processor for a major client, just like it envisioned when announcing its IDM 2.0 business model a little less than two years ago. 

Intel's IFS has been steadily gaining momentum. The foundry unit has already signed Qualcomm and Amazon Web Services (AWS) as initial customers and won a contract from the U.S. Department of Defense. In addition, Intel says it is actively working with many of the largest fabless chip designers on specific engagements, which is quite an achievement for a contract chipmaker that has been in the business for less than two years. 

"We also have an active pipeline of engagements with seven out of the 10 largest foundry customers coupled with consistent pipeline growth to include 43 potential customers and ecosystem partner test chips," said Gelsinger. "Additionally, we continue to make progress on Intel 18A, and have already shared the engineering release of PDK 0.5 (process design kit) with our lead customers and expect to have the final production release in the next few weeks."

A Major Bump

However, not everything is rosy for Intel Foundry Services. The proposed $5.4 billion acquisition of Tower Semiconductor, a specialty foundry with a vast portfolio of clients, has hit a bump. According to Seeking Alpha, China's State Administration for Market Regulation (SAMR) suspended the clock in its review of the transaction. 

It is believed that Intel will not invest in its Chinese operations as U.S. authorities prohibit investments in China for companies that receive money or other benefits from the CHIPS fund. Meanwhile, if the deal is not approved by SAMR, it will not move forward. 

Intel needs Tower to gain an experienced foundry executive team and a huge portfolio of clients. In fact, the former head of IFS already stepped down back in November as Intel planned to close the deal in Q1 2023. Now that the deal cannot close, IFS is operating without a head, which is not particularly good for business. However, Intel remains optimistic about the Tower buyout. 

"In addition, we are working hard to complete the Tower acquisition, which will further amplify our momentum as our foundry business becomes even more compelling to customers," said the head of Intel.

TSMC remains poised to produce CPUs, GPUs, SoCs, and FPGAs for AMD on its leading-edge fabrication processes for years to come. However, AMD will keep outsourcing production of its older products to GlobalFoundries and Samsung Foundry, according to DigiTimes. Given the current geopolitical situation, one might speculate that the chip designer is exploring SF as a potential alternative not only to GF, but to TSMC as well. 

Older Products for Now

DigiTimes claims that TSMC will make Zen 4 and Zen 5-based products for AMD using its N5/N4 (5nm and 4nm-class) and N3 (3nm-class) fabrication processes in the coming years. Meanwhile, GlobalFoundries and Samsung Foundry will produce AMD's processors based on older Zen and Zen+ microarchitectures as well as previous-generation GPUs using SF 14LPP and GF 12LP production nodes.  

This is not the first time we've heard of AMD working with Samsung Foundry to produce some of its chips. For example, last August, TechGoing reported that AMD would outsource some of its older products to Samsung Foundry. Keeping in mind that GlobalFoundries licensed Samsung's 14LPE and 14LPP technologies, it is easy for AMD to dual-source some of its older products. In fact, AMD confirmed to analyst Patrick Moorhead in 2016 that it could use Samsung Foundry's capacities if needed.  

GlobalFoundries is busy producing chips for various customers, including Intel, Qualcomm, MediaTek, and NXP. So AMD might want to secure additional 12nm/14nm-class capacity at Samsung Foundry for a just-in-case scenario. Meanwhile, considering the current geopolitical situation, AMD may explore Samsung Foundry as a potential alternative to TSMC. Here is why.

Possibly Something New Later

Once AMD's leading production partner, GlobalFoundries ceased to develop leading-edge process technologies in 2018 as it pivoted to specialty production nodes. When the company switches to a specialized sub-10nm fabrication process, the bulk of AMD's primary requirements will probably shift to 5nm-class or even 3nm-class manufacturing technologies. As a result, AMD might outsource some of its parts requiring specialty nodes to GlobalFoundries, but we would not expect GF to produce mass-market products for AMD soon. 

By contrast, Samsung Foundry keeps developing leading-edge fabrication technologies as it competes fiercely against TSMC. While Samsung Foundry's yields on leading-edge fabrication processes are reportedly not as high as those of TSMC, the company can offer its services to chip designers that require leading-edge technologies. 

For now, the skies are blue for TSMC. But if tensions between China, Taiwan, and the U.S. lead to a military conflict or even China's invasion of Taiwan, AMD will need an alternative to the world's largest foundry. For now, the only other option for TSMC is Samsung Foundry, as Intel Foundry Services' (IFS) 20A technology won't be available to fabless chip designers until 2024.  

Of course, AMD is almost certainly exploring IFS and its future nodes, just like other leading chip developers.

Some Thoughts

DigiTimes and TechGoing both reported about AMD's outsourcing to SF within a matter of six months. Working with Samsung Foundry allows AMD to learn how to collaborate with this contract maker of chips, how to use its packaging, and what to generally expect from SF. 

AMD's possible intention to use Samsung Foundry's leading-edge production nodes if something occurs to TSMC is our speculation not confirmed by any industry source at this point. To use Samsung Foundry's advanced technologies, AMD would need to redesign its chips for SF's nodes, which costs hundreds of millions of dollars, which is a lot of money even for AMD. Yet, if the situation between China, Taiwan, and the U.S. becomes dire, AMD's experience with GlobalFoundries could become more than valuable. 

Being the world's No. 1 foundry, TSMC was almost immune to the softening high-tech market in Q2 and Q3, but it looks like the party is over even for the world's largest contract chip manufacturer. TSMC's leading customers began to revise their orders to the company, affecting the foundry's results starting from the fourth quarter of 2022, reports DigiTimes. 

According to the report, virtually all TSMC clients will experience a downturn and have to cut orders, so TSMC's utilization will decline significantly in Q1 2023. For example, the utilization rate of TSMC's N7-capable lines (7nm, 6nm-class technologies) will decline to around 50% in early 2023. Furthermore, even TSMC's N5/N4-capable lines will be underutilized, though this may not come as a surprise since these are used to make leading-edge products, like Apple's smartphone SoCs, and demand for advanced handsets is usually down in the first half of the year. More alarming is that even N28-capable fabs — which have been fully loaded since the beginning of the chip deficit in early 2021 — will be underutilized. 

Due to the slowing economy in China as well as its COVID lockdowns, an economic downturn in numerous European countries, and reduced demand for many products in the U.S., large computer hardware, PCs, and smartphone makers lowered their procurement of new chips from companies like AMD, Intel, MediaTek, and Nvidia. As a result, the fabless chip designers reportedly had no choice but to slash orders to TSMC. 

The order cuts will come into effect in Q4 2022, which will increase TSMC's inventories, though it is unclear how the significantly reduced orders will affect the company's revenue this quarter. Meanwhile, DigiTimes estimates that TSMC's Q1 2023 sales will decrease by 15% quarter-over-quarter. By contrast, TSMC's Q1 2022 revenue exceeded its Q4 2021 revenue by 12.1%. 

Slashing orders to a contract manufacturer is not trivial since fabless chip designers are obliged to procure a fixed number of wafers in certain quarters. Nevertheless, TSMC is reportedly willing to accept compensation (as it will hold wafers with chips from AMD, Intel, Nvidia, etc., before they are ready to buy them) and even renegotiate deals on long-term supply contracts (i.e., increase the number of wafers that a company is committed to buying in the future) in exchange. Such measures will not make TSMC's life any easier in Q4 2022 or Q2 2023, though. 

Market observers generally remain optimistic about the demand for advanced chips returning to normal in 2023. As a result, TSMC is still expected to post revenue growth for the year, but the sales increase may not be as impressive as the 2021 – 2022 period.

Randhir Thakur, the head of Intel's contract chip production unit, has resigned, according to a report from The Register that was confirmed by Intel. He will continue to lead Intel Foundry Services through Q1 2023 to ensure a smooth transition to a new leader. 

Pat Gelsinger, chief executive of Intel, sent an email to the company's employees thanking Randhir Thakur for helping establish IFS and being instrumental to the company's IDM 2.0 business model. 

"Randhir has been a key member of the Executive Leadership Team for the past two and a half years and has served in several senior leadership roles since he joined us in 2017," Intel's CEO wrote in the email quoted by The Register. "His contributions to our [Integrated Device Manufacturing] 2.0 transformation are many, but most notable is his leadership in standing up our IFS business." 

Indeed, Randhir Thakur has done quite a lot to be proud of at IFS. During his tenure, Intel announced the pending acquisition of Tower Semiconductor (which would immediately make Intel one of the biggest contract makers of chips). He was also instrumental in inking deals with giant chip developers like MediaTek, which also happens to be one of TSMC's largest customers. This contract was a major win for IFS, and the outfit also won a contract with the U.S. Department of Defense for the Rapid Assured Microelectronics Prototypes - Commercial (RAMP-C) program. 

But IFS is far from an ideal contract chipmaker, at least based on what we know about Intel's process technology roadmap through 2025. The company's plans announced so far largely align its production nodes to its own fabrication processes, which is good for Intel as an IDM (Integrated Design Manufacturer) player, but might not be enough to persuade companies like Apple, AMD, and Nvidia to use IFS services for their high-volume products. 

Intel's most logical next move is perhaps to hand the foundry division over to the management of Tower Semiconductor. The transaction is set to close in February 2023, which is around the time Randhir Thakur is set to leave. Tower Semiconductor has developed its roadmap together with its customers for years, so the management experience will be critical for the future of IFS.

The Intel Foundry Services business unit earned $171 million in the company’s fiscal third quarter of 2022, which may not sound much given that it is about 1.1% of the company’s revenue. The company has yet to offer its leading-edge technologies to big customers.

Apple's rapid transition to Arm-based system-on-chips has shown the industry how swift such changeover is possible if architected correctly. Analysts from Canalys believe that Arm's architecture advances so fast that Arm-powered SoCs will grab a sizeable share of the PC market and half of the cloud server market in just four years. But not everyone in the industry is so optimistic.

"By 2026, not 2050 but 2026, four years from now, half of the cloud processors will be ARM-based, 30% of PCs will be ARM-based," said Steve Brazier, president, and CEO of market research firm Canalys, at an event, reports DigiTimes. "It is an extraordinary event and an industry-changing event that simply has not been taken seriously enough.

30% of PCs by 2026

Arm already controls a sizeable chunk of the PC market. Almost 100% of the PCs that Apple sells are based on the M-series SoCs these days, and in Q3 2022, it controlled 13.5% of the PC market regarding units, according to IDC. Moreover, the company increased its unit shipments from 7.174 million Macs in Q3 2021 to 10.060 million systems in Q3 2022 amid deteriorating demand, which is quite an achievement. Meanwhile, Arm-powered SoCs also power loads of inexpensive Chromebooks. While such systems are not very popular, it is probably safe to say that the Arm architecture already commanded at least 15% of the PC boxes sold in Q3 2022.

According to Dean McCarron of Mercury Research, this assumption may be a bit too optimistic. He estimates that in Q3 2022, Arm commanded 13.1% of PC client processors, up from 9.4% in Q2 2022 and up from 8.9% in Q3 2022. It should be noted that while IDC counts PCs sold, Mercury Research counts sold-in CPUs and GPUs, which might get sold this quarter or next. Given inventory correction at PC makers and increasing sales of Apple's PCs, it is reasonable that the share of sold-in x86 CPUs dropped in the third quarter. In any case, the share of Arm's processors is growing because of Apple and Chromebooks.

"We note that within the Chromebook market, is appeared that Arm's share looks to have increased in the quarter, though the market was very weak," McCarron told Tom's Hardware.

Arm vs x86 Consumer Market Share Q2 2022

The head of Canalys believes that Arm has the potential to grow further in the PC space, especially if other leading PC suppliers like HP, Dell, Lenovo, Acer, and Asus come up with their own Arm-based SoCs. However, he never mentioned — at least according to the DigiTimes story — Qualcomm's efforts to drive its Snapdragon platforms into always-connected laptops by Lenovo and HP and then come up with something more powerful based on technologies from Nuvia. Yet, Qualcomm could become a significant player, particularly in notebooks.

We can hardly imagine HP, Dell, or Lenovo developing their own Arm-based SoCs for their PCs to differentiate from each other and offer unique competitive advantages. Developing chips are expensive nowadays, and since these companies sell boatloads of PCs featuring completely different configurations, they will need several different SoCs to address all of their markets. Keeping in mind the relatively thin profit margins on the PC market, it is unlikely that these companies will try to make their processors replace those from AMD or Intel. Meanwhile, they may start ordering customized versions of chips from the leading CPU designers. With multi-chiplet designs like Intel's Meteor Lake, it should be pretty easy to add third-party IP to the processor, though we are speculating.

It remains to be seen whether the Arm architecture will indeed capture an additional 15% - 17% of the client PC market in the next four years because Apple gains share (although we do not expect Apple to command one-third of the PC market by 2026) or because the joint effort of Apple, Qualcomm, and MediaTek will play its role. But given that so much work is happening around Arm, the technology is poised to increase its adoption.

Meanwhile, Arm adoption in the Windows space will largely depend on the experience of Arm-based machines running Windows 12, which happens to be when Qualcomm rolls out its Nuvia-based Snapdragons.

"It will be an incredible shock for Intel and AMD, to suddenly find themselves fighting to protect their business," said Brazier.

50% of Cloud Servers by 2026

Servers are another frontier that the Arm architecture entered reasonably recently thanks to datacenter-grade SoCs from Ampere, AWS, and Huawei. According to Omdia, Arm SoCs controlled around 7.1% of the server market in Q2 2022. Omdia only tracks data center machines, so we are talking about blades, rack servers, white-box servers used by hyperscalers, tower servers, and hyper-converged infrastructure servers, but not about edge servers, mission/business critical machines, and other niche markets.

In the coming years, we are going to see more Arm server processors from other makers, such as SiPearl and Chinese cloud giants, assuming that they can develop SoCs that they can produce in volume with decent yields and ship them to China without violating the recent U.S. export regulations. Therefore, the share of Arm in data centers will increase.

As costs of running cloud machines are increasing, designing highly custom SoCs for specific applications makes sense, and CPU cores based on the Arm architecture can be customized to meet the requirements of particular applications. Moreover, SoCs are, by definition designed with specific requirements in mind.

In fact, even Intel acknowledges that custom chips make a lot of sense for cloud servers, which is why it is willing to produce both Arm- and x86-based SoCs for IFS customers as part of the IDM 2.0 strategy.

But neither Intel nor AMD is giving up half of the cloud server market without a battle. AMD is set to release its 128-core codenamed Bergamo processor based on the Zen 4c microarchitecture optimized for cloud and hyperscalers. Intel, on the other hand, is building-in loads of special-purpose accelerators into its 60-core Sapphire Rapids CPU to address the specific needs of both enterprise and hyperscale customers.

Not everyone needs Arm-based servers, though. Enterprises that use tens of thousands of servers will unlikely adopt Arm since they run applications developed for x86, and Arm's advantages for them are not obvious. Therefore, Brazier believes cloud servers will be the primary market that Arm-based SoCs will address initially. He believes that Arm's advantages in this market are so significant that the architecture will be used by 50% of cloud server SoCs.

In the coming years we are going to see more Arm server processors from other makers, such as SiPearl and Chinese cloud giants, assuming that they can develop SoCs which they can produce in volume with decent yields and ship them to China without violating the recent U.S. export regulations. Therefore, the share of Arm in datacenters will increase.

As costs of running cloud machines are increasing, designing highly-custom SoCs for specific applications makes sense and CPU cores based on the Arm architecture can be customized to meet requirements of very specific applications. Moreover, SoCs are by definition designed with specific requirements in mind.

In fact, even Intel acknowledges that custom chips make a lot of sense for cloud servers, which is why it is willing to produce both Arm- and x86-based SoCs for IFS customers as part of the IDM 2.0 strategy.

But neither Intel nor AMD are giving up half of the cloud server market without a battle. AMD is set to release its 128-core codenamed Bergamo processor based on the Zen 4c microarchitecture optimized for cloud and hyperscalers. Intel, on the other hand, is building-in loads of special-purpose accelerators into its 60-core Sapphire Rapids CPU to address specific needs of both enterprise and hyperscale customers.

Not everyone needs Arm-based servers though. Enterprises that use tens of thousands of servers will unlikely adopt Arm since they run applications developed for x86 and Arm's advantages for them are not obvious. Therefore, Brazier believes cloud servers will be the primary market that Arm-based SoCs will address initially. Apparently, he believes that Arm's advantages on this market are so significant that the architecture will be used by 50% of cloud server SoCs.

Taiwan's parliament is scheduled to discuss a draft of Article 10-2 of the 'Industrial Innovation Act' tomorrow. This could be highly impactful in Taiwan, and worldwide, as it is being characterized as Taiwan's own CHIPS Act. Overall, this legislation looks like an understandable reaction to the recent flight of semiconductor investments out of Taiwan.

The key attraction of the Industrial Innovation Act for domestic businesses, and what is hoped that will help steer investment back into the island, is a 25% tax break on R&D expenses for high-tech companies. There are also going to be tax breaks on the purchase of manufacturing equipment, compensated with up to an additional 5% cut in tax. Thus it could be possible to see tech businesses in Taiwan enjoying and attractive and healthy 30% tax reduction.

According to reports from sources like Taiwan's Central News Agency (CNA), the amendments to business taxation will be discussed by the ministers of the government's Executive Yuan on Thursday. It is widely expected that the amendments, with the tax breaks, will be then passed to the Legislative Yuan for further deliberation and voting.

The Taiwanese Ministry of Economic Affairs' official explanation for the new 'Taiwan CHIPS Act' is that it will encourage homegrown tech businesses like TSMC, MediaTek, GlobalWafers, UMC, Phison etc to invest more in Taiwan, especially for advanced research and development. This and the tax break for equipment purchases should encourage breakthroughs in advanced manufacturing on Taiwan's home turf.

Regular readers will be well aware that big Taiwanese tech businesses have increasingly been attracted to build facilities in countries like the USA and Japan since the tightening of US tech sanctions on China. However, highly important and strategic businesses like TSMC have pledged to keep leading edge plants in Taiwan. Whether the Taiwan CHIPS Act will be successful in rebalancing internal and outside investments to the extent the government wishes, remains to be seen. Foregoing tax revenue in the short term, with these tax breaks, must be seen to be working or political opponents will highlight this policy in the presidential elections in 2024.

After becoming one of the leading suppliers of system-on-chips for inexpensive Chromebooks, MediaTek wants to address the market of Windows on Arm PCs. To meet the performance expectations of Windows users, MediaTek plans to develop SoCs with enhanced CPU and GPU performance; the company reiterated this week.

"In CPU and GPU we are having to make some bigger investments as a foundational capability [for PC-oriented SoCs]," said Vince Hu, a corporate vice president of MediaTek at the company's event, reports PC World.

MediaTek's Kompanio platforms for Windows on Arm PCs will include 'some of the technology' applied to high-end Dimensity SoCs for smartphones as well as 5G, Bluetooth, Wi-Fi, and even display driver ICs (DDICs) specifically designed for laptops, the company revealed.

The company's latest Dimensity 9200 SoC for smartphones features eight general-purpose CPU cores and an 11-cluster graphics processing unit with hardware raytracing support. The CPU department includes one Arm Cortex-X3 performance-enhanced core operating at 3.05 GHz, three Cortex-A715 high-performance cores at 2.85 GHz, and four Cortex-A510 energy-efficient cores. Also, the SoC is compatible with LPDDR5X-8533 memory.

By contrast, MediaTek's top-of-the-range Kompanio 1380 SoC for higher-end Chromebooks features four standard high-performance Arm Cortex-A78 cores at 3.0 GHz, four standard energy-efficient Arm Cortex-A55 cores, 5-cluster Arm Mali-A57 graphics, and an LPDDR4X-2133 memory subsystem. However, the Kompanio 1380 is less capable than the Dimensity 9200, so MediaTek wants to enhance its PC SoCs before addressing Windows on Arm machines with them.

It is unclear whether MediaTek plans to use performance-enhanced Arm Cortex-X cores for its notebook SoCs or will develop its custom Arm-compatible cores like Apple. On the one hand, it is entirely logical for MediaTek to build custom high-performance cores if MediaTek plans to compete against Qualcomm's next-generation Snapdragon SoCs that will use custom cores from Nuvia. But on the other hand, this requires significantly more investments and effort than licensing high-performance out-of-box cores.

MediaTek has been gradually transforming itself from a developer of mainstream SoCs for consumer electronics and handsets to a supplier of premium application processors for advanced smartphones. As a result, it will be logical for MediaTek to start developing custom performance-enhanced IP in-house to differentiate from its rivals (most notably Qualcomm, Samsung, and Unisoc) and offer unique capabilities. However, it is unclear whether the company has such plans for now.

TSMC (Taiwan Semiconductor Manufacturing Co.) is set to start production using its N3 (3nm-class) fabrication process shortly, the company said at an event on Tuesday. Formally, the company does not disclose customers who plan to use its N3 manufacturing technology first, but unofficial sources indicate that Apple will be TSMC's alpha customer for the production node.

TSMC intends to kick off high volume manufacturing of chips on its N3 fabrication technology 'soon,' according to C.C. Wei, chief executive of the world's largest contract maker of semiconductors, reports Nikkei. Previously, unofficial sources indicated that TSMC intended to start producing chips on its N3 production node in September, but this is the first time the company has confirmed that it has not started to make products on its most advanced node yet and that it plans to start high-volume manufacturing (HVM) of 3nm chips 'shortly.'

When compared to TSMC's N5 fabrication process that entered HVM in 2020, the foundry's first generation N3 fabrication process promises to deliver a 10% to 15% performance improvement (at the same power and complexity), reduce power consumption by 25% – 30% (at the same speed and transistor count), and increase logic density by around 1.6 times.

Eventually, TSMC plans to add more nodes to the N3 family. N3E will deliver an improved process window, N3P targets enhanced performance, N3S will offer increased transistor density, and N3X will have further performance optimizations for applications like CPUs. As you might expect with the slowing down of Moore's Law, the 3nm family of nodes will be used for many years to come.

One of the key features of N3 is TSMC's FinFlex technology that greatly enhanced design flexibility for fabless chip developers. The tech allows designers to mix and match different kinds of standard cells within one block to accurately optimize performance, power consumption, and area, which will be especially advantageous for complex devices like CPU or GPU cores that have to offer both high transistor density and high clocks. As a result, companies who build complex system-on-chips — such as Apple, AMD, Intel, MediaTek, and Nvidia — will benefit from FinFlex.

Apple, which is TSMC's largest customer, is expected to be the first company to adopt the foundry's N3 technology, though it is unclear what exactly the company plans to produce using this node. Later on, the technology will be used by Intel as well as other clients of TSMC, including AMD, Nvidia, MediaTek, and Qualcomm.

As AMD is ramping up production of its next-generation CPUs and GPUs, it is set to become the second largest client of TSMC's N5 (5nm-class) production node platform in the second half of 2022, according to a DigiTimes report. But will it become the foundry's second largest customer overall? If so, it would be behind Apple, but ahead of MediaTek. 

AMD is set to launch a multiple new products made using various nodes that belong to TSMC's N5 family. First up, next month the company will release its long-awaited Ryzen 7000-series processors for desktops based on its Zen 4 microarchitecture. Later this year, the firm will start ramping production of its 4th Generation EPYC processors codenamed Genoa featuring the same Zen 4 microarchitecture. Also, later this year AMD plans to release its Radeon RX 7000-series graphics processors featuring its RDNA 3 architecture. While ramp up of datacenter-oriented EPYC will be fairly slow, client-oriented Ryzen and Radeon should ramp relatively quickly, so AMD will have several high-volume N5 chips in production by the end of the year.  

Going into 2023, AMD is set to launch its EPYC processors codenamed Bergamo featuring Zen 4C cores for cloud native workloads in Q1 as well as EPYC CPUs codenamed Siena for telecommunications equipment later on. Both are set to be made using one of TSMC's N5 process technologies. In addition, AMD is projected to introduce its Zen 4-based Ryzen 7000 codenamed Phoenix processor for mainstream and high-performance laptops in the second half of next year. 

To sum up, AMD is going to order a boatload of chips for various applications (desktops, notebooks, high-performance servers, cloud servers, telco servers) made on multiple TSMC's N5 process technologies in the next six to nine months. As a result, it may well become TSMC's second largest N5 customer if not in late 2022, but in the first half of 2023. 

Readers who follow AMD closely may ask whether AMD becoming TSMC's second largest customer on N5 platform could also mean that the company is poised to become the foundry's second largest customer in general.  

Based on what we know about TSMC's largest customers, AMD was somewhat behind MediaTek last December, but the gap between the companies has shrunk since then, partially due to AMD acquiring Xilinx. MediaTek tends sell boatload of system-on-chips for consumer electronics, smartphones, and tablets as well as to adopt TSMC's latest nodes more aggressively than the CPU and GPU company, which is why its spending on production is high. But MediaTek's SoCs have a very long lifespan, so they are cheap to make.  

Meanwhile, AMD's N5-based product family will be very broad in terms of applications, so the volumes that the company will need will also be huge. Since TSMC charges more for N5 chips than it does for N7 products, AMD's next-generation N5 chips will be more expensive to make than its existing devices, so the company's spending at TSMC will also raise. Will it be enough to leave MediaTek behind? We do not know for sure, but AMD clearly has chances to do so. 

Being TSMC's second largest customer has its benefits as it strengthens AMD's position when negotiating allocation, prices or talking about process technology customizations. Allocation is very important for companies like AMD in general and now that it has so many product lines that benefit from leading-edge nodes, it is crucial for AMD to strengthen its relationship with TSMC. Being the second largest customer is certainly a good way of strengthening ties with any foundry.

Intel and MediaTek announced a strategic partnership today that will see the fledgling Intel Foundry Services (IFS) producing chips for Mediatek, the fourth largest chip designer in 2021, for a range of smart edge devices. Intel will fab the chips on its 'Intel 16' node, a revamped version of the 22FFL node (a legacy process optimized for low-power devices). The announcement comes as the US-based semiconductor industry, and Intel in particular, is on the cusp of receiving significant subsidies from the government to increase US-based chip fabrication.

MediaTek currently uses TSMC for most of its foundry services, but it is also looking to diversify its supply chain by adding production capacity in both the US and Europe. Intel's IFS fits the bill with facilities in both regions, and Intel says that it anticipates a long-term partnership that could span multiple technologies and applications. Intel declined to comment on the timeline for shipping MediaTek products, but says the 'Intel 16' node will be available for tapeouts (first revisions of the silicon) for its customers in 2022, and then available for initial volume ramps in early 2023.

MediaTek currently creates more than two billion devices per year, but it isn't clear how many of those will soon come from Intel's foundries. Intel also isn't specifying the ratio of MediaTek's production that will occur in either the US or Europe, telling Tom's Hardware, "We can’t comment on the details of customer products. IFS customers have access to corridors of capacity across Intel’s global factory network, including existing wafer fabrication sites in Oregon, Arizona, Ireland and Israel, as well as plans for new greenfield fabs in Ohio and Germany."

The overwhelming majority of processors made worldwide are based on older legacy nodes, not the bleeding-edge tech Intel has coming as it looks to execute its process node roadmap that promises five new nodes in four years.

MediaTek's planned smart edge devices mesh nicely with the 'Intel 16' process, an improved version of the company's mature 22FFL node that began shipping in 2018. The 22FFL (FinFET Low Power) process is optimized for low-cost and low-power chips that still deliver high performance while also offering design simplicity for a faster time to market.

For the 'Intel 16' node, Intel has modernized the 22FFL tech further and added support for third-party chip design tools, which stands in contrast to Intel's proprietary design tools that it uses internally. For IFS, supporting third-party Electronic Design Automation (EDA) software for chip design is a critical step forward if it plans to woo chip designers to its production services.

"This is a chance for IFS to build out a real foundry business. There's likely going to be some growing pains along the way, so IFS needs a customer that will be willing to work with it,' Kevin Krewell of Tirias Research told Tom's Hardware.

Intel's decision to plow an initial $20 billion of its own money into the Intel Foundry Services (IFS) wing comes as the company looks to reverse years of decline, in part by offering its manufacturing services to chip designers like MediaTek. IFS has some momentum — it has already signed Qualcomm and Amazon Web Services (AWS) as initial customers and won a contract from the US Department of Defense. It's also generated interest from other industry heavyweights, like Nvidia.

But the first wave of customers alone won't build a thriving third-party foundry, so Intel has been heavily investing in building out its initiative. Intel slapped down $5.4 billion to acquire existing third-party fab Tower Semiconductor, a specialist in high-volume trailing-edge node production with a vast portfolio of clients, and has recruited seasoned leaders like Suk Lee from TSMC to expand its design technology ecosystem. The company is also expanding its horizons by plowing $1 billion into the RISC-V ecosystem, pledging to build Arm chips if needed, and licensing its own x86 IP for its customers to build their custom designs. 

Adding the MediaTek partnership to the list is another important achievement as Intel adjusts to the foundry business model. MediaTek currently partners with TSMC to produce most of its chips. Still, it doesn't seem likely that the initial Intel partnership will steal much of TSMC's business, and the two companies aren't disclosing any financial information with today's announcement. However, Intel now has its foot in the door with yet another large chip designer, leaving room for further opportunity in the future. 

Apple's transition to its own custom Arm-based system-on-chips enabled the company to increase its systems' sales and made it a dominant supplier of PC-grade Arm SoCs. As for Arm, its instruction set architecture now commands nearly 9% of the global PC market. 

"Apple established itself as a distant market leader in Arm-based notebook PC processors with almost 90% revenue share [in 2021]*," wrote Sravan Kundojjala, Director of Handset Component Technologies service at Strategy Analytics.  

Companies like Acer, Dell, and HP have been shipping Chromebooks and Windows-on-Snapdragon always-connected PCs (ACPC) based on various Arm-powered SoCs for years. However, these PCs have never been truly popular due to mediocre performance and/or uncompetitive pricing. With its M1-based iMacs, MacBooks, and Mac Minis introduced in 2020 and throughout 2021, Apple not only managed to offer competitive performance and appealing design, but it also priced those systems very competitively (e.g., below previous-generation systems featuring Intel's CPUs). So, it attracted sales both from its loyal customers and from new clients. 

Apple outmaneuvered Arm-based Chromebooks and ACPCs in terms of sales and in revenue since its systems are still premium machines priced well above average Chromebooks.  

Since Qualcomm's Nuvia-based SoCs will not launch until late 2023, Apple will continue to offer the fastest Arm-powered SoCs for PCs and will likely lead the market in Arm desktops and laptops for quite a while. 

"Apple's M-series family of processors set the benchmark and gave Apple a 2–3-year lead over the rest of the Arm-based PC processor vendors. Qualcomm captured just 3% revenue share in the Arm-based notebook PC processor market in 2021 and lags Apple in CPU performance," wrote Kundojjala. "Despite its low share, Qualcomm continues to invest in notebook PC processors with its Nuvia CPU cores. We believe that Arm-based notebook PC processor offers an attractive opportunity to Qualcomm, given the company's growing collection of high-performance processor assets including CPU, GPU, AI, audio, imaging, connectivity, gaming and security." 

The industry shipped around 348.8 million PCs in 2021 and 80.5 million systems in Q1 2022, according to IDC. Sales of Chromebooks totaled 37 million units in 2021 as well as 5.1 million systems in Q1 2022. In the first quarter, Apple shipped 7.2 million Macs and had a market share of 8.9%.  

Since the vast majority of Apple's PCs solid in Q1 this year were powered by its own Arm-enabled SoC, it is clear that Arm commands a sizeable share of the PC market due to Apple's M-series SoCs alone. Meanwhile, there are also several popular Chromebooks based on Qualcomm's Snapdragon as well as MediaTek SoCs. While those systems are not as popular as Apple's MacBook Air or MacBook Pro laptops, it is safe to say that Arm's share in PCs is at least 10%, a significant achievement for the British CPU designer. 

It should be noted that for now, Arm-based SoCs power mainstream and even entry-level workstation machines, but they still cannot compete against high-end x86-powered desktop PCs, especially in the field of gaming. Therefore, Arm, Apple, and Qualcomm still have a lot of work to do in a bid to successfully compete against AMD and Intel across all fields.  

*Note: Since Strategy Analytics does not disclose how it estimates revenue that Apple gets for its notebook processors, our story is focused on volume sales of Apple Macintosh systems in 2021 (27.775 million) and in Q1 2022 (7.2 million) compared to shipments of Chromebooks in 2021 (37 million) and in Q1 2022 (5.1 million). We also note that the vast majority of Chrome OS-powered machines use x86 processors from AMD or Intel.

In the first quarter of 2022, the Top 10 fabless chip designers increased their cumulative sales to $39.43 billion, or by a whopping 44% year-over-year, according to TrendForce. Qualcomm and Nvidia continued to remain on top of the list, but AMD — now in the fourth place — is slowly but surely increasing its sales and has all chances to become the third largest fabless chip designer in the coming quarters.

Being the world's leading supplier of smartphone SoCs and RF modules, Qualcomm enjoyed the natural growth of these businesses in Q1 2022. In addition, the company experienced growth in its automotive and IoT sales, so Qualcomm's earnings for the first quarter totaled $9.548 billion (excluding its licensing business), up 52% from the same quarter a year ago.

Nvidia continued to benefit from the growing demand for discrete graphics processing units for client PCs and datacenter GPUs for AI and high-performance computing applications. Hence, its revenue for Q1 reached $7.904 billion, up 53% annually. By contrast, Broadcom was not that lucky, so its sales 'only' increased by 26% YoY to $6.11 billion.

AMD replaced MediaTek from the No.4 spot in Q1 mainly because it added Xilinx sales to its revenues. Still, AMD's CPU and GPU sales significantly increased in the first quarter due to strong demand for high-performance processors and graphics cards. As a result, AMD's earnings exceeded $5.887 billion in Q1 2022, up 71% from the same quarter last year. Once the company adds Pensando to its results and ramps up production of CPUs, GPUs, and console SoCs due to seasonality later this year, it will probably outpace Broadcom in terms of earnings. It will become one of the world's Top 3 fabless designers of chips, a landmark event for a company that almost went bankrupt about half of a decade ago. Of course, it remains to be seen how smooth integration of Pensando and Xilinx will proceed, but for now, the sky is blue for AMD. 

While MediaTek switched its place with AMD, its Q1 revenue grew a healthy 32% year-over-year to $5.007 billion. The company continues to address all kinds of smartphones and tablets with its mobile SoCs. As its portfolio expands and the performance of application processors increases, the company can charge more significant premiums for its products. 

Marvell was another company in Q1 2022 to demonstrate an over 70% revenue growth year-over-year (to $1.412 billion) mainly due to its acquisition of Innovium, a cloud and edge data center networking solutions designer, in October 2021. Meanwhile, Marvell was not the only company to improve its business results by adopting an M&A strategy. For example, Cirrus Logic (No.10) took over Lion Semiconductor to boost its mixed-signal business in mid-2021, so its Q1 2022 sales reached $490 million, up 67% year-over-year. 

Demand for different chips continues to remain strong, particularly in 5G, automotive, AI, high-performance computing, and edge computing spaces. But the geopolitical and macroeconomic situation may pose challenges even for the seemingly resilient semiconductor market. As a result, it will be fascinating to watch how the leading chip companies will address these challenges in the coming quarters.

A U.S. district judge dismissed a proposed class-action lawsuit against Apple that accused it of concealing Meltdown and Spectre security flaws and reducing performance of its devices by mitigating the vulnerabilities, which reduced value of their devices. The plaintiffs demanded a compensation to themselves as well as other owners of Apple devices affected by Apple's actions. 

Accusers stated that Apple misrepresented the privacy and security of iPhones and iPads as well as performance of its custom SoCs on which Apple argued that plaintiffs' alleged misrepresentations are too general to be actionable and they have failed to allege that any specific representation was false. 

"Plaintiffs have failed to allege an affirmative misrepresentation, an actionable omission, and actual reliance on misstatements by Apple," U.S. District Judge Edward Davila in San Jose, California, wrote in his ruling, reports Reuters. 

Meltdown and Spectre security vulnerabilities were discovered in mid-2017 and then made public in early 2018. While Meltdown affected chips from Intel, Spectre hit virtually all modern high-performance processors with speculative and out-of-order execution. Affected chips included those from AMD and Intel as well as custom system-on-chips based on various Arm architectures, including those from Apple, Qualcomm, MediaTek and Samsung. 

Mitigation of Meltdown and Spectre leads to degraded performance. Since Apple knew about vulnerabilities before they were made public and mitigated them before they were revealed, plaintiffs alleged that it deceived its customers by advertising security advantages of its iOS platform as well as performance advantages of its then latest system-on-chips compared to its previous generation SoCs. The accusers called Apple's actions fraudulent.  

"Plaintiffs allege that Apple was notified of the defects in June 2017 but did not publicly disclose them until January 4, 2018, after a New York Times article leaked the vulnerabilities," the complaint reads. "Apple addressed speculative execution and Meltdown, disclosing that its December 2, 2017, iOS 11.2 update included a software update to address the vulnerability. On January 8, 2018, Apple separately released iOS 11.2.2, a software update to address Spectre." 

The plaintiffs stated that if they had known about potential security vulnerabilities of Apple's hardware and software, they would not buy the company's products or would not pay the prices they did. Furthermore, after Apple made appropriate announcements, their devices declined in value. 

"Plaintiffs assert that these vulnerabilities are material because, 'had they known data stored on their systems would be compromised and made available to unauthorized third parties,' they would not have purchased their iDevices or paid the price they did," the lawsuit reads. "After Apple made the announcements, the iDevices allegedly declined in value." 

The court sided with Apple as all security and performance-related claims were correct at the time they were made.  

"Because plaintiffs are unable to identify any statement from Apple that is both sufficiently specific to be actionable and was false when made, plaintiffs have failed to state a claim for fraud under an affirmative misrepresentation theory," the ruling reads. 

Furthermore, since Apple's mitigations affected all its SoCs, performance comparisons between subsequent generations continued to be accurate even after mitigations were implemented as generation-to-generation performance advantages are not only conditioned by improved speculative or out-of-order execution advantages.

While the class-action lawsuit looks to be blown out of proportion, the ruling is an example of a case that exonerates companies from liability when they make certain claims about performance before other information comes to light that may change it.

The organizers have shared the schedule for the Hot Chip Symposium 2022. Intel has booked a lot of presentation time with experts sharing information and insight about its latest GPUs, CPUs, foundry technologies, etc. Other key PC technology firms at Hot Chips 34 include AMD, Arm, and Nvidia. Well-known tech industry pioneers like Samsung, Tesla, MediaTek, and Cerebras will be there. Due to the ongoing pandemic situation, Hot Chips 34 will be a virtual conference that will take place live over the days specified. Still, all presentations will be recorded and made available after processing.

Intel's Hot Chips 34 Schedule

• Intel’s Ponte Vecchio GPU: Architecture, System and Software. Monday, August 22, 9 – 11am
• Heterogenous Integration Enables FPGA Based Hardware Acceleration for RF Applications. Monday, August 22, 11.30am – 1.30pm
• Semiconductors Run the World. Monday, August 22, 2.30 – 3.30pm
• Meteorlake and Arrowlake : Intel Next Gen 3D Client Architecture Platform with Foveros. Tuesday, August 23, 5 – 7pm
• Next-Generation Intel processor build for the edge - Intel Xeon D 2700 & 1700. Tuesday, August 23, 5 – 7pm

Our readers' most interesting Intel presentation will concern Meteor Lake and Arrow Lake technologies on Tuesday. Intel's Wilfred Gomes will host this presentation.

Meteor Lake first successfully booted up in Windows, Chrome, and Linux last month, and it is due to begin shipping to customers in 2023 as 14th Gen Core processors. It looks like Meteor Lake will be a mobile-first architecture, and the presentation will take place in a segment of Hot Chips devoted to mobile and edge processors. Meteor Lake chips will scale from 5W on the mobile side up to 125W for desktops.

To remind readers of the importance of Meteor Lake for Intel, it will be a platform that debuts several essential technologies. Meteor Lake utilizes the Intel 4 process node. We already know it will feature a flexible tiled architecture with chiplets and hybrid cores and a next gen integrated GPU and AI acceleration (a trend set by Arm mobile processors).

The Hot Chips 34 presentation descriptions confirm that Meteor Lake will feature a "Next Gen 3D Client Architecture Platform with Foveros," or a new generation 3D Foveros implementation, with doubled connection density compared to the previous gen. As far as we know, right now, Arrow Lake, or the 15th Gen Core processors, will carry through most of the features of Meteor Lake with a die shrink to Intel 18A. Meteor Lake and Arrow Lake might share a common platform like we have seen confirmed that Alder Lake and Raptor Lake do.

AMD, Nvidia, and Tesla

AMD will also have multiple presentations at Hot Chips. It will discuss its latest Instinct GPU accelerators for HPC on Monday. On Tuesday, it has a presentation about an SoC targeting networking hardware. Lastly, and probably the most interesting to us, will be a late Tuesday presentation about AMD Ryzen 6000 series processors (mobile). We saw this Ryzen APU family first unwrapped at CES, and they have only recently started to feature in laptop designs that are shipping. Since this presentation is part of the mobile segment, it could be that there are some more ULV models on the way. The current H-series are 35 to 45W+, and the U-series are 15-28W. Could we see some sub-15W APUs between now and the end of August?

If we had to pick a highlight from Nvidia's upcoming Hot Chips presentations, we would probably jump with both feet onto the Nvidia Grace CPU talk. We reported the 144-core Nvidia Grace CPU in quite some depth last month. However, Nvidia has promised more architectural insight later, and the Hot Chips 34 presentation looks like the right place at the right time.

The Tesla Dojo features two separate presentations at Hot Chips 34. Tesla decided to split the presentations regarding this exascale computer into an architectural segment, and a super compute scaling ML training segment, led by respective experts from the company. In our most recent report on the Tesla Dojo, we talked about this 7nm and 50 billion transistor chip's custom ASIC-for-AI design and performance potential. We are looking forward to an official update on the hardware, how it will be implemented, and more.

MaxLinear, a designer of connectivity and infrastructure solutions for broadband, connectivity, and infrastructure applications, on Thursday said that it had reached a definitive agreement to acquire Silicon Motion (SMI), a developer of SSD controllers and turn-key solid-state drives. The transaction greatly enhances MaxLinear's product portfolio with a line-up of competitive SSD controllers and a successful team of engineers. 

"Combining Silicon Motion with MaxLinear creates significant economies of scale, accelerates our expansion into enterprise storage markets and unites unparalleled intellectual property to continue serving our customers with high-quality expertise and technical support," said Wallace Kou, President and CEO of Silicon Motion. 

In fact, the merger is highly complementary as most designers of networking and infrastructure chips do not have storage controllers, whereas developers of SSD controllers do not sell networking or infrastructure ICs. MaxLinear believes that the takeover expands its total addressable market to around $15 billion. 

"The enhanced scale of the combined organization creates a new significant $2B+ player in the semiconductor industry with compelling positions across a diversified set of end-markets," said Kishore Seendripu, Ph.D., Chairman and CEO of MaxLinear.

It is noteworthy that MaxLinear was not the only company which considered buying Silicon Motion. Earlier today Bloomberg reported citing people familiar with the matter that MediaTek, a developer of system-on-chips for smartphones, tablets, PCs, and consumer electronics, was also interested in SMI. 

MaxLinear will acquire Silicon Motion in a cash and stock transaction. Under the terms of the agreement, each owner of American Depositary Share (ADS) of Silicon Motion will get $93.54 in cash and 0.388 shares of MaxLinear common stock (valued at $44.36, based on $114.34 per MXL share as of May 4, 2022 closing price), which represents a premium of 48% over Silicon Motion's market value as of April 22, 2022. In total, MaxLinear values Silicon Motion at $3.8 billion. It estimates that the combined company will feature an enterprise value of $8 billion. The transaction is expected to close in 18 months.  

The $3.1 billion of cash consideration will be funded with cash on hand from the combined companies and fully committed debt financing from Wells Fargo Bank.

MaxLinear develops and sells chips used for home and enterprise Ethernet networks, datacenter, broadband, home, and industrial applications. The acquisition of Silicon Motion greatly expands MaxLinear's competitive advantages over datacenter rivals in general and makes it significantly more competitive with Marvell in particular. The latter is renowned for its client and enterprise -oriented HDD as well as SSD controllers and while SMI is competitive in the client space, it is only starting its journey on the enterprise market. 

Solid-state storage is one of the key ingredients of next-generation datacenters and personal electronics. Developing a proper SSD controller for future types of 3D NAND (QLC, PLC, HLC, 600+ layers, high-speed interfaces, etc.) is a major challenge, so acquiring an SSD developer (whether an established player or a start-up) is one of the ways to ensure that your platforms are competitive several years down the road. Meanwhile, taking over a reputable designer of SSD controllers is a relatively easy way of getting a steady revenue stream and enhance product portfolio with offerings that are in high demand. 

Taiwan Semiconductor Manufacturing Corp. is set to start high volume production of chips using its N2 (2 nm-class) process technology in late 2025 and deliver the first batch of these chips in early 2026. According to two media reports and sources within the financial community, the first customers to adopt N2 will be Apple and Intel.

Apple and Intel First to Use TSMC's N2

Apple has been TSMC's largest customer by revenue contribution for about a decade, so it is not surprising that it will also be an alpha customer for N2. As for Intel, the company intends to use TSMC's services to make graphics processing units (GPUs) and various SoCs, two types of applications that benefit from leading-edge nodes. So, it is not surprising that Intel will also be one of the early adopters of N2, according to reports by DigiTimes and UDN. Furthermore, given Intel's volumes, it will quickly become one of the foundry's primary customers. 

Since the first batch of N2 chips should be delivered in early 2026, it is unclear which of Apple's system-on-chips (SoCs) will use it then. Meanwhile, analysts from China Renaissance Securities speculate that Intel will use TSMC's N2 for the graphics tile in its codenamed Lunar Lake processor.  

"We also see more clarity around TSMC's N2 expansion schedule in Fab 20 (Hsinchu)," Sze Ho Ng, an analyst with China Renaissance Securities, wrote in a note for clients. "Tool move-in is expected to start by end-2022, based on company plans, ahead of risk production in late 2024E with Intel (client PC Lunar Lake's graphic 'tiles', while the CPU 'tiles' are fabbed using Intel’s 18A) and Apple being the anchor customers for dedicated capacity support." 

The usage scenario of N2 for Lunar Lake is speculation at this point. However, Intel's own slide that describes graphics tiles of Meteor Lake, Arrow Lake, and Lunar Lake processors clearly indicates that the latter's GPU will be made externally using a technology that is more advanced than N3.

AMD, Nvidia, MediaTek to Use N3, N2

AMD, Broadcom, Nvidia, and MediaTek have formally confirmed that they will use various nodes from TSMC's N5 family (N5, N5P, N4, N4P, N4X). MediaTek has already formally introduced its N5-based Dimensity 8000/8100 application processors and N4-based Dimensity 9000 SoC, whereas Nvidia will use a custom 4N fabrication process for its Hopper and presumably Ada Lovelace GPUs. AMD's Genoa and Raphael processors will also be made using a 5 nm technology. 

According to the DigiTimes report, all of these companies are currently in talks with TSMC over allocation for N3-capable capacities starting late 2023 or sometime in 2024. In addition, these companies are also expected to start talks regarding N2-capable allocations next year, though they will certainly adopt N2 significantly later than Apple and Intel.

N2: Coming in 2026

TSMC's N2 will be the foundry's first technology to adopt gate-all-around field-effect transistors (GAAFET), years after Samsung's 3GAE (2023) and over a year after Intel 20A (2024). So far, the world's largest contract maker of chips has not disclosed what to expect from N2 in terms of power, performance, and area/transistor density improvements over N3. However, considering the fact that this will be a brand-new node, it is reasonable to expect tangible advantages over its predecessors. The new fabrication process will continue to rely on proven extreme ultraviolet (EUV) lithography scanners with a 0.33 numerical aperture. By contrast, Intel's 18A is set to use innovative ASML's Twinscan EXE EUV scanners with High-NA (0.55NA).

For the last few years, TSMC's N5 nodes have been used almost exclusively by Apple for its system-on-chips aimed at smartphones and PCs. But as more companies adopt these fabrication technologies, TSMC has had to increase its production capacities. A new report says that TSMC will increase its N5 production capacity by around 25% this year to meet the demand for N5 chips from the likes of AMD, Nvidia, and MediaTek.

TSMC's N5 (5nm-class) family of manufacturing processes includes vanilla N5, performance-enhanced N5P, N4, N4P, N4X, and Nvidia-specific 4N. Apple is believed to use N5 and N5P for its existing A14, M1, and A15 system-on-chips, but companies like AMD, MediaTek, and Nvidia, are set to use various technologies from the lineup. Meanwhile, Apple's next-generation A16 is also projected to migrate to N4.

For example, Nvidia has tapped 4N for its Hopper compute GPUs (and perhaps for Ada Lovelace consumer GPUs), whereas MediaTek uses N5 for its Dimensity 8000/8100 and will use N4 for Dimensity 9000. 

With N5, TSMC has the capacity for up to 120,000 wafer starts per month (WSPM), according to a report by DigiTimes. 120,000 N5 WSPM is what TSMC planned to achieve by early 2022, so the foundry has the exact capacity it planned to have. However, to serve its existing and future customers interested in one of N5 processes, TSMC will install some additional equipment to increase N5 output to 150,000 WSPM by Q3 2022, DigiTimes claims. Indeed, TSMC will need more N5-capable tools at its fabs by mid-2022.

Nvidia plans to start shipping its Hopper compute GPUs commercially in Q3, so given the length of modern cycles, we're pretty sure that TSMC is already ramping H100 production using the Nvidia-tailored N4 node. While the production volumes dedicated to these GPUs isn't very high, the chips are very large, meaning they'll eat a significant share of TSMC's N5-capable capacity. 

Meanwhile, Apple traditionally ramps up its new iPhone SoCs in April or May, so expect TSMC to kick off production of A16 in the coming weeks. Apple's smartphone SoCs are used for hundreds of millions of devices, so Apple will remain TSMC's biggest customer both in terms of revenues and in terms of processed wafers. Also, since MediaTek sells a boatload of advanced SoCs these days, it will need tens of millions of its Dimensity 8000/8100/9000 application processors and will therefore remain TSMC's No. 2 customer. 

We also fully expect AMD to introduce its next-generation Zen 4-based Epyc and Ryzen CPUs along with RDNA 3-based Radeon RX 7000-series GPUs this fall. Nvidia will also launch its Ada Lovelace-powered GeForce RTX 40-series consumer offerings around the same time as well. All of these products are meant to be widely available, so TSMC will need a lot of capacity to produce these chips. 

So far, both AMD and Nvidia have spent billions of dollars on securing production capacity at TSMC to ensure that they can get all the chips they need. As a result, all the capacity that TSMC will add has essentially been paid for already, months before it goes online.

Meanwhile, TSMC plans to kick off mass production of chips using its next-generation N3 node (3nm-class) sometime in the middle of the year. Apple will be the first to adopt this node, and Intel is expected to follow. The first N3 products will be shipped in 2023, so we won't see them in 2022. 

According to a report from Nikkei, debt-stricken Tsinghua Unigroup will scrap plans to build two new large 3D NAND and DRAM fabs and reshuffle Unisoc as part of its restructuring plan. Both moves represent major setbacks for China's 'Made in China 2025' semiconductor self-reliance plan.  

Tsinghua Unigroup will axe two major semiconductor fab projects in China as part of the ongoing restructuring procedures, a serious blow for the country's semiconductor self-sufficiency goal as local memory production is a key part of the plan. The move will impact the global supply of 3D NAND and DRAM and thus affect memory prices two or three years from now. In addition, Tsinghua will reorganize Unisoc, the country's largest developer of a mobile system-on-chips (SoCs), which might be good news for MediaTek and Qualcomm but another blow for the self-sufficiency plan.

New Fabs Scrapped

Tsinghua Unigroup already owns Yangtze Memory Technologies Co. (YMTC), a rapidly developing yet still fairly small 3D NAND producer. The company announced intentions to build a 3D NAND and a DRAM fab worth $30 billion in 2017. Still, the initiative has been reconsidered several times (the locations of the fabs were changed, the amount of money to be invested was projected to increased) and has never materialized. After Tsinghua Unigroup faced major financial difficulties last year, the destiny of both projects was completely uncertain, especially amid tensions between China and the U.S. 

Tsinghua Unigroup's second 3D NAND project was essentially meant to replicate YMTC's success. Yet, since YMTC is yet to become a rival for Samsung Semiconductor and SK Hynix — two South Korean companies that produce 3D NAND in China — it did not really make much sense to build yet another fab, which could cost as much as $24 billion, reports Nikkei. 

To build its DRAM operation, Tsinghua hired Yukio Sakamoto, a former chief executive of Elpida Memory who has experience competing against companies like Samsung and Micron. However, Mr. Sakamoto left Tsinghua in the second half of last year after the company faced bankruptcy, even before he could use his knowledge and connections. Since it would have taken years and cost billions of dollars to develop a competitive DRAM process technology amid uncertainties with the supply of fab tools in 2025 ~ 2026, Tsinghua reportedly axed its DRAM intentions, too.

Unisoc to Be Restructured

Unisoc, a smartphone and tablet SoC developer, is yet another Tsinghua Unigroup business unit to be restructured. The company, which consolidates assets of consolidated Spreadtrum and RDA Microelectronics, reportedly controls about 9% of application processors for mobile devices globally. Still, since it mostly sells inexpensive 4G devices, it barely brings in huge profits and cannot compete against SoC designers with 5G-supporting products.  

The current plan includes further consolidation and relocation of existing engineers from multiple sites across China. While this might increase efficiency due to better collaboration, it will also postpone the realization of ongoing projects, possibly including the development of 5G SoCs. 

New Strategic Investors

With many assets under its control and huge plans for expansion, Tsinghua Unigroup owed approximately $31 billion and had $8 billion in cash and cash equivalents as of mid-2020. Yet it had defaulted or cross-defaulted on seven bonds worth $3.6 billion by the end of 2020, according to Reuters. As a result, Hong Kong-listed state-owned Huishang Bank requested the First Intermediate People's Court of Beijing Municipality to start insolvency procedures for the conglomerate in 2021. 

Late last year, two new investors — Beijing Jianguang Asset Management Co. and Wise Road Capital — emerged as new backers of Tsinghua Unigroup. Nikkei believes that the two funds acquired a portion of the stake controlled by Beijing Jiankun Investment Group, a private investment group founded by Zhao Weiguo, the chairman of Tsinghua Unigroup. 

Both Beijing Jianguang Asset Management Co. (also known as JAC Capital) and Wise Road Capital are keen high-tech investors. Over the past several years, they acquired several semiconductor companies, including chipmaker Nexperia, once a subsidiary of NXP Semiconductor, and Chinese chip packaging facilities of ASE Technology Holding. Last year they attempted to buy Magnachip Semiconductor but were denied by the U.S. and South Korean governments. 

With new investors on board, Tsinghua Unigroup will be more financially stable than before, but scrapping two major semiconductor projects emphasizes that the goals of the 'Made in China 2025' project are not going to be achieved. 

Taiwanese semiconductor company and wireless communication specialist Mediatek (via PR Newswire) has, for the first time demonstrated what it’s calling Wi-Fi 7, offering higher speeds and lower latencies than the current king of the wireless frontier, Wi-Fi 6E. 

Mediatek’s demo was designed to show off its Wi-Fi 7 Filogic connectivity SOCs, based on an Arm Cortex-A53 CPU and a Mediatek network accelerator, transferring data at 802.11be’s maximum speed of 30 Gbps using multi-link channel aggregation to allow data to flow seamlessly around network congestion. Wi-Fi 7 is expected to add new technology to all bands while retaining backward compatibility, including the stagnant and cluttered 2.4GHz spectrum, and should deliver speeds 2.4x faster than Wi-Fi 6 using the same number of antennas.

"The rollout of Wi-Fi 7 will mark the first time that Wi-Fi can be a true wireline/Ethernet replacement for super high-bandwidth applications," said Alan Hsu, corporate vice president and general manager of the Intelligent Connectivity business at MediaTek. "MediaTek's Wi-Fi 7 technology will be the backbone of home, office and industrial networks and provide seamless connectivity for everything from multi-player AR/VR applications to cloud gaming and 4K calls to 8K streaming and beyond." 

Based on the 802.11be protocol, Wi-Fi 7 (also snappily known as Wi-Fi Extremely High Throughput) has yet to see an initial draft of its amendments to the Wi-Fi standard, with a final version not due until 2024. It is expected to continue Wi-Fi’s invasion of the 6 GHz spectrum band, and could provide up to 320MHz of bandwidth , improvements to MIMO (up to 16 streams) and multi-band/multi-channel aggregation. What this means is faster, more stable Wi-Fi necessary for demanding applications such as cloud gaming and 8K video streaming.

It may be a week late, but Samsung has now officially announced the outline specs for its premium Exynos 2200 mobile processor. Samsung's news SoC features the AMD RDNA 2-based Samsung Xclipse GPU which Samsung claims offers "Console quality graphics".  But while the chip is in mass production we don't as yet have any details of the device it will make its debut within.

With the launch of the Exynos 2200, Samsung makes a plethora of confident claims. The Korean tech giant's own designed SoCs have sometimes trailed in popularity, with consumers preferring Qualcomm if given a choice, and it clearly hopes to break this trend with the advanced new processor. To try and get a feel for whether Samsung really has broken through with this SoC design, let us look at each major constituent part in turn.

Samsung Xclipse GPU

Starting with the headlining GPU, when this SoC becomes available to third party testers we are sure we shall see some fascinating results. We don't have a tech specs deep dive on the GPU today, as all we have is a rough outline of its purported capabilities.

According to Samsung, its new AMD RDNA 2 architecture GPU positions the graphics capabilities of a mobile phone closer to a console than ever. The Xclipse inherits important technologies from its RDNA 2 brethren on PC, such as hardware accelerated ray tracing (RT) and variable rate shading (VRS).

Marking an "exciting new chapter" in mobile graphics, Samsung asserts the Xclipse GPU will "offer the most immersive graphics and user experiences even on mobile." The new GPU is also claimed to be "power efficient," which may allay some of the pre-launch fears that this SoC had been delayed due to getting far too toasty when the GPU is taxed.

In the background, Samsung's multi-IP governor (AMIGO) will enhance overall performance and efficiency by smartly using GPU acceleration depending on the task at hand.

Octa-core CPU

Samsung says that the Exynos 2200 is one of the first SoCs to make use of Arm's latest Armv9 CPU cores. MediaTek, with its Dimensity 9000, seems to have beaten it for claiming the actual first consumer Armv9 SoC to launch. The Samsung Exynos uses a tri-cluster CPU design with one powerful Arm Cortex-X2 flagship-core, three performance and efficiency balanced Cortex-A710 big-cores and four power-efficient Cortex-A510 little-cores. We don't have deeper specs, such as clock speeds etc, at this time.

Other important Exynos 2200 SoC constituents

Elsewhere in the SoC design, Samsung has included an upgraded dual-core NPU, claimed to offer double the performance of its predecessor in AI performance. Moreover the NPU has much higher precision with FP16 support, in addition to power efficient INT8 and INT16 calculations, says Samsung.

A new ISP enhances imaging tasks and supports the latest image sensors for up to 200MP photography, up to 30fps in 108MP single camera mode, and up to 64+36MP in dual camera mode. It is capable of driving up to seven image sensors, and up to four at a time. Working alongside the NPU the on-device AI can recognize multiple objects, faces, and scenes.

Last but not least the Exynos 2200 SoC includes an onboard modem supporting both the sub-6GHz and mmWave frequencies, as well as 4G LTE. GPS, GLONASS, BeiDou, Galileo satnav systems are supported too. Devices packing the Exynos 2200 will use LPDDR5 memory and UFS 3.1 storage.

Why the delay?

Above we touched upon some recent rumors relating to the delayed launch of the Samsung Exynos 2200. About a week ago, prominent smartphone leaker Ice Universe alleged that this upcoming flagship SoC was being held back as engineers frantically tweaked it. The purported issues were largely due to the GPU generating too much heat at its target frequencies. This official SoC announcement allays fears somewhat, with its confident language, but a Samsung Statement to Business Korea said that this SoC has been delayed to fall in line with a new premium Samsung device.