China's LineShine supercomputer has dethroned El Capitan as the world's number one supercomputer, going straight to the top of the charts after the National Supercomputer Center in Shenzhen (NSCS) submitted its results.

LineShine hit 2.198 FP64 ExaFLOPS in the Linpack benchmark and became the industry's first machine in the Top 500 list to sustain more than 2 ExaFLOPS of double-precision performance using only CPUs. The system is deployed at the National Supercomputing Centre in Shenzhen and was built by the Shenzhen Cloud Computing Center using semi-custom 304-core LX2 processors based on the Armv9 instruction set architecture and running at 1.55 GHz. The machine employs 13.79 million cores in total, uses proprietary LingQi interconnect, and consumes 42.2 MW of power.

From a performance-per-watt point of view, the LineShine machine delivers 52.07 GFLOPS/W, which is below El Capitan's 60.94 GFLOPS/W. However, LineShine by far outperforms Fugaku — another CPU-only supercomputer that used to be the No.1 HPC system several years ago — that can only deliver 14.78 – 16.84 GFLOPS/W depending on whether its efficiency is optimized or not.

LineShine also moved to the top of the HPCG ranking with 22.00 HPCG-PFLOPS. However, the supercomputer achieved 7.92 mixed-precision EFLOPS in HPL-MxP, which puts it behind El Capitan, Frontier, and Aurora. This limits LineShine's usability for AI training and inference, but this can be justified with its exceptional performance for traditional supercomputer tasks.

Each LX2 CPU relies on two compute chiplets and has a total of 304 CPU cores organized into eight CPU clusters containing 38 cores each. Every core includes Arm SVE (Scalable Vector Extension) and SME (Scalable Matrix Extension) units that accelerate vector and matrix operations used in AI training and scientific computing that support FP64, FP32, BF16, FP16, and INT8 data formats. The chip features a rather unusual memory architecture that pairs 32 GB of on-package HBM, offering up to 4 TB/s of bandwidth with as much as 256 GB of external DDR5 memory to maximize both bandwidth and capacity.

Despite this, the processor only gains 3.6X performance when moving from FP64 to mixed-precision data, which is lower compared to systems that integrate low-precision accelerators, such as AMD's Instinct MI300A or Intel's Ponte Vecchio. While an Armv9 CPU with SVE/SME can accelerate FP16/BF16/INT8 workloads, its mixed-precision uplift remains limited compared to systems with accelerators due to many reasons, including memory bandwidth, software maturity, and interconnect efficiency. That said, it may be too early to make final conclusions about the LX2 and its usability for mixed-precision workloads.

In any case, the very fact that a Chinese supercomputer has achieved extraordinary FP64 performance is remarkable. Furthermore, the fact that NSCS has actually submitted results to Top 500 indicates that the organization is confident that the LineShine supercomputer relies exclusively on domestic technologies and the U.S. government cannot affect the production of these technologies.

Servers running x86 processors from AMD and Intel used to rule the market, both unit and money-wise, less than a decade ago, but fast forward to today, Arm-based machines command well over 45% of the server market, according to data released by IDC. While technically x86 machines still control 52% of the market in terms of revenue, the real winner is a different category altogether: GPU- and ASIC/FPGA-accelerated systems, which generated over 70% of the global server revenue in the first quarter of 2026.

Server market reaches $122.6 billion in a single quarter, Dell leads the game

IDC estimates that the global server market generated a record $122.6 billion in revenue in the first quarter of 2026, up 30.4% year-over-year, as spending on AI infrastructure remained particularly strong.

Sales of ODM Direct servers — custom machines ordered by hyperscalers that run merchant or custom silicon — accounted for 50.2% of the revenue (down from 64.1% in Q1 2025) and reached $61.53 billion, up modest 2.1% year-over-year*. By contrast, sales of standard servers from well-known brands grew at a much higher pace, which suggests that branded vendors such as Dell, HPE, Supermicro, and others won a larger portion of AI infrastructure deployments than they did a year earlier. That was probably made possible by accelerating enterprise AI deployment and sovereign AI projects, which tend to buy machines from branded vendors, as well as hyperscalers increasingly turning to well-known suppliers for AI hardware.

When it comes to vendor rankings, Dell remained the largest server supplier by revenue with a 16.5% share of the market after its revenue surged 244.1% year-over-year to $20.3 billion, which was driven by exceptionally strong AI server demand. Supermicro remained in second place with $9.3 billion in revenue and a growth of 128.9%.

Lenovo ranked third with $5.6 billion and 36.5% growth, while IEIT Systems (which is a part of the sanctioned Inspur Group) dropped to fourth after revenue declined 7.0% to $4.0 billion. HPE was No.5 with $3.7 billion in revenue, up 17.2%. Other vendors — from Asus to Atos and from ASRock Rack to Gigabyte — commanded 14.8% of the market with $18.11 billion in revenue, up from 13% and $12.21 billion in the same quarter a year ago.

Arm-based machines rapidly gain revenue share

As AI servers dominated the market in Q1 2026, systems with various types of accelerators accounted for over 70% of the revenue. However, the rise of Arm-powered machines is the elephant in the room that is hard to miss, as it represents a tectonic shift in the whole market, both to the Arm instruction set architecture (ISA) in general and custom-built Arm CPUs designed by hyperscalers.

Non-x86 platforms generated $58.7 billion in revenue, a 107.6% increase year-over-year, which lifted their share of the market to 47.9%. Most of the non-x86 systems are Arm-based AI machines (think Nvidia's NVL72) as well as systems running custom CPUs, AWS, Google, and Microsoft, just to name a few. Still, also keep in mind IBM Z mainframes and IBM Power Systems (including storage) that use CPUs featuring proprietary non-x86 and non-Arm ISAs and which still generate $1 billion or more in revenue. IDC claims that Arm-based machines accounted for more than 95% of non-x86 revenue, so it is safe to say that Arm-based machines commanded over 45% of server revenues in Q1 2026.

One of the reasons why Arm-based machines now command a huge chunk of the server market is because they are used inside such systems as Nvidia's NVL72 'Blackwell' that sell for up to $6.5 million per unit. Each NVL72 rack-scale solution carries 36 compute trays with two Blackwell GPUs and one Grace CPU per unit, so while unit-wise each we are only talking about 36 processors, dollar-wise one NVL72 machine is as expensive as 928 entry-level 1P server (for $7,000) for cloud or edge applications or 433 higher-end 2P servers (for $15,000) for cloud or virtualization applications.

Given the fact that Nvidia will continue bundling its own Arm-based Vera CPUs with NVL72 'Vera Rubin' machines that will be more expensive than their Blackwell ancestors, we will not be surprised that Arm-based machines will account for well over 50% of the server market revenue in the second half of this year or in 2027. Also, keep in mind that Nvidia plans to sell server racks featuring only Vera CPUs for agentic AI applications, which will further drive sales of Arm-based machines.

Accelerated servers: The real winner

Since AI servers dominate server sales, it is not surprising that sales of accelerated servers are increasing. Systems equipped with GPUs produced $68.9 billion in revenue during the quarter (up 24.8% compared to the same period a year earlier) and accounted for 56.2% of all server sales. Servers based on other accelerator types, including custom ASICs and FPGAs, expanded to $17.7 billion, up 122.1% YoY. As a result, accelerated servers earned $86.6 billion in Q1 2026, which is around 70.6% of all server revenue.

X86 servers remain unit volume champions, but suffer from shortages

In contrast, x86 server revenue declined 2.9% to $63.9 billion, though IDC attributes this weakness to supply limitations rather than deteriorating demand. The market research firm claims that the industry's primary constraint is no longer customer appetite for general-purpose servers, but rather the availability of key components, including CPUs, DRAM, NAND memory, and hard drives.

Without any doubt, x86 servers remain working horses for the industry. In fact, many of them use accelerators, including ASICs, FPGAs, and GPUs, as they are used for a wide range of workloads, including AI, supercomputing, simulations, encryption, video transcoding, and many more.

AMD and Intel shipped nearly 20 million EPYC and Xeon SP processors for data center systems in 2025, according to Dean McCarron, the head and principal analyst at Mercury Research. He believes Nvidia is on track to ship four million Grace and Vera CPUs this year, which is considerably lower compared to shipments of AMD and Intel. It is hard to estimate how many custom Arm-based CPUs are deployed by AWS, Alibaba, Google, and Microsoft, but it is safe to say that we are talking millions of CPUs here; otherwise, the companies would not be able to justify development and production of custom silicon.

From a volume perspective, x86 servers remain the most popular machines, and it will probably take some time before ARM can challenge x86 in mainstream general-purpose servers. Nonetheless, it is safe to say that Arm-based data center CPUs are catching up with x86 parts in terms of volumes.

Summary

The global server market hit a record $122.6 billion in the first quarter of 2026 as AI infrastructure spending continued. Accelerated systems powered by GPUs, custom ASICs, and FPGAs generated more than 70% of server revenue, while Arm-based platforms — including Nvidia's Grace Blackwell as well as custom CPUs from Arm, Google, and Microsoft — captured nearly half of the market.

Although x86 servers based on AMD EPYC and Intel Xeon processors remain dominant in shipment volumes, supply shortages of CPUs, memory, and storage components constrained revenue growth, which further enabled Arm-powered AI-optimized systems to gain share. But while at 20 million data center processors per year, x86 volumes are untouchable for Arm today, things may change in the coming years. Nvidia is on track to ship 4 million CPUs in 2026, and other developers of custom Arm-based CPUs are certainly not standing still.

*There is one significant difference with IDC's 'ODM Direct' classification. IDC classifies revenue according to which company invoices the customer, not necessarily who manufactures the hardware. As a result, while many AI servers are built by ODMs like Compal, Foxconn, or Quanta, they are sold under brands like Dell or HPE. As a result, while the latter get more business from enterprises or sovereign AI deployments, this does not mean that big ODMs are losing business; they are actually gaining it, as the appetites of hyperscalers like AWS, Google, Meta, or Microsoft are not going anywhere, just demand from new entrants emerges.

Computex 2026 is moving full steam ahead, like an AI train running down a track made of gold-plated DDR5 DIMMs. We’ve moved into the first full day of the trade show, and the announcements are really starting to fill in. You can catch our Day Zero coverage here and keep track of our dedicated Computex 2026 hub.

Arm PC chips are back in focus

Windows on Arm is not new; the first-generation Surface RT launched way back in 2012 with an Nvidia Tegra 3 processor. Since then, we’ve seen various other takes on Arm processors running on Windows, from the Surface Pro X with its Microsoft SQ1 SoC to a slew of laptops running Qualcomm’s PC-centric Snapdragon processors.

Now, we’re seeing an even more interest in the segment with Nvidia RTX Spark “Superchip” and the new Snapdragon C from Qualcomm. From all accounts, the RTX Spark targets the high end of the PC market with its 20-core Arm CPU, Blackwell RTX GPU, and 128GB of unified memory.

The Snapdragon C takes a different approach, instead aiming at the budget laptop segment. Laptops using Snapdragon C are expected to be priced as low as $300. However, that price point will be highly dependent on memory pricing, which remains a real pain point not only for OEMs, but also consumers looking to get the most bang for their computing buck. In fact, things have gotten so bad that even Intel has said that "something has to give” with memory prices.

• Nvidia's RTX Spark could capitalize where Qualcomm's Arm-based efforts have not
• Jensen Huang says Nvidia wants to 'reinvent the single most important tool of humanity' with RTX Spark
• Qualcomm aims the Snapdragon C laptop chip at the budget laptop segment, as manufacturers feel the DRAM squeeze
• Intel says 'something has to give' with memory prices

We get hands-on time with Asus’ latest hardware at Computex

Asus always has a large presence at Computex, and this year was no exception. The company had an extensive cast of new characters in the laptop field, with new Vivobooks, Zenbooks, Expertbooks, and Strix Scar gaming laptops. Of the new models introduced, the Zenbook 14 with an Intel processor and 14-inch 2880 x 1800 OLED display caught my attention.

Asus also had a rather cool-looking ROG 20th anniversary Harpe II Extreme Edition wireless mouse and the ROG Azoth Extreme Edition 20 mechanical keyboard. The devices are finished primarily in black, but feature 24-karat gold accents. The keyboard also includes transparent switches and keycaps. And we can’t get past its 3.5-pound heft, thanks to its all-metal chassis.

• Asus shows off updated Zenbook and Strix Scar laptops, along with a TUF-based Gaming desktop
• Hands-on with Asus’ ROG Harpe II Extreme Edition 20 gaming mouse
• Hands-on with Asus’ ROG Azoth Extreme Edition 20 mechanical keyboard

PCIe 6.0 SSDs are on the horizon

A year ago at Computex 2025, we saw prototype PCIe 6.0 SSDs capable of 30+ GB/s speeds. This year, we’re slowly inching towards production-capable hardware, with Phison showing off its latest PCIe 6.0 SSD controller: the X3. The X3 is a 16-channel design that is capable of 28 GB/s sequential read/write speeds and 6.8 million random read/write IOPS.

According to Phison, the controller will begin sampling to customers by the end of 2026, with volume production starting in mid-2027.

• Phison shows PCIe 6.0 X3 SSD controller with 28 GB/s of bandwidth and 6.8 million IOPS, supports 2 petabytes per drive

Everything else

We can’t get into detail on everything we’ve seen so far at Computex in this wrap-up — we’ll let the individual news stories speak for themselves. Here’s everything else we’ve covered for Computex 2026 Day One:

• Noctua announces new thermal pad for AMD chips in partnership with Carbice
• Gigabyte debuts fourth-gen Tandem WOLED and multi-mode Mini LED gaming monitors
• Gigabyte showcases new Infinity products for its 40th anniversary
• Cooler Master shows off new HAF 500 chassis, aluminum fans, and new air coolers
• AMD ‘had to re-engineer’ the Ryzen 7 5800X3D for a re-release

Qualcomm announced the Snapdragon C Platform on May 28th, ahead of Computex 2026 in Taipei, an entry-level Arm processor built to anchor Windows 11 laptops priced from $300. The chip abandons the Oryon CPU cores that define the Snapdragon X family in favor of an older Kryo design pulled from Qualcomm's smartphone parts, runs in machines carrying as little as 8GB of memory, and skips Microsoft's Copilot+ certification. We went hands-on with Snapdragon C at Computex 2026 today.

Meanwhile, analysts from TrendForce, Gartner, and IDC are all warning that a surge in memory prices is making the sub-$500 laptop market financially unviable, which could lead to its disappearance in its entirety.

Kryo cores, not Oryon

The "C" stands for Compute, and the new platform sits beneath every Snapdragon X and X2 part Qualcomm sells. Where those chips use the Nuvia-derived Oryon cores Qualcomm acquired in 2021, Snapdragon C reuses Kryo cores from its phone lineup, the architecture Qualcomm built laptop silicon on before the Oryon transition. Mandar Deshpande, senior director of product management at Qualcomm, told reporters at a pre-launch briefing that the platform "is not built to scale up to the Copilot+ requirements," meaning it clears neither Microsoft's 40 TOPS neural-engine floor nor the 16GB memory minimum tied to the Copilot+ PC program.

Qualcomm has disclosed little else; core counts, clock speeds, neural-engine throughput, the manufacturing node, and the supported memory type were all absent from the announcement, with the company saying it would detail them during its Computex keynote this week. Reported leaks point to a 6nm-class part with eight cores, though none of that is confirmed.

The first machine is Acer's Aspire Go 15. Acer's specification sheet lists a 15.6-inch 1920 x 1080 display, up to 8GB of memory, up to 512GB of storage, a 53Wh battery, and Windows 11 with a Copilot key but no Copilot+ branding. Acer hasn’t given the laptop a price or a release date, and HP and Lenovo, both named as launch partners, have yet to unveil their own machines.

Memory prices dictate retail prices

Snapdragon C enters a market where memory has arguably become the deciding factor in what a laptop ultimately retails for. TrendForce projects that conventional DRAM contract prices rose 90% to 95% in the first quarter of 2026 and will climb a further 58% to 63% in the second, with mobile DRAM — the LPDDR type Snapdragon C depends on — rising as much as 93% to 98% quarter over quarter.

Meanwhile, Gartner expects combined DRAM and SSD pricing to increase 130% by the end of 2026, lifting average PC prices 17% and pushing memory from 16% to 23% of a typical laptop's bill of materials. Ranjit Atwal, senior director analyst at Gartner, said in a February forecast that the increases have removed vendors' ability to absorb the cost, and that "the sub-$500 entry-level PC segment will disappear by 2028." IDC reached a similar conclusion, cutting its 2026 global PC shipment forecast to a decline of 11.3% and warning that bargain-priced PCs are, for now, behind us.

The squeeze is already visible further up the Windows line-up. HP told investors that memory now makes up roughly 35% of its PC bill of materials, up from the mid-to-high teens a quarter earlier, and Lenovo told TechRadar there was "no way around" the price increases it would pass to buyers. Microsoft's cheapest Surface Laptop now starts at $1,149, a position that no longer competes for budget buyers at all, and a significant jump from its original $899 launch price.

A new entry-level tier?

Qualcomm isn’t the first manufacturer to repurpose a binned phone processor; Apple’s $599 MacBook Neo, announced in March, runs the A18 Pro from the iPhone 16 Pro alongside 8GB of unified memory and 256GB of storage in a fanless 13-inch body. That’s the same maneuver Qualcomm is attempting by using Kryo silicon in the Snapdragon C.

Apple claims that the Neo runs up to 50% faster than a comparable Intel Core Ultra 5 laptop and three times faster in on-device AI, but that’s all based on its own internal benchmarks against an unnamed machine. Asus co-CEO S.Y. Hsu, on the company's first-quarter earnings call, called Apple's pricing "a shock to the entire industry."

On the Windows side, Intel launched Wildcat Lake as its Core Series 3 family in April, built on the 18A node with a roughly 17 TOPS neural engine and more than 70 laptop designs in the pipeline. Its Project Firefly reference platform targets sub-$600 x86 machines, and Chinese vendor CHUWI has already shown a Wildcat Lake laptop, the “UniBook,” at $449.

AMD's budget option remains the aging Ryzen 7020 "Mendocino" series, a 6nm part pairing Zen 2 cores with RDNA 2 graphics and no neural engine, with no direct successor yet below $500.

At this level, Qualcomm’s most immediate competition may well be its own back catalog. The Snapdragon X-powered Asus Vivobook 14, with 16GB of memory and 512GB of storage, has sold for $379.99 at Best Buy, undercutting the price Snapdragon C is chasing while offering twice the memory and faster Oryon cores.

$300 and up figure is guidance

The $300 figure is guidance Qualcomm shares with manufacturers, not a price it sets, and Acer's refusal to confirm a number for the Aspire Go 15 is arguably an immediate tell that it’s not going to retail anywhere near that price.

With LPDDR5 contract pricing reported above $10 per gigabyte, the memory alone in an 8GB configuration now costs roughly what a whole budget laptop's components once did, and OEMs are passing those costs through. A first wave landing between $349 and $449 in the U.S. is a more realistic price point.

Dropping Copilot+ is, of course, a concession that has made this price possible; buyers lose Recall, Cocreator, and the richer Windows Studio Effects modes, and because Copilot+ has yet to drive measurable Windows upgrade demand, Qualcomm and its partners appear willing to trade the badge for a lower bill of materials.

Whether the platform holds its line depends on the specifications Qualcomm has so far withheld, potentially coming imminently at Computex, and on whether HP and Lenovo ship Snapdragon C machines this year or wait for memory prices to ease.

Some people are pretty excited that we are on the cusp of a new Windows PC era ignited by technologies from the mighty Nvidia and Arm. Our article on the teasers for the rumored wave of new N1X laptops makes it clear that companies like Nvidia and Microsoft are excited, at least. However, Microsoft veteran Steven Sinofsky is here to remind folks that it has been done before, with a similar level of simmering excitement, when the first Surface hybrid PC ran on Nvidia Tegra Arm silicon back in 2010/11.

Sinofsky shares some interesting media in the above embedded Tweet. There’s a slide from the promotional deck, shown at CES 2011, where the “strong partnerships” behind this Windows on Arm thrust would surely lead to unstoppable momentum. That’s what you may have believed if you swallowed the effusive presentations at the time.

This prior push for Windows on Arm didn’t exclusively support Nvidia Tegra. Sinofsky points out that Windows 8 on Arm also supported Qualcomm and TI processors. However, we saw Microsoft switch to a Qualcomm partnership focus in the years following. That collaborative effort also didn’t break through, even when the much-lauded Nuvia Oryon architecture arrived with the Snapdragon X family.

The former-President of the Windows Division (July 2009 to Nov 2012) also shares a video that should be preserved for posterity. Check it out to witness “an old school Windows Phone video of the first time Windows ran on Nvidia Tegra Arm using the desktop compositor which was a BIG deal. Sept 2010.”

The video appears to show a Tegra dev kit running Windows 7, with multi-window and Start menu manipulation, which looks rather lethargic to my eyes. However, that was a “first” achievement several months before CES 2011, and it would be over a year and a half until the Surface RT shipped to customers (October 2012).

Will the latest Nvidia and Arm thrust into Windows work out better? We’ll probably have a much better idea about that in the coming days at Computex. Sinofsky replies to some comments on his Tweet to say the result could be “a reliable platform for graphics compute,” using Windows. Pricing is also obviously going to be a sticky issue for any new platform looking to attract buyers right now.

The creators of the Flipper Zero “portable multi-tool device for geeks” have announced the Flipper One. This new pocketable gadget hugely expands the original's feature set with compute, modularity, and expandability to make what is claimed to be a different category of device. The Flipper One isn’t actually ready yet, though. Instead, the Flipper Devices team is asking for help from the community to help steer and finesse the final stages of Flipper One development to meet their ambitious goals.

“Flipper Zero taught us how much you can do with a tightly scoped, open product and a community that pushes it further than you can,” said Pavel Zhovner, Co-Founder and CEO of Flipper Devices. “Flipper One is what happens when we apply the same approach to a much bigger problem — building a fully open ARM Linux device that doesn't go obsolete the moment it ships. To be honest, it’s hard, and we can't do that alone, which is why we're opening the development process from day one.”

Before we go on, the team wanted to stress that the Flipper One isn’t an upgraded Flipper Zero. They assert that “Flipper Zero and Flipper One are completely different projects built for different tasks.” However, after digesting the announcement material, we’d probably sum up that the Flipper One is a device that adds a very useful chunk of Linux compute to the geek multitool form factor established by the Zero. Whatever the case, the Flipper team has created a side-by-side infographic to compare the two devices, and we’ve embedded that below.

Key to the Flipper One’s expanded abilities are the inclusion of an Arm processor capable of running Linux with about the same performance level as a Raspberry Pi 5, according to the press release, plus the addition of modular M.2 expansion capabilities. These are big additions, and at this stage, the Flipper team openly admits it is still wrangling with getting everything working as intended.

Some important foundational work has been done in preparing Arm Linux for the Flipper One. For example, the team has partnered with Collabora “to push full support for the Rockchip RK3576 SoC into the mainline Linux kernel.” This is a work in progress, though, with current effort focused on power management and USB DP Alt-mode support. Moreover, drivers for the SoC’s NPU, hardware video decoding, and other accelerators aren’t fully upstream yet.

To move forward with the above and related tasks, the Flipper team has created the Flipper One Developer Portal, a public wiki with all the development documentation for Flipper One. Due to the complexity of this new networking and computing multitool, the Wiki houses sub-projects focused on Hardware, Mechanics, Linux software, MCU Firmware, User Interface, Documentation, and Testing. Anyone can join and is welcome to contribute.

It sounds like there’s a lot of work left to do, but there’s a lot of potential in this new computing multitool. Its coprocessor architecture mixes the aforementioned octa-core Rockchip RK3576 SoC, which also packs Mali-G52 graphics, an NPU, and comes with 8GB of RAM. It is partnered by the RP2350 low-power MCU. Importantly, the MCU can work alone, bringing a lot of functionality to Flipper One without even getting into Linux. The CPU and MCU communicate and work together using an interconnect system.

To make the most of the power inside a Flipper One, the OS must be optimized for the task(s). So, the development team is making Flipper OS, based on Debian. A key project in this development is FlipCTL, a framework for interacting with the device on a tiny screen using just a D-pad and a few buttons.

If you have access to a big screen, Flipper One can charge, output video to a monitor, and connect USB peripherals — all via a single USB-C DisplayPort Alt Mode cable. The built-in full-size HDMI port is also envisioned as a big-screen boon, with Flipper One connecting and becoming a “hacker’s TV media box.”

Flipper One’s open hardware module system with a widely compatible M.2 slot and GPIO port also boosts its capabilities beyond the network multitool skills it seems a natural fit for.

Lastly, no new device press release can neglect to mention AI. In this case, the Flipper One is touted as a device with a built-in AI accelerator (Rockchip's integrated NPU, remember) that can run LLMs locally. There are a couple of wrinkles to achieving this functionality right now. However, Flipper One will support external AI agents through integrations when you have internet connectivity.

Arm Holdings, the maker of the popular Arm architecture used by Qualcomm, Apple, and several other companies, is facing an antitrust investigation as the U.S. Federal Trade Commission (FTC) looks into the company’s operation. Bloomberg reports that the FTC is determining if the company is trying to monopolize the Arm architecture and either only give customers lower quality designs for their own semiconductors or outright deny them access to its licenses. The move comes as Arm launched its own AGI CPU focused on data centers in March 2026, a significant departure from the company, whose business previously focused on licensing its chip designs to other companies.

The company’s legal troubles began when it sued Qualcomm, the biggest manufacturer of smartphone chips, for using Nuvia’s ARM licenses after it acquired the startup in 2022. Arm Holdings argued that the smartphone chipmaker cannot use Nuvia’s licenses after its acquisition, and that it should have acquired a new one to continue using the startup’s designs based on Arm licenses. Arm ultimately lost the case, allowing Qualcomm to continue using the Oryon cores it acquired from Nuvia.

The Qualcomm lawsuit broke the longstanding relationship the two companies had, and it also opened a can of worms for Arm. The former launched a global antitrust campaign against the latter because of the case, saying that it was using its dominant market position to prevent competition. It reached out to the European Commission, the U.S. FTC, and Korea’s Fair Trade Commission to present its case, with the Korean government agency raiding Arm’s Seoul office in November 2025.

While the x86 processor still has the advantage when it comes to desktops and laptops, its advantage is slowly being eroded by Arm-based Apple Silicon and Qualcomm’s Snapdragon X-series chips. On the other hand, it has cornered the mobile market, with Apple, Qualcomm, Samsung, MediaTek, and other mobile device chipmakers using Arm architecture. Some analysts also say that it will eventually dominate the AI server industry, with over 90% of custom processors expected to use an Arm chip by 2029.

Arm’s launch of its own AGI processor has got its customers (and now potential competitors as well) wary of its status as both chip designer and manufacturer. Arm did not introduce a processor for consumer computing, meaning it hasn’t directly challenged some of its biggest customers. However, the fact that it has expanded into building physical chips has worried them that it could leverage the popularity of the Arm architecture to gain an unfair advantage and limit competition.

Chipmakers are taking to social media to confirm their partnerships with Google on its newly announced Googlebook laptop lineup.

In a post shared on X, Intel said it is collaborating on the lineup. Meanwhile, over on Instagram, Qualcomm made its own confirmation. Both used similar wording, saying that the laptops will be "powerful" and "premium" "devices designed for Intelligence." (Qualcomm used "built" instead of designed."

The announcements came shortly after Google gave a preview of its upcoming platform at the Android Show: I/O Edition, and confirmed that it is working with various PC manufacturers, including HP, Dell, Acer, Asus, and Lenovo.

During the showcase, Google refrained from discussing the core hardware and instead focused entirely on its brand-new operating system, which combines elements of Android and ChromeOS with deep Gemini Intelligence integration. It was initially assumed that the new Googlebook lineup would be based on Arm SoCs, since many aspects of the platform resemble an Android smartphone or tablet experience. However, with Intel now officially involved, there is a possibility that Google’s new AI-focused OS could also support x86 hardware, unless Intel has an Arm-based chip up its sleeve.

In an exclusive interview with Chrome Unboxed, Google VP John Maletis further confirmed Intel’s involvement in the Googlebook project, revealing that the upcoming notebooks will ship with processors from Intel, Qualcomm, and MediaTek. According to Maletis, the Googlebook is an entirely new category of premium AI-first laptops that deeply integrate Gemini into the core experience rather than treating AI as an add-on. He also noted that Google is establishing strict hardware standards across memory, storage, keyboards, and overall build quality to ensure every Googlebook delivers a consistent premium experience.

The interview also shed more light on what users can expect when Googlebook devices officially launch later this fall. According to Maletis, the first wave of laptops will focus heavily on premium hardware from its partners, while also bringing back the iconic Glow Bar LED lighting seen on older Chromebook Pixel devices. He additionally confirmed that Googlebook laptops will run native Android applications without emulation, promising significantly better app performance alongside tighter Android smartphone integration and Gemini-powered features such as the new Magic Pointer interface.

Interestingly, the Googlebook partnership comes just a month after Intel and Google announced a separate multi-year agreement focused on next-generation AI cloud infrastructure. Under the deal, Google Cloud will deploy Intel Xeon processors alongside custom IPUs for large-scale AI workloads, suggesting that the relationship between the two companies now extends from cloud AI infrastructure all the way down to consumer AI-focused devices.

Updated May 13, 3:18 PM ET with further confirmation from Qualcomm on its partnership with Google

When Arm introduced its first 'physical' AGI processors in late March, the company expressed optimism about their adoption by select customers. In less than two months, the company attained around $2 billion in commitments for its AGI CPU over the next several years, smashing the company's expectations two-fold. But despite this heightened interest, Arm's market share will remain in the low single digits even if it manages to ship $2 billion worth of CPUs in two years, Mercury Research told Tom's Hardware.

"Customer response to the Arm AGI CPU is already strong, with more than $2 billion of customer demand across FYE27 and FYE28 – more than double what was stated at Arm Everywhere," Arm declared in its earnings press release.

Arm officially introduced its AGI CPU on March 24, 2026, and referred to it as 'production silicon,' meaning the design of the processor itself is final. Actual production of the CPU is expected to begin in the second half of 2026, with initial customer shipments expected in Q4 2026. Arm expects to ship $90 to $100 million worth of AGI CPUs in Q4 2026 alone.

Given the rising interest in the new chip, the company expects to generate $15B in AGI CPU sales and $10B in IP revenue by FY 2031 (ending on March 31, 2031), which will drive its total revenue to $25B per year, up from $2.61B in FY 2026.

Generating $15 billion in data center CPU sales in a single year is a big deal; Intel earned $16.8B selling server processors last year, after all. Given the rising demand for CPUs, particularly for agentic AI workloads, Arm's revenue may indeed increase by almost a factor of 10, with actual CPUs accounting for 60% of that total figure.

Single-digit percent of the server market

While $100M worth of AGI CPUs in Q4 2026 and over $2B of demand for the next two fiscal years looks like a lot of money (especially given the fact that Arm's current annual revenue is $2.61B), Arm's presence in the server and data center CPU market (silicon CPUs, not IP) will be negligible (yet still quite hard to achieve) if compared to share of merchant CPUs.

AMD and Intel sold just under 20 million data center-oriented EPYC and Xeon SP processors worth tens of billions of dollars in 2025, according to Dean McCarron, president and principal analyst at Mercury Research, a leading CPU market research firm. If we consider only 2025 data center CPU shipments, Arm would need around 4% unit share of the current server CPU market to achieve its $2 billion revenue target.

"In round numbers for 2025, AMD's EPYC average selling price was about $1,325," Dean McCarron told Tom's Hardware Premium. "For Intel, the 2025 ASP for Xeon SP* is about $1,125. What Arm gets of course might be different, and prices are rising, but something like $1,250 probably is not a bad starting place."

At this point, it is hard to estimate the actual ASP of Arm's AGI since while the company advertises processors with up to 136 cores, we can only wonder how many SKUs there will be and how many cores entry-level models will have. If Arm behaves like a typical CPU maker — balancing recovery of development and manufacturing costs against maximizing margins — then AGI's ASP will be comparable to that of EPYC or Xeon.

"So, $2 billion would take roughly 1.6 million CPUs, if that is done over the course of a couple years — eight quarters — that is an average of 200,000 units per quarter," McCarron explained. "For comparison, in 2025, the combined EPYC and Xeon SP markets averaged just under 5 million units per quarter, and of course, that is going to be growing rapidly in 2026 and beyond. So, Arm's $2 billion in server CPU revenue requires them to sell just 4% of the total units right now, and this would be an even smaller percentage of the total in a couple years."

Since Meta is a co-designer partner and lead customer for Arm's AGI CPU, it might get a considerably lower price, which means that Arm will have to supply more units to meet its revenue target, which will mean a higher market share at the cost of lower profits.

"While those [ASP] figures span entry-level to the largest cores, the volumes (and ASPs) are dominated by the hyperscalers," explained McCarron. "When you buy hundreds of thousands of units at a single time, there are some volume discounts, which is why the ASPs are in the low thousands and not $10,000+."

*Other Intel server products were excluded from the comparison as they are not direct competitors to Arm-based data center CPUs.

But can Arm supply?

Given the widespread shortages of everything from wafers at TSMC to memory and from storage devices to advanced chip packaging capacity, we can only wonder whether Arm can increase its output of its AGI CPUs in the next two years by a factor of two. The company has not given a positive answer straight away, but it claims that it is working on it.

"How quickly can we get units?" Rene Haas asked rhetorically. "The number that we talked about end of March was supply in place to support $1 billion of demand, and that includes memory, that includes wafers, that includes packaging, that includes access to test equipment. For the $2 billion, we are now in the process of securing supply to support that. The teams are working around the clock to make sure we can find the right answers for our customers."

Strategic positioning

Strategically, Arm positions its AGI CPUs not as traditional off-the-shelf processors competing directly with merchant CPU vendors and/or custom silicon designed by (or for) leading hyperscale cloud service providers, but as scalable compute platforms and subsystems that hyperscalers and OEMs can use for specific workloads and vertically integrated AI stacks.

The first-gen Arm AGI processor was co-developed with Meta, which will be the first and lead customer for the CPU. Nonetheless, Cerebras, Cloudflare, F5, OpenAI, Positron, Rebellions, SAP, and SK Telecom plan to deploy the Arm AGI CPU for a variety of use cases that include agentic AI CPU workloads. These include accelerator management and control plane processing, as well as other CPU workloads that support AI agent infrastructure or typical cloud workloads.

While the AGI processors will not be available on demand like server CPUs from AMD and Intel, interested parties will be able to get AGI-based rack-scale solutions from such OEMs and ODMs as ASRock Rack, Lenovo, Quanta Computer (which is the leading supplier to Meta), and Supermicro.

On the hardware side, Arm claims that its AGI processor is the world's most efficient agentic CPU. In particular, Arm claims that its AGI CPU was purpose-built as a new class of processor for sustained parallel performance at rack scale, delivering high 'per-task performance' without throttling across thousands of cores and retaining modern data center power and cooling limits.

Arm's 1^st Generation AGI is a data center-bound processor that features up to 136 high-performance Neoverse V3 cores at up to 3.70 GHz, based on the Armv9.2 instruction set architecture, equipped with dual 128-bit SVE2 (Scalable Vector Extension 2) units per core, as well as 2MB of L2 cache per core.

The CPU features a 12-channel DDR5 memory subsystem supporting up to 6 TB of 8800 MT/s memory, providing up to 6 GB/s of bandwidth per core, and has an I/O that supports 96 PCIe Gen6 lanes with CXL 3.0 on top for caching and memory expansion. The CPU is comprised of two identical chiplets (with their own memory interfaces and I/O) made using a 3nm-class process technology and has a thermal design power of 300W.

Arm has a roadmap for its own AGI processors that spans years. While the company does not disclose it to the public, its management implies a consistent and significant core count increase, and believes that agentic AI workloads will call for racks full of CPUs rather than racks that pack a few CPUs and tens of AI accelerators. When it comes to agentic AI workloads, they will not call for more CPUs, but rather for more CPU cores; hence, the rapid core count increase seems to be a logical evolution for Arm's own AGI processors.

"The way I think they think about it is that while the ratios may not go to more CPUs than GPUs from a chip standpoint, they probably will from a core count standpoint," said Rene Haas, chief executive of Arm, during the recent earnings call. " CPUs today, the Arm AGI CPU, for example, has 136 CPU cores. [Nvidia's] Vera, that is 88. As I mentioned earlier, could I see those core counts doubling or quadrupling over the next number of years? Absolutely. […] Will you see many more CPUs inside a data hall, dedicated racks of CPUs that are doing agentic orchestration and scheduling and management? 100%."

With up to 136 highly high-performance cores optimized for agentic AI and data center workloads and available starting from Q4 2026, Arm's AGI CPU is poised to be in high demand from those who need high-end CPUs to run their AI agent infrastructure and whose software stack is already optimized for Arm.

Arm braces for AGI influx

Orders for Arm's 136-core AGI CPUs have doubled to over $2 billion since their announcement on March 24. The development is a result of the skyrocketing growth of demand for CPUs for agentic AI infrastructure and reflects similar occurrences at AMD and Intel. The company now expects to generate $15 billion in AGI CPU sales and $10 billion in IP revenue in fiscal 2031 (which ends on March 31, 2031), increasing its revenue by 9.5X in five years.

However, while $2 billion by FY2028 and $15 billion in FY2031 look like a huge amount of money, Arm will remain a strong contender, rather than a major supplier of data center CPUs, as AMD and Intel earn tens of billions per year selling their EPYC and Xeon parts and are projected to earn hundreds of billions in the 2030s.

Mercury Research believes that Arm could ship roughly 1.6 million of AGI CPUs over the next two fiscal years, which looks pale compared to nearly 20 million of EPYC and Xeon processors sold in 2025. Still, it should be noted that Arm does not plan to compete directly with merchant CPUs as its AGI processors will be available to select hyperscale CSPs and through OEMs and ODMs that will offer rack-scale solutions based on AGI CPUs.

Valve has released Proton 11.0-Beta1, but this perhaps inconsequential-sounding test version comes with something revolutionary – support for Arm Linux devices. It isn’t mainstream yet; it is a beta version, of course, but folks have already shown off the Steam UI running on a Nintendo Switch. So, Steam gaming support on more open gaming devices from brands like Retroid, AYN, and Ayaneo should be far easier.

Steam Linux ARM64 beta on Switch

— @aagaming.me (@aagaming.me.bsky.social) 2026-04-17T11:10:52.424Z

Steam Linux ARM64 beta running on Ubuntu Linux Noble Numbat on Nintendo’s popular handheld.

AAGaming, which posted the Steam on Switch demo on BlueSky, embedded above, says they have shared “a working copy of proton arm + steamrt arm that you can drop right into conpatibilitytools.d to play games with this client.”

You can find the official Proton 11.0-Beta1 release on Valve’s Proton GitHub, alongside some release notes. The key change, as far as Arm gamers are concerned, is flagged in a section under the heading ARM64 Builds. Not much is said, but handheld gaming enthusiasts on social media have been quick to realize the implications. The newly bundled FEX 2604 can enable Steam’s built-in x86 Windows to ARM Linux gaming compatibility on your Arm Linux device.

This new Arm Linux compatibility for Steam is thought to have been implemented now to prepare for the Arm-powered Steam Frame gaming headset, which we went hands-on with last November. That headset was described as a ‘streaming-first device’ but packs very respectable Arm mobile hardware, such as a Qualcomm Snapdragon 8 Gen 3 backed with 16GB LPDDR5X.

During our Steam Frame hands-on, we also learned about FEX bringing Steam games to Arm and running locally. In our demo, we saw the x86 version of Hades 2 running standalone with respectable performance at 1400p. The Valve rep clarified, “It’s actually running on Linux, running on Arm,” at the time. So, now we are seeing this ability surfacing in public betas, and already being used to get Steam OS and games running on alternative Arm hardware.

Valve is building a ‘Verified’ style list for the Steam Frame Arm hardware, much like its efforts seen on Steam Deck. Thus, users will know which games will run well locally and which they might prefer to stream from their gaming PCs.

Not interested in Arm? There's more...

In addition to the juicy Arm/Linux compatibility news, the new Proton 11 Beta includes goodies for existing Steam OS gamers. There’s a host of new certified playable titles for users, such as titles from the Resident Evil and Dino Crisis stables, plus Warhammer: Vermintide 2, SHOGUN: Total War, Breath of Fire IV, and more.

Numerous bug fixes are implemented in the release, too. For example, Valve has fixed Steam Overlay not working correctly with many EA games, and got the intro video playback in Crimson Desert to work properly.

Getting Proton 11 is easier if you already have a Steam OS device like the Deck. Just search for the update in the Steam Library.

Robust financial performance and strong growth expectations from ASML and TSMC this week boosted sentiment across the tech industry, lifting stock prices and market capitalizations throughout the AI supply chain. Among the most notable beneficiaries of this AI-driven momentum are AMD, with a market capitalization that reached an all-time high on Thursday, and Intel, whose market cap hit a 25-year high. Other notable beneficiaries are Arm Holdings and Nvidia.

AMD's market cap hit $454 billion as its stock price touched $278, an all-time high for the company. Intel continued its rally that began in early March, and its market cap reached nearly $340 billion on April 16 as its stock price tested $68. For Intel, this is its second-best result since August 2020, when its valuation briefly achieved $502.71 billion. Arm has been on the rise since early March; its current market capitalization is $174 billion as its stock approached $165.

All three CPU stocks are driven by AI hype — specifically, rapid adoption of agentic AI and retrieval augmented generation (RAG) systems that benefit from high-performance CPUs and high-bandwidth memory subsystems. While it remains to be seen whether AMD and Intel can actually increase sales of their EPYC and Xeon CPUs tangibly due to the adoption of agentic AI and RAG systems by various parties, it is obvious that both companies benefit from the AI infrastructure buildout and AI adoption by enterprises, as these drive demand for data center-class processors.

If you look at AMD's historical market performance, you'll notice that for roughly 20 years, AMD's market capitalization remained low and volatile, generally in the single-digit to low tens of billions. Peaks around the early 2000s (Athlon era) and mid-2000s (Athlon 64 and Opteron success) were followed by prolonged declines tied to execution issues and performance disadvantages compared to Intel. AMD's inflection started around 2017 with the launch of the Zen microarchitecture as well as EPYC and Ryzen processors on its base, though it was not until around 2020 when the company's market cap hit $100 billion. That was after the market finally believed that the company could execute.

Meanwhile, the majority of AMD's current valuation was created after 2020 and the valuation is based on the market's belief that AMD can benefit from industrial megatrends like AI and cloud computing. So far, AMD has been successful.

When it comes to Intel, its market cap was below its book value last August, but began to climb after the U.S. government, accompanied by SoftBank and Nvidia, injected some cash into the company. To a large degree, the company's current valuation is based on expectations about Intel's ability to execute and sell a boatload of processors for AI systems.

For Arm Holdings — which became a CPU maker a few weeks ago and which gets most of its revenue by licensing CPU technologies rather by selling hardware — the positive sentiment reflects its unique position in the client and AI ecosystems. Arm is the indisputable leader in smartphones and tablets and the rising star in automotive and client PCs. Meanwhile, the Arm ISA has become the architecture of choice for hyperscalers developing custom silicon.

As for Nvidia, its undeniable leadership in the AI infrastructure in general and AI accelerators in particular is driving its market capitalization towards $5 billion. Nvidia's market cap on Thursday reached $4.82 trillion, down slightly from $4.92 trillion on October 21, 2025.

Virtually all hyperscale cloud service providers (CSPs), as well as some of the leading developers of AI accelerators nowadays, have their own custom-silicon programs that are focused not only on developing AI accelerators, but also on custom general-purpose CPUs usually based on the Arm instruction set architecture (ISA). Over the next several years proliferation of custom CPUs based on the Arm ISA inside AI servers will increase to 90%, leaving x86 and Arm around 10%, according to Counterpoint Research.

x86 processors from AMD and Intel have long dominated general-purpose servers, which is why most of the AI servers initially relied on Opteron and Xeon processors. However, Arm-based custom CPUs that are tailored for specific data-intensive AI workloads are more cost and power-efficient. Furthermore, given the fact that AI workloads are emerging workloads, backward compatibility with x86 is not vital. To that end, AWS, Google, and Microsoft have developed their own proprietary Arm-based processors for their own workloads, whereas Meta is the alpha customer for Arm's own AGI processor.

As a result, adoption is unfolding across multiple hyperscalers in parallel. AWS is expanding the role of its Graviton processors across Trainium-based systems, while still retaining x86 in some configurations for compatibility reasons; Google's next-generation TPU infrastructure relies on its Axion Arm CPU; while Microsoft has paired its Azure Cobalt Arm CPU with its Maia accelerators from the beginning to build a vertically integrated AI infrastructure. Meta is also set to begin deploying Arm's own AGI CPUs shortly.

"The transition from x86 to Arm in AI servers is not a single switch," said Neil Shah, vice president of research at Counterpoint Research. "It has played out generation by generation, configuration by configuration. Hyperscalers are making deliberate choices based on their specific deployment needs, writing compatible and interoperable software, and the economics are very encouraging. The transition is expected to accelerate meaningfully in the second half of 2026, driven by the broad deployment of in-house Arm CPUs alongside next-generation ASIC platforms across major hyperscalers."

Nowadays, the majority of CPUs powering AI servers are still x86, but this is going to change shortly, and by 2030, 90% of AI servers that use custom processors will rely on Arm, leaving only 10% for x86 and RISC-V. It should be noted that loads of AI servers will continue to rely on off-the-shelf EPYC and Xeon processors from traditional suppliers, though broad adoption of Arm by hyperscalers for their custom silicon programs should be a signal for AMD and Intel to make their custom CPU programs more appealing to customers.

"Our analysis projects Arm-based CPUs will account for at least 90% of host CPU deployments in custom AI ASIC servers by 2029, up from around 25% in 2025, a structural shift driven by the accelerating rollout of in-house Arm CPU programs across major hyperscalers," Shah added.

AMD builds its own vertically integrated AI platforms featuring x86 EPYC processors, Instinct MI-series AI accelerators, Pensando DPUs, and Pensando NICs, so it is reasonable to assume that these CPUs are tailored for AI workloads. Meanwhile, Intel is developing custom Xeon processors for Nvidia's next-generation AI platforms, which suggests that these processors will also be optimized primarily for AI workloads. All in all, while Arm will get significantly bigger in the AI server realms over the next four to five years, x86 will continue to command a sizeable share of this market.

IBM and Arm on Thursday announced a strategic collaboration to co-develop dual-architecture enterprise platforms that would enable software designed for the Arm ecosystem to work on IBM Z mainframes and LinuxONE systems in emulation mode. The collab is designed to enable enterprises to run AI and cloud-native workloads originally developed for Arm on mission-critical IBM Z enterprise hardware with ultimate reliability, availability, and security.

Nowadays, a lot of AI frameworks as well as data-intensive cloud-native applications are developed for the Arm ecosystem, whereas IBM Z platforms (based on the Z390x or z/Architecture ISA) excel in reliability, availability, and serviceability but have a narrower native software stack. This is why enterprises increasingly operate a mix of legacy transaction processing alongside AI inference and microservices, which are typically deployed on separate Arm or x86 servers, according to IBM.

Running Arm workloads on IBM Z is designed to enable running a broad software ecosystem on IBM's Z mainframe systems, particularly those that are based on the Telum II processor and Spyre AI accelerator, through virtualization or emulation without porting them to IBM Z, which is costly, time consuming, and not common for the modern industry that relies more on x86 and Arm and less on IBM Z. Therefore, by bringing these newer workloads onto the same system, IBM reduces architectural complexity, lowers integration overhead, and simplifies operations.

Furthermore, this approach keeps workloads close to where critical data already resides: financial systems, government databases, and high-value transactional engines, which reduces latencies, minimizes security and compliance risks, and eliminates the need to replicate datasets across external platforms.

"IBM's defining role in shaping enterprise infrastructure spans decades, showcasing the breadth and commitment required to support our clients' most intensive and sensitive workloads," said Christian Jacobi, Chief Technology Officer and IBM Fellow, IBM Systems Development. "This moment marks the latest step in our innovation journey for future generations of our IBM Z and LinuxONE systems, reinforcing our end-to-end system design as a powerful advantage."

The model is not intended for performance-hungry applications. In addition, emulation and virtualization introduce a host of additional performance penalties, so do not expect IBM Z systems running Arm workloads on Telum II CPUs and Spyre accelerators to demonstrate leading performance. That being said, enterprise decision-making does not prioritize performance per se, but rather total cost of ownership, operational stability, reliability, risk mitigation, and scalability.

As a result, the trade-off may well be justified, particularly for those companies that already use IBM Z for mission-critical workloads and yet have to run additional workloads on different types of hardware. At the end of the day, IBM customers do not want to replace all of their hardware and mission-critical applications, but rather want their already deployed hardware and software to evolve, which includes running modern applications alongside legacy software. Whether or not this could lead to eventual inclusion of Arm-based CPUs or accelerators into IBM servers is something that remains to be seen, but IBM does not talk about it at this point.

Arm's introduction of its AGI CPU last week was an expected yet, still a milestone event that marked the transformation of Arm from a technology IP licensor into a supplier of standard CPUs competing directly with AMD, Ampere, and Intel. But the surprises did not end there, as the company intends to sell AGI processors to customers in China, despite the fact that Neoverse V3 cores that power the new silicon cannot be licensed to Chinese CPU developers due to sanctions.

"We just do not have any customers today that we are able to talk about publicly," said Rene Haas, chief executive of Arm, in an interview with ChinaDaily. "But we would expect the demand for this product to be just as strong in China as it is in the rest of the world."

Arm's AGI dual-chiplet processor packs 136 Neoverse V3 cores with 2 MB L2 cache per core running at 3.70 GHs, features a 12-channel DDR5 memory subsystem supporting 8800 MT/s memory, and has an I/O that supports 96 PCIe Gen6 lanes with CXL 3.0 on top for caching and memory expansion. The CPU is made using a 3nm-class process technology and consumes around 300W.

While Arm certainly would like to sell its AGI CPU everywhere, an interesting wrinkle from the announcement is that these processors will be available to buyers in China. This is a bit surprising as Arm cannot license its HPC-oriented Neoverse V cores to entities from the People's Republic (PRC) due to export restrictions. If Arm were to provide this core (or related design, such as CSS) directly to a Chinese chip designer, that would transfer advanced semiconductor know-how to a potential foe, something current U.S./U.K.-aligned controls aim to limit. However, the Arm AGI CPU is a finished product, not IP, which enables Arm to sell it to Chinese entities.

As Arm's AGI processor is a commercial semiconductor device, and by selling it, no design transfer occurs, its export is governed by different rules, such as performance thresholds like absolute performance, compute density, interconnect bandwidth, and so on. As it turns out, Arm's AGI processor with 136 cores offers performance that is compliant with the current U.S./U.K.-aligned export controls.

Arm positions Neoverse V3 as its fastest infrastructure and supercomputer core to date, but unlike prior generations, the company has not disclosed FP32/FP64 throughput, leaving only indirect clues from SIMD width and pipeline count. As a result, any FLOPS estimate today is inferred, not specified, so we cannot really say whether or not Chinese entities can now get a high-performance HPC-oriented processor with a high core count without any restrictions.

This is particularly interesting as Arm offers a reference-design 2-node blade, which is designed to fully populate a standard air-cooled 36kW rack that delivers 8160 cores. Arm has also teamed up with Supermicro to develop a 200 kW liquid-cooled system that can accommodate 336 Arm AGI processors and deliver over 45,000 CPU cores in a single deployment. Obviously, such building blocks can be used to build actual supercomputers with decent FP32 and FP64 throughput, possibly giving Chinese entities an opportunity to get modern Western supercomputer technologies.

Arm today announced the AGI CPU, an up-to 136-core data center processor family that the company designed and will sell as finished silicon. The chip, built on TSMC's 3nm process with Neoverse V3 cores, was co-developed with Meta and represents the first time in Arm's 35-year history that the company has shipped its own production processor rather than licensing IP to partners.

The AGI CPU has been designed for what Arm calls "agentic AI infrastructure," the CPU-side orchestration work required to coordinate accelerators and manage data movement in large-scale AI deployments.

136 Neoverse V3 cores at 300 watts

The chip packs up to 136 Neoverse V3 cores running at up to 3.2 GHz all-core and 3.7 GHz boost across two dies, all within a 300-watt TDP. It supports 12 channels of DDR5 memory at up to 8800 MT/s, delivering more than 800 GB/s of aggregate memory bandwidth or 6GB/s per core with a target of sub-100ns latency. I/O includes 96 PCIe Gen6 lanes and native CXL 3.0 support for memory expansion and pooling.

Arm's reference platform is a 10U dual-node server compliant with the Open Compute Project's DC-MHS standard. Two AGI CPUs fit per blade, and a standard air-cooled 36kW rack holds 30 blades for 8,160 cores total. Arm has also partnered with Supermicro on a liquid-cooled 200kW configuration that houses 336 chips and more than 45,000 cores.

Arm claims the AGI CPU delivers more than two times the performance per rack compared to the latest x86 platforms. That figure is, of course, based on the company’s own internal estimates at this stage, not independent benchmarks.

GPUs have made most of the headlines surrounding AI hardware to date, but there’s a demand for more powerful general-purpose compute as agentic systems like OpenClaw explode in popularity. Arm is clearly hoping it can meet and cash in on this demand — and hopefully that won’t be to the detriment of non-AI customers, who seem to have long since been forgotten by the likes of Nvidia and Micron.

OpenAI among early customers

Meta served as the lead partner on the project and plans to deploy the AGI CPU alongside its custom MTIA accelerators. Santosh Janardhan, head of infrastructure at Meta, said the two companies worked together on the chip and are committed to a multi-generation roadmap.

Beyond Meta, Arm confirmed commercial commitments from Cerebras, Cloudflare, F5, OpenAI, Positron, Rebellions, SAP, and SK Telecom. Sachin Katti, head of industrial compute at OpenAI, said the AGI CPU will play a role in OpenAI's infrastructure by strengthening the orchestration layer that coordinates large-scale AI workloads.

Arm has historically operated as an IP licensing company. Its partners, from Apple to Nvidia to AWS, design their own chips using Arm's instruction set architecture and core designs. The AGI CPU adds a third option alongside IP licensing and Arm's Compute Subsystems (CSS) program: Arm-designed, production-ready silicon that customers can deploy directly.

Arm said the AGI CPU product line will continue in parallel with the Arm Neoverse CSS product roadmap, and that follow-on products are already committed. The company seems keen to point out that this is an additive move rather than a pivot that competes with existing licensees, though how Arm manages that as it sells chips into the same data centers as Nvidia Grace, AWS Graviton, Google Axion, and Microsoft Cobalt remains to be seen.

EA released its 'Javelin' anti-cheat last year to power the multiplayer gameplay in Battlefield 6. Like every other anti-cheat software out there, it runs at the kernel level to spoof out any bad actors or code, and EA has said it's been very successful thus far. So much so that the company is now seemingly looking to port it over to Arm and eventually Linux, according to a new job listing.

Titled "Senior Anti-Cheat Engineer, ARM64," the listing is pretty self-explanatory. The main responsibility endowed upon this engineer would be to "develop a native ARM driver" for the Javelin anti-cheat. Save an official announcement, this confirms that EA wants games like Battlefield to run Arm-based devices soon. Apple Silicon is not EA's interest, and Qualcomm's X-series hasn't been enough of a hit to justify this either.

That leaves us with Nvidia's upcoming N1/N1X chips that are highly rumored to launch this year. The timing for this role aligns with the chips' reported debut. The N1 silicon is said to finally bring Windows-on-Arm to the mainstream, with a focus on gaming that Arm has generally never been known for. After all, there's a 20-core Arm CPU and an RTX 5070-level GPU in there, according to the rumor mill, so it's plenty powerful.

The job description specifically mentions, "development of [our] Windows on ARM support," so there's no ambiguity over what platform we're talking about. Previously, Epic added support for Easy Anti-Cheat on Arm and Linux devices, but its flagship title, Fortnite, still doesn't run on those machines. Microsoft also released the Xbox app on Arm-based Windows 11 devices just a month ago.

Most anti-cheat software is only designed to run on x86-based Windows machines — part of Valve's push for Linux gaming has been held back by this. Similarly, Arm-based devices can't run anti-cheat either, which shuts out a large chunk of the most popular games from the platform. Therefore, getting a native Arm version of Javelin to work on N1-powered machines would be a huge step forward for non-x86 PC gaming.

In fact, the job listing also includes a point that reads: "Chart a path for EA Javelin Anticheat to support additional OS and hardware in the future, such as Linux and Proton." Namedropping Proton is a big deal; it shows clear intent to eventually have support for Javelin on things like the Steam Deck and Steam Machine — another audience EA doesn't want to miss out on, which can influence others like Riot and Epic to follow suit.

Semiconductor and software design company Arm Holdings signed a deal with the Malaysian government in March 2025, wherein the latter would pay US$250 million to the former over ten years for the provision of licenses and technical know-how to local companies. However, three non-government organizations have alleged that there was financial misconduct relating to the agreement, with the Malaysian Anti-Corruption Commission (MACC) seizing MYR 5 million or around US$1.27 million from a safehouse owned by a major political figure, according to Digitimes. Former Economy Minister Datuk Seri Rafizi Ramli, who spearheaded the negotiation with Arm, is also being linked to the investigation, although Ramli claims that the move is politically motivated.

Ramli resigned from his cabinet position in mid-2025 after being defeated in an internal election by the prime minister’s daughter for the ruling party's second-highest position. Since then, he has been a vocal critic of the prime minister and also publicly attacked the head of the MACC for abuse of power during a rally in Kuala Lumpur last month. Immediately after this, the former minister alleged that the MACC opened an investigation against him, framing it as a move to politically persecute him. Ramli even said that he will pursue legal action if the investigation does not result in a court case. Alongside scrutiny into Ramli, Digitimes reports that a former Army Chief has also been arrested by MACC.

The MACC investigation is putting Malaysia’s high-tech strategy at risk, especially as this could dampen investor confidence. Despite that, the Economy Ministry reiterated that the Arm deal was a Cabinet policy decision, and that it will move forward despite the investigation. In fact, two companies have reportedly received access to Arm IP, and local chip design capacity will continue growing under the government’s plan. Still, it intends to cooperate with the MACC and provide documents and statements, as needed.

Malaysia is taking advantage of the falling out between the U.S. and China by building up its own semiconductor industry, with many U.S. tech companies that previously manufactured in China moving to the Malaysian peninsula. Because of this, the government is pushing its local industries to level up towards high-value manufacturing, and the Arm deal is designed to help it achieve that. It seems, though, that corruption and politics could affect its progress towards its goal.

We've been waiting on Nvidia's long-rumored N1X Arm chips for a while at this point. Through several leaks and official teases, the company's ARM-based consumer SoC has excited many as it's poised to open the gates of high-end Arm performance on Windows machines. Interestingly, it seems like the "Chinese Nvidia" has beaten the Green Team to the punch with its own custom Arm chip in a new laptop.

Moore Threads, the region's local darling, has just launched the "MTT AI Book" — a new thin-and-light laptop powered by an in-house "MT1000" CPU. What's special about this chip is that it's Arm-based and features 12 CPU cores clocked at 2.65 GHz (base), along with an unknown GPU that's based on its MUSA microarchitecture. The NPU is capable of delivering up to 50 TOPS of AI compute, similar to AMD's Strix Point.

The SoC as a whole is called "Yangtze" (translated) and is paired with 32 GB of LPDDR5X-7500 unified memory, meaning that's shared between the MT1000 CPU and the MUSA-based GPU. The 1 TB SSD onboard carries a Linux-based operating system called "AIOS," but the device can also run Windows. That's what makes this exciting, because now we're in N1 territory. Unfortunately, we're not looking at Windows-on-Arm here, but rather a virtualization-based approach where Windows just sits inside a VM.

The Nvidia N1 silicon is said to have a 20-core ARM CPU and an RTX 5070-level GPU because Jensen Huang himself confirmed it powers the GB10 Superchip inside the DGX Spark.

Even if we imagine for a second that Moore Thread's Arm SoC is natively running Windows and is optimized — it just doesn't stack up to what Nvidia is cooking. The Green Team's offering is supposed to open up AI and gaming in a whole new way for Windows-on-Arm. Meanwhile, Qualcomm is already trying with its own X series of SoCs.

AMD and Intel, on the other hand, don't have competing Arm products so we can't really speculate much. The Red Team's Strix Halo chips, which feature desktop-level integrated graphics, and the rumored Nova Lake-AX (cancelled?) lineups are both technically in the spot that N1X might gun for: a powerful, portable machine with strong battery life.

Moore Thread's Yangtze doesn't seem to be there yet, and that's proven by its Geekbench listing: it scores 1,127 points in the single-core test and 7,420 points in multi-core. Those numbers are very underwhelming. The most modern CPU we could find around these results was the Ryzen 3 7320U at 1,112 single-threaded points; even a recent Core i3/Core Ultra 3 SKU scores more than that.

Apart from its intriguing silicon, the MTT AI Book features a 2.8K 14-inch OLED display running at 120 Hz. The port selection seems to be limited to just 3x USB-C ports, and the battery is rated at 70Wh. The laptop weighs 1.5 kg despite being CNC-milled out of a "6-series" aluminum alloy. It also looks very similar to a MacBook Air and is priced at 9,999 CNY on JD.com, or about $1,447 USD.

We hope to see more mainstream media coverage of this device with independent reviews that test the Arm-based SoC's capability. This was just one Geekbench listing, so there's still a chance that with the right drivers and firmware tuning, the MTT AI Book can deliver better performance. It's aimed at AI applications, though, so we may not be that impressed by its graphical prowess when it surfaces.

Nvidia has sold off the last of its shares in Arm Holdings, some six years after it attempted to buy the company outright for $40 billion, but was stopped by EU and UK regulators, as Bloomberg reports. This amounts to 1.1 million shares worth an estimated $140 million. Though this draws to a close Nvidia's ownership ambitions, it won't end the partnership, with Nvidia and ARM still set to work together on its Arm-based CPUs, including in its Vera Rubin platform.

The actual share disposal took place sometime in the last few months of 2025, with only the filing now revealing the divestment of Nvidia's ARM holdings. It feels a little anticlimactic after talk of such enormous deals not so long ago. But in 2026, that deal has in turn been dwarfed by the scale of investments among the major tech firms, often reaching into the hundreds of billions.

Even Nvidia's recent chip deal du jour, its non-exclusive licensing agreement with AI inferencing hardware firm Groq, worth $20 billion, feels somewhat pedestrian now.

The deal that never was

In September 2020, while the world was grappling with the global pandemic, Nvidia was making moves to purchase British software and semiconductor design firm, Arm Holdings. Its majority shareholder, Softbank - the same Japanese investment firm that has invested so heavily in OpenAI and many other AI players approved the deal, and the agreed figure was set at $40 billion. If it had been finalized, it would have been one of the largest semiconductor company acquisitions ever, falling just behind Dell's $64 billion purchase of EMC in 2016.

Nvidia pledged to maintain Arm's licensing model and would continue to work with major partners like Samsung, Qualcomm, and Apple, among others. But those same companies weren't so keen, and many raised concerns to regulators over anti-competitiveness.

Authorities in the UK and EU, which had already been scrutinizing such a large-scale deal, dived into the details and announced investigations in early 2021. They raised concerns that a deal of this magnitude between two key players in the technology hardware space could result in worse products for consumers, alongside higher prices. With Arm's hardware present in a range of components from different companies, it was felt that with ownership, Nvidia would have too much power over its competitors.

The back and forth went on for several more months before the deal ultimately started to fall apart in early 2022. Despite Nvidia's efforts, regulators finally put a halt to it, forcing Nvidia to pay a $1.25 billion exit fee.

Softbank ultimately took ARM into an initial public offering the next year, valuing the company at just under $55 billion. As of February 2026, ARM has a market cap of $133 billion.

Onwards and upwards

Nvidia continued to work with Arm on the CPU cores of its leading chips, and will continue to do so even after divesting itself of all remaining shares. Its Grace CPUs and Rubin CPUs are a major part of its AI hardware offerings, and that's unlikely to change in the future.

That deal arose in the wake of all the other major AI infrastructure deals of 2025, including a number of circular deals among the main players. Nvidia now holds stakes in many of the companies that are helping to make it the most valuable in the world. It has part stakes in Intel, CoreWeave, Nokia, and Synopsys. Its most recent deal saw it invest $20 billion in Groq to acquire both its inferencing hardware technology and some of its key staff members, who will now work for Nvidia to advance its various AI hardware efforts.

As one of the few companies making serious profit from the AI boom, Nvidia has used large portions of that funding to invest it back into the industry, as well as ensuring it has its fingers in as many pies as possible. Although it's claimed $100 billion deal with OpenAI may not pan out, it has pumped funds into many companies, which have helped maintain the AI rollout momentum over the past year.

GPU dominance

As much as the Arm Holdings deal would have given Nvidia a supreme position within the hardware space, it's made its way there anyway without them; it just did it with its GPUs instead. Now the world is on the AI hype train, its enterprise revenue has exploded, and it's become far and away the most valuable company in the world, with a current market cap of 4.5 trillion. That's 3,000% higher than it was when Nvidia tried to buy Arm way back when.

Today, Nvidia's throwing out tens of billion-dollar deals like they're pocket change, and at that kind of valuation, they almost are. After the Groq deal in December, it just struck a new deal with Meta to provide millions more GPUs and Spectrum-X Ethernet switches.

So, despite losing out on the overall Arm deal, Nvidia will continue to work with the company for its Rubin systems, DGX Spark CPUs, and presumably, we'll see the heavily rumored N1X chip rear its head around at some point soon.

IoT specialist M5Stack has released an alluring, translucent, RGB-infused, tetrahederal mini-PC dubbed the AI Pyramid Computing Box. Despite its curb-appeal, and great connectivity, computing enthusiasts should be clear that this cute little computer is designed for local-AI and edge deployment. M5Stack’s octa‑core Cortex‑A55 and NPU-powered box of tricks isn’t really a contender to replace your x86 productivity and gaming workhorse.

So, we’ve established the purpose of the AI Pyramid, let’s look closer at the compelling specs for edge-AI purposes. This delightful device is based around an Axera AX8850 SoC, which integrates an octa-core Arm Cortex-A55 processor, a 24 TOPS INT8 NPU, and 8K H.264/H.265 hardware encode/decode engines.

With the 4GB model ($199) there is 2GB available to the system and 2GB reserved for the NPU and video engines. M5Stack’s AI Pyramid Pro 8GB model ups the split to 4GB/4GB. Even the higher-spec model would be stingy for a desktop Linux OS, which is one of the major reasons we said this isn’t going to find favor as a general purpose mini-PC.

M5Stack reckons that the true sweet spot deployments of its AI Pyramid will include: AI PC & Edge Intelligent Terminals; Smart Interactive Devices such as Home Assistant, AIGC, voice cloning, meeting transcription; AI Vision Gateways; Local AI Smart Photo Album (Immich); and AI-Powered Intelligent Security (Frigate).

For hardware clarity, we’ve tabulated the full, extensive specs list of the AI Pyramid Computing Box, below.

Additionally, the M5Stack AI Pyramid comes with some nice frills such as the headlining RGB lighting (consisting of 48 LEDs in four segments), a small 32×128 pixels OLED status display, quad-microphones, a built-in speaker, is cooled by a single thermal sensing active fan, and has two user-configurable buttons.

M5Stack's compact translucent device comes in at an almost impulse-buy price (in 4GB and 8GB versions). Maybe that’s why it has completely sold out so quickly. Hopefully, we will see restocks soon, without any RAMpocalypse-style inflation changing those sticker prices for the worse.

A developer has ported Doom to a pair of earbuds. That kind of statement might not have the impact it once had, after tales of high jinks and ports of Doom stretching from quantum computers, to USB chargers, to lawn mowers. However, as earbuds don’t have displays (yet), Arin Sarkisian also devised a canny method to stream the Doom action to another device, or even via the internet.

You can’t just use any earbuds for this latest Doom-on-x episode of development gymnastics. Currently, this Doom port only works on the PineBuds Pro, which Arin-S says are the only earbuds with open source firmware. And, yes, these buds are made by the same folks behind the RISC-V-powered soldering irons, like the Pinecil V2 we reviewed in Aug 2022.

In terms of improbable CPU horsepower, the PineBuds Pro are right up there. Inside their snug ear-fitting shells, alongside the audio drivers and battery cells, there is an Arm Cortex-M4F processor. Arin-S tweaked the open-source firmware to boost the CPU clock from 100 to 300 MHz (an astounding OC nowadays), and disabled the low-power mode for the most Doom-tastic experience.

Some more hurdles of using the PineBuds were the less-than-ample RAM and storage. The developer managed to get Doom to run in under 1MB by “pre-generating lookup tables, making variables const, reading const variables from flash, disabling Doom's caching system, removing unneeded variables.” Moreover, the shareware Doom 1 WAD assets file was 4.2MB, just over the 4MB storage on the PineBuds. This payload was reduced to 1.7MB after borrowing some pre-modded-for-size Doom resources.

Lastly, the no-screen conundrum presented its own set of challenges. Arin-S decided the PineBuds Pro’s UART connection was the best choice for achieving game visuals (the only other choice was Bluetooth). A mix of bandwidth, image compression shenanigans, and demands on the Cortex-M4F ultimately meant the best achievable performance was about 18fps, in practice. In theory, it should have been nearer 25fps, but the dev reckons the CPU hadn’t the grunt to keep up with converting the MJPEG stream at that rate.

No PineBuds Pro, no problem

Arin-S shares all the resources needed to get your own set of PineBuds Pro up and running, with Doom loaded. However, the developer has also generously decided to set up a website that streams Doom action from his own pair of PineBuds Pro.

Head on over to this nicely crafted webpage to join the queue (yes, it is popular) and partake in some online Doom powered by the remote earbuds.

Lastly, the developer says that he is currently looking for work if you think you might need someone with his unique set of skills.

Microsoft just announced that you can now install the Xbox app on all Arm-based Windows 11 devices. The Windows Experience Blog said that more than 85% of the Game Pass catalog is compatible with Windows 11 devices that run on Arm-based processors, and that it’s continuing to expand compatibility with the remaining titles. This news comes as Nvidia is reportedly debuting Arm-based N1X gaming laptops this quarter, challenging the established x86 architecture that still dominates the gaming PC industry and offering an alternative to Qualcomm’s Snapdragon X SoCs.

“Players can now download and enjoy a wide range of titles from the Xbox PC app game catalog on Arm-based Windows 11 PCs,” the company said in its blog post. “Today, more than 85% of the Game Pass catalog is compatible with these PCs, and we’re actively working with partners to expand support even further. For titles not yet supported running locally, Xbox Cloud Gaming remains an easy way to jump in while we work with partners to broaden compatibility.”

The company achieved this with the continued development of Prism, an emulator designed to run x86/x64 apps on Arm hardware. The addition of AVX and AVX2 support has expanded its compatibility with several modern games, while Epic Anti-Cheat support means that you can continue playing popular multiplayer titles on Arm-based Windows 11 PCs.

This is going to be excellent news for gamers, as it will allow them to have more hardware options when it comes to gaming. Even though many feel that x86 processors still deliver more raw horsepower, the efficiency of Arm-based chips is certainly attractive to those who prefer gaming on the go. Handheld console maker Ayaneo has been making Arm-based devices for several years now with the Snapdragon G3x Gen 2-powered Pocket S, but the lack of PC titles for Arm SoCs meant that it’s designed for Android gaming. This announcement would likely push handheld gaming console manufacturers to develop Arm-powered options, giving us the option to play PC games on battery for much longer than what we usually get from x86-powered devices.

Arm’s threat to x86’s dominance has got long-time rivals Intel and AMD cooperating, with the two companies forming the x86 Ecosystem Advisory Group in 2024. By the time of its first anniversary, we’ve already seen AMD and Intel agreeing to support some technologies, like the Advanced Matrix Extension, AVX10, Flexible Return and Event Delivery, and x86 Memory Tagging, for upcoming and future processors.

The Xbox app’s support for gaming on Arm-based processors will pose a threat to x86, especially now that a lot of games are compatible with the system. Even Steam is experimenting with Arm compatibility, with the client app adding native support for Apple silicon in 2025. But while this might make Intel and AMD nervous about their duopoly, especially in the gaming PC space, this is just going to be good news for gamers and consumers, as it drives competition between different ISAs and could even be the engine that would lead to more innovations in processors.

Things aren't looking too rosy for Snapdragon X Elite PCs. Although they made an initial splash in the market, combining excellent battery life with mid- to high-end performance, the machines' popularity suffered due to software support issues. As an example, Phoronix ran Linux performance tests on the SoC, and there have been significant regressions in just a few months.

Phoronix's software setup is arguably a best-case scenario for the chip as of this writing. The site used the latest Ubuntu 25.10 "Questing Quokka" release, with the X1E Concept packages added. The distro offers about as fresh a stable kernel as you can get, while the X1E packages add Snapdragon-specific enhancements, most of them courtesy of the Linaro consortium.

Given that progress on Arm64 support on Linux has been improving at a fairly nice clip, as evidenced by the now-present nested virtualization support and Guarded Control Stack (GCS), you'd expect the fresh version to be an upgrade. Alas, compared to testing done in September, Phoronix saw rather significant regressions in performance and thermal behavior — quite the odd double-whammy.

The regressions are so bad that in many benchmarks, the test machine actually performed worse than when it was first tested, still under Ubuntu 24.10 without as many enhancements. The situation is currently so dire that the reviewer concluded that performance is back to the level of now-aging Intel Tiger Lake and Ryzen 7 Pro machines. Needless to say, Phoronix advises Linux users to shop for a conventional laptop rather than invest their cash in a Snapdragon machine, at least for the time being.

Things look much rosier on the Windows side of the fence, with 2025 seeing the introduction of an Arm-native Google Drive client, and Adobe launching Arm versions of both Adobe Premiere and After Effects. Likewise, Microsoft's x86-to-Arm emulation layer called Prism saw some fixes and added AVX/AVX2 support.

It's worth noting that the thermal-induced crashes and/or throttling can be caused by simple bad luck with the tester machine, an Acer Swift 14 AI laptop. Although Snapdragon X machines have been known to experience BSOD issues in Windows land due to firmware bugs, the laptops themselves are generally considered high quality and run cool and quiet.

This month, Amazon Web Services introduced the Graviton5, its fifth-generation custom general-purpose server processor, designed to compete against industry-standard CPUs from AMD and Intel in AWS's data centers. The new processor extends AWS's in-house Arm-based CPU program with a CPU that packs up to 192 cores and 180 MB of L3 cache, and is designed to compete with higher-end AMD EPYC and Intel Xeon processors, potentially replacing some of them in AWS data centers.

At a glance

The AWS Graviton5 processor is fabricated using a 3nm-class process, likely by TSMC. The processor integrates 192 Neoverse V3 cores alongside an assumed 180 MB L3 cache. AWS says that the new CPU will deliver 25% higher performance compared to its predecessor, which appears to be conservative, as the Graviton5 offers a twofold increase in the number of cores. The chip uses the Armv9.2 ISA that brings several microarchitecture enhancements and a fivefold increase in L3 cache size.

The new processor is now available in Amazon EC2 M9g instances in preview, while compute-optimized C9g and memory-focused R9g variants are scheduled for launch in 2026. The current EC2 M9g instances are up to 30% faster for databases, up to 35% faster for web applications, and up to 35% faster for machine learning workloads compared to M8g, according to AWS.

Diving deeper: 192 Neoverse V3 cores

Amazon Web Services is intentionally opaque about the exact specifications and internal design of its Graviton5 CPU. Nonetheless, it offers comparisons with the previous-generation Graviton4 chip, which allows us to decode some details and delve into them with a little more depth.

AWS and Arm officially confirm that Graviton5 integrates 192 Neoverse V3 cores per package, fabricated using a 3nm-class process, making it the densest CPU in the Graviton lineup and the densest Armv9.2 processor available to date. The internal layout of the processor has been redesigned to reduce communication overhead, and AWS claims up to 33% lower inter-core latency, which is particularly noteworthy given the twofold increase in core count.

When we discuss Neoverse V3, we cannot help but think about the Arm-developed compute subsystems (CSS). While Arm confirmed that we are dealing with Neoverse V3, neither Amazon nor Arm has confirmed that Graviton5 uses Arm-developed CSS. That means we're likely dealing with a unique design in Graviton 5.

In performance comparisons between the Neoverse V3 core and its predecessor, Arm claims a 9%-16% uplift over Neoverse V2 across general cloud workloads and up to 84% in AI data analytics. This is one of the reasons why AWS is so conservative about performance upticks, both for Graviton5 as well as compute-intensive M9g instances in general. Another reason for AWS's conservative performance estimate is that it does not sell leading-edge performance like AMD or Nvidia, but rather predictable performance per dollar and scalability in the cloud. Nonetheless, with a 192-core processor, AWS puts itself into the highest league among CPU developers.

L3 cache replaces system-level cache

One interesting thing to note about Graviton5 is that it comes with L3 cache, not system-level cache like Graviton4. While L3 and SLC in data center CPUs have a lot in common, they are not the same thing. Traditionally, L3 cache is a last-level cache located inside each compute tile or core cluster in a data center CPU. L3 primarily serves CPU workloads by reducing DRAM access; it is optimized for low latency and directly participates in the core's coherence protocol. Therefore, L3 is tightly coupled to the cores and is physically close to them.

By contrast, SLC sits outside the core clusters on the SoC fabric and is shared by all CPU cores, various other accelerators, I/O devices, NICs, and DMA engines. It tends to be much larger (often 100–300+ MB) and optimized for throughput rather than latency, as it acts as a global buffer that reduces pressure on DRAM and provides coherent access for heterogeneous compute blocks. SLC can improve scaling for very high core counts and enables unified memory semantics across CPUs, GPUs, and on-die accelerators, a role traditional L3 caches cannot fulfill on their own.

Amazon has not publicly explained the design decision, but based on Graviton4’s architecture and what we know about Graviton5, the reason is almost certainly architectural scalability. The move from SLC in Graviton4 to a large 180 MB L3 in Graviton5 is not cosmetic; it reflects fundamental changes in how a 192-core processor moves data, manages latency, and maintains coherence.

Graviton4's architecture — 96 Neoverse V2 cores, a CMN-700 mesh, 12 DDR5-5600 channels — operates efficiently with a centralized or semi-centralized SLC. But doubling the core count to 192 dramatically increases mesh traffic, hop distances, and contention on any unified cache structure. At this scale, a monolithic SLC could almost certainly become a latency bottleneck and would not support AWS’s claim of up to 33% lower inter-core communication latency. A distributed L3 sliced across the die allows hot data to remain physically close to compute clusters, reducing average access latency and improving overall coherence behavior.

The fivefold cache expansion AWS advertises reinforces this architectural necessity. Scaling Graviton4's 36 MB SLC by that factor yields 180 MB, and AWS's additional statement —2.6X more cache per core, at double the core count — implies ~187 MB total, which aligns with a large, multi-slice L3 rather than a single SLC block, which would create routing complexity.

Finally, L3-based designs offer stronger multi-tenant performance predictability, which is crucial for AWS. Under cloud workloads, shared caches experience heavy cross-tenant interference and variable latency, so when designing cache subsystems, developers must take into account AWS's use case. To sum things up, the shift to a distributed L3 was a necessary architectural evolution for Graviton5.

New memory subsystem, I/O, and security features

Just as AWS didn't disclose many details about other design aspects of Graviton5, it also didn't disclose much about the memory subsystem of the CPU. It goes without saying that Graviton5's memory subsystem is more powerful than that of Graviton4, as it supports higher memory speeds, which likely means that it at least retains a 12-channel memory subsystem of the Graviton4, but with higher data transfer rates (i.e., higher than DDR5-5600).

A 12-channel DDR5 design operating at 6400 MT/s would provide around 614 GB/s of aggregate bandwidth, which translates to approximately 3.2 GB/s per core, which is actually lower than 5.6 GB/s per core in the case of Graviton4. However, the larger L3 cache could compensate for this decrease in memory bandwidth. Then again, we do not know the exact number of memory channels supported by Graviton5.

Input/output throughput is similarly increased, according to AWS: network bandwidth is up by 15% on average across instance sizes, with as much as double the throughput for the largest configurations. Storage bandwidth through Amazon EBS rises by around 20% on average, according to AWS. These gains are designed to improve performance not only for compute-heavy applications, but also for distributed systems that depend on fast storage and networking.

On the security side, Graviton5 is built on the AWS Nitro System, with sixth-generation Nitro Cards that handle virtualization, networking, and storage. AWS has also introduced a new component called the Nitro Isolation Engine, which the company describes as a formally verified isolation layer. Instead of relying solely on conventional security validation, the Isolation Engine uses mathematical proofs to demonstrate that workloads are separated from each other and from AWS operators. The architecture enforces a zero-operator-access model, and AWS plans to allow customers to review the implementation and the formal proofs behind it to ensure maximum security. Such security measures could be a part of the company's effort to attract clients who have traditionally used on-prem servers.

Wrapping it up

AWS's new Graviton5 processor offers a 192-core, 3nm Arm-based CPU with around 180 MB of L3 cache. This positions the cloud giant as a competitor to the high-end AMD EPYC and Intel Xeon solutions for data centers. The CPU integrates Neoverse V3 cores and delivers an advertised 25% performance uplift, which is conservative given the twofold increase in core count, large microarchitectural improvements in the Armv9.2 ISA, and a fivefold increase in cache capacity. Also, AWS confirms 33% lower inter-core latency due to a redesigned internal layout but has not disclosed whether it uses Arm's CSS, suggesting that Graviton5 may be a unique Annapurna Labs design built around Neoverse V3 cores.

A key architectural shift is replacing Graviton4's SLC with a large distributed L3 to enable better coherence scaling across 192 cores and predictable latency. The processor also gains a faster memory subsystem (likely retaining 12 channels at higher DDR5 speeds), improved network and storage bandwidth, and the new Nitro Isolation Engine, which uses formal verification to guarantee tenant isolation and enforce zero-operator access.

At present, Graviton5 powers new EC2 M9g instances — up to 30% – 35% faster for databases, web services, and machine learning — and compute-optimized C9g and memory-optimized R9g variants will follow in 2026.

Korea's Fair Trade Commission (KFTC) has carried out an on-site investigation of Arm's office in Seoul, according to a report published by The Korea Times. The article cites unnamed sources — which is standard practice for the publication — and says the raid is tied to allegations of unfair market practices involving Arm's long-running licensing dispute with Qualcomm. The KFTC declined to comment on the report.

According to the report, the investigation focuses on whether Arm has improperly limited access to its chip-architecture technologies after years of maintaining a more open and widely accessible licensing model. The inquiry reportedly stems from a complaint Qualcomm filed in March with antitrust authorities in Korea, the United States, the European Union, and other jurisdictions. Qualcomm alleges that Arm restricted it from continuing to use licenses originally held by Nuvia, the custom-CPU startup Qualcomm acquired in 2021 that is responsible for the design of its powerful Snapdragon X Elite processors and their soon-to-be-released successors.

The licensing battle between the two companies has been underway for more than three years. It began when Arm terminated Nuvia's Architecture License Agreement (ALA) shortly after the acquisition, arguing that Qualcomm needed to renegotiate the license to continue using Nuvia's custom-core designs. Qualcomm countered that its own longstanding Arm architecture license allowed it to develop and use the Nuvia technology without any new agreement.

In September, Qualcomm won a decisive victory in U.S. federal court. The judge affirmed a jury finding that Qualcomm's use of Nuvia-derived technology was covered under its existing Arm license and rejected Arm's remaining claims, shutting down Arm's attempt to block Qualcomm from using the designs. Arm has said it intends to appeal the ruling, but legal wins do not resolve the broader commercial and regulatory tensions around Arm's licensing practices.

Qualcomm itself filed a separate lawsuit against Arm earlier this year, accusing the company of withholding required technical deliverables, interfering with Qualcomm's customer relationships, and attempting to shift its business model in ways that disadvantage existing licensees. In that suit, Qualcomm argues that Arm's recent conduct represents a departure from its traditional role as a neutral technology supplier.

Regulators appear to be paying attention. Korea is home to several major Arm licensees — including Samsung, both a rival and a large partner for Qualcomm — and the country's antitrust authorities have a history of aggressive enforcement in cases involving global technology suppliers. An on-site inspection typically signals that regulators have found the complaint credible enough to warrant closer scrutiny, though it does not indicate any preliminary finding of wrongdoing.

Whether the KFTC probe will broaden into a deeper multinational regulatory investigation remains unclear, and Arm has not yet issued public comment on the Korea Times report. Qualcomm, for its part, has been escalating its complaints worldwide following the court victory, seeking to frame Arm's behavior as a systemic issue rather than a contractual dispute limited to Nuvia's chips.

For now, the Korea Times report adds only one concrete data point: Korean regulators are taking at least an initial look. With Arm's appeal pending in the U.S., Qualcomm's countersuit still underway, and multiple antitrust agencies reviewing the allegations, the global fight over Arm's licensing model isn't close to finished. As more filings surface and regulators weigh in, the full scope of the conflict as well as its implications for the broader Arm ecosystem will become clearer in the months ahead.

Arm and Nvidia announced at the Supercomputing '25 conference that Arm had joined the NVLink Fusion ecosystem, marking a major advance for the technology, which is now supported by two major microarchitecture developers and four CPUs developers in total. For Nvidia, this means that Arm's customers will develop processors that can work with Nvidia's AI accelerators, whereas Arm will also be able to design CPUs that could compete against Nvidia's own or Intel processors in Nvidia-based systems.

"Arm is integrating NVLink IP so that their customers can build their CPU SoCs to connect Nvidia GPUs," said Dion Harris, the head of data center product marketing at Nvidia. " With NVLink Fusion, hyperscalers can significantly reduce design complexity, save development costs, and reach the market faster. The addition of Arm customers provides more options for specialized semi-custom infrastructure."

Arm is a large company with diverse businesses, including ISA and IP licensing, and the development of custom CPUs and system-on-chips (SoCs) for large customers. For each type of business, NVLink Fusion support gives certain benefits.

As an IP provider, Arm gets a major new competitive lever in the data-center market by supporting NVLink Fusion. By integrating NVLink IP directly into its architecture portfolio, Arm can offer its licensees a ready-made pathway to build CPUs that plug natively into Nvidia's AI accelerator ecosystem. In theory, this makes Arm-based designs far more attractive to hyperscalers and sovereign cloud builders who want custom CPUs and compatibility with market-leading Nvidia GPUs for AI and HPC. Previously, Nvidia's Grace CPUs were the only processors compatible with Nvidia GPUs for NVLink connectivity.

While Nvidia only mentions Arm as an IP provider, Arm also gains benefits as a developer of its own CPUs aimed at hyperscalers and sovereign organizations. Specifically, Arm gains the ability to compete directly inside Nvidia-based systems. With native NVLink Fusion integration, future Arm-designed server CPUs can compete head-to-head with both Nvidia's Grace and Vera, as well as Intel Xeon, in systems where Nvidia GPUs are the central compute element. With NVLink Fusion, Arm CPUs can become first-class participants in rack-scale NVLink solutions, assuming that Nvidia allows this to happen, which is not guaranteed.

Also, NVLink Fusion support strengthens Arm's position as an ISA licensor, as it makes the Arm architecture inherently more attractive to hyperscalers and chip designers who want custom CPUs tightly integrated with Nvidia GPUs. By ensuring that Arm-based CPU designs can work with Nvidia GPUs using the coherent NVLink fabric — rather than being limited to PCIe — Arm gains ecosystem gravity and 'future-proof' relevance that competing ISAs like x86 and RISC-V cannot match today. For sure, this poses risks to both AMD and Intel as the former is barely interested in supporting NVLink, while the latter is years away from building custom NVLink-supporting Xeon CPUs for Nvidia's rack-scale systems. Then again, we have to keep in mind chip development cycles and other factors here, as by the time Arm-based CPUs with NVLink are ready, Intel's custom Xeon CPUs will be ready as well.

Arm's support for NVLink Fusion benefits Nvidia by massively expanding the pool of CPUs that can serve natively in Nvidia-centric AI systems using NVLink, without Nvidia having to build all those CPUs itself. By enabling Arm licensees — such as Google, Meta, and Microsoft — to integrate NVLink directly into their SoCs, Nvidia ensures that future Arm-based processors will be either architected around Nvidia GPUs, or at least compatible with them. On the one hand, this could reduce the appeal of open alternatives like UALink; on the other hand reduce the appeal of AI accelerators from companies like AMD, Broadcom, and Tenstorrent in general.

As an added bonus, it also strengthens Nvidia's position in sovereign AI projects that use Arm CPUs (at least in the next few years): governments and cloud providers that want custom Arm CPUs for control-plane or data-loading tasks can now adopt them without leaving Nvidia's GPUs.

All in all, Arm's addition to the NVLink ecosystem is a win for both Arm, Nvidia, and a bunch of their partners, but could pose great risks for AMD, Intel, and Broadcom.

China-based Mini PC manufacturer Minisforum has just globally launched the MS-R1, the company's first foray into the ARM market, with a compact yet powerful device. Building on the interest in Nvidia's DGX Spark, the ongoing AI boom has made desktop-class ARM systems a hot topic, and Minisforum's MS-R1 differentiates itself by including support for dedicated GPUs and a UEFI.

The MS-R1 is based on Cixin's P1 SoC, using a rebranded variant of it called the CD8180. It's a 12-core SoC, packing 8x Cortex-A720 and 4x Cortex‑A520 cores, alongside an Immortalis-G720 MC10 integrated GPU. It features up to 64 GB of error-correcting LPPDR5 RAM, running at 5500 MT/s. Minisforum is quoting up to 45 TOPS of AI compute, with 28.8 coming from the NPU.

All that power doesn't require a lot of space, as the MS-R1 is only 1.7-liters in size and should run quietly thanks to its 28W max TDP. The 12-core SoC can boost up to 2.6 GHz, but looks to be fairlyweak, even compared to a last-gen Snapdragon X Elite. For connectivity, you get Bluetooth 5.3 and Wi-Fi 6E, plus the IO is packed, with a bunch of fast USB ports, dual 10Gbps Ethernet, and HDMI out.

Despite touting UEFI boot support — which should enable even beginners to run mainstream operating systems like Linux easily — Minisforum does not mention Windows on ARM, so we assume it's not supported. The company says this is a "major breakthrough" for ARM computing, allowing both complex setups and everyday environments to be booted into with just a USB drive.

Again, the MS-R1 has a dedicated PCIe x16 slot for discrete GPUs, making it one of the first and only compact ARM-based solutions to have that capability. Discrete GPUs can massively accelerate AI inference and training, and pair nicely with the ARM cores to offer an efficient yet powerful computer for AI, for much less than a DGX Spark.

Speaking of, Minisforum is asking $503 for the base unit with 32 GB RAM, which does not include any storage. The top-end model with 64 GB RAM and a 1 TB SSD is $695.90, and you do get an expandable (up to 8 TB) M.2 slot. It looks like the NAND and DRAM shortage has already hit Minisforum as well. These are early-bird discounted prices that will go up after the initial batch sells out (which it already has) showing just how much in demand desktop-class ARM is at the moment.

OpenAI is reportedly working with SoftBank-owned Arm on a new CPU to complement the custom AI accelerator it is co-developing with Broadcom. The collaboration, first reported by The Information, would see Arm design a server-class CPU that anchors OpenAI’s next-generation AI racks, potentially representing one of Arm’s biggest steps into the data center market to date.

The chip in question is OpenAI’s in-house AI accelerator, part of plans announced on October 13 to deploy custom AI accelerators and rack systems in collaboration with Broadcom. The SoC, specialized for inference workloads, is expected to enter production in late 2026 and scale up to support roughly 10 gigawatts of compute capacity between 2026 and 2029. The Broadcom accelerator, said to be fabricated by TSMC, has been in development for roughly 18 months.

According to The Information, Arm’s new role goes well beyond supplying architectural blueprints. The company has recently started designing and manufacturing its own CPUs rather than just licensing cores to partners, and sees the OpenAI contract as a chance to expand its server ambitions. People familiar with the discussions told the outlet that OpenAI could use the Arm-designed CPU not only with its Broadcom chip, but also with systems from Nvidia and AMD.

The potential revenue from OpenAI’s CPU program could reach into the billions, the report also said, representing a major windfall for SoftBank, which owns nearly 90% of Arm and has borrowed heavily against its stake. SoftBank has also pledged to invest tens of billions of dollars into OpenAI’s data center build-out and to buy AI technology from the startup to help accelerate Arm’s own chip development cycle.

Together with earlier agreements with Nvidia and AMD, OpenAI says its chip programs now total as much as 26GW of planned data center capacity. If successful, OpenAI’s custom chip deployments could reach a total installed base that analysts estimate could cost more than $1 trillion in construction and equipment in tandem with its Nvidia and AMD purchases.

The OpenAI–Broadcom chip could also give the ChatGPT developer more leverage in pricing talks with Nvidia, whose H100 and forthcoming Blackwell GPUs still dominate the AI training market. If Broadcom and TSMC can scale production, OpenAI’s inference chips may offer a partial hedge against the tight GPU supply that has constrained AI labs for much of the past year.

SoftBank is arranging a new $5 billion margin loan to invest in OpenAI and use its shares in Arm Holdings as collateral, reports Bloomberg. The move will enable the company to increase its stake in OpenAI and/or invest in the company's infrastructure, while taking the company's total borrowing against Arm shares to $18.5 billion.

The Japanese investment company is said to be negotiating with several international banks to finalize the deal, which will be collateralized by Arm stock. The company had already tapped $13.5 billion in margin loans from Arm shares by March 2025, with $5 billion still unused at that time, which gives it some flexibility to expand its credit base further.

The approach is not uncommon for SoftBank, as it is usual practice for the company to draw cash from its most valuable asset rather than sell equity. Arm's shares were up 38% this year, which gives Softbank more collateral room.

SoftBank's campaign to become a dominant force in the AI sector has become increasingly costly. The company joined the $500 billion Stargate project with OpenAI and Oracle to build massive U.S. data center infrastructure early this year, pledged up to $30 billion to OpenAI, purchased ABB's robotics business for $5.4 billion, took over Graphcore for an undisclosed sum, and agreed to purchase Ampere Computing for $6.5 billion.

Bloomberg Intelligence's Sharon Chen estimates SoftBank’s total financing needs could surpass $30 billion, considering the possible purchase of Ampere Computing and other ventures.

The latest borrowing follows a series of large financings tied to the AI push. SoftBank previously arranged $8 billion in margin loans ahead of Arm's 2023 IPO through lenders such as JPMorgan Chase, Barclays, BNP Paribas, Crédit Agricole, and Goldman Sachs, and later secured a $15 billion one-year loan to fund AI projects in the U.S. Together, these moves have transformed Arm's stock into the group's main source of liquidity.

The analyst warned that these commitments could push the group close to its 25 % loan-to-value threshold, raising concerns about its credit profile. While SoftBank may use asset sales or structured financing to offset risk, its growing demands for cash highlight how heavily its strategy depends on the market value of Arm and the continued strength of AI-related equities. On the one hand, this puts SoftBank at the center of the AI boom, but on the other hand, it also places it at the frontier of its potential instability.

This week, Qualcomm secured a final legal win against Arm in a high-profile licensing dispute around CPU cores developed by Nuvia, which Qualcomm acquired in 2021. A U.S. District Court has now rejected all of Arm's claims, affirming Qualcomm's right to use Nuvia's technology under its architecture license agreement (ALA). Without any doubt, this is a big legal win for Qualcomm. However, this win could have reverberations throughout the entire industry, as it could set a precedent that allows architecture licensees to develop custom CPUs in whichever way they deem necessary.

Origins

The legal conflict between Arm and Qualcomm began shortly after Qualcomm acquired Nuvia, a startup focused on high-performance Arm-compatible CPU designs for the HPC segment, and promised to build CPUs for consumer PCs based on Nuvia's IP.

Arm claimed that the transfer and use of Nuvia's designs under Qualcomm's umbrella required renegotiation of Nuvia's original architecture license agreement (ALA), as Nuvia had only planned to use the cores in the data center space, whereas Qualcomm had much broader plans for the IP. Qualcomm, on the other hand, maintained that its own existing ALA was sufficient to incorporate Nuvia's work and continue development and deployment of custom cores based on the Arm instruction set architecture. Arm then revoked Qualcomm's ALA in October 2024, citing a violation of the agreement for not renegotiating it after acquiring Nuvia.

Arm opposed Qualcomm's use of Nuvia’s CPU IP across broader market segments, which would include everything from entry-level consumer devices to high-end servers, because it threatened Arm's control over licensing boundaries and its ability to extract additional royalties. Arm likely viewed the transfer of that IP to Qualcomm, followed by its broad deployment across various markets, as a breach of scope, as the original Nuvia agreement had narrower terms.

Furthermore, Qualcomm used Arm's off-the-shelf Cortex cores under a technology license agreement (TLA) license. This provides Arm more control, as well as a per-core license. This clearly hurt Arm's bottom line, as ALA royalties are considerably lower than TLA royalties.

If Qualcomm can freely use Nuvia's custom Arm v8 cores across segments under an existing ALA, it would theoretically enable other licensees to do the same thing. Companies would be able to acquire CPU startups that receive an ALA on certain conditions, then roll them into their own ALA if they have one, and sidestep core-level royalties.

By allowing Qualcomm to integrate Nuvia's custom cores under an existing architecture license, Arm risked weakening its tiered licensing model and losing leverage over other major partners that considered similar custom silicon strategies. Perhaps Arm also viewed Qualcomm's move as a competitive threat to its own Neoverse core roadmap, Neoverse CSS roadmap, and ultimately its processor or custom processor roadmap. Therefore, blocking this reuse was an attempt to preserve both licensing revenue and product relevance in high-performance markets.

In December 2024, a jury sided unanimously with Qualcomm, finding no breach of the Nuvia ALA and confirming the legitimacy of Qualcomm's use of the technology. On September 30, 2025, the U.S. District Court in Delaware reaffirmed that position, dismissing Arm's final remaining claim and denying a request for a retrial. The judgment, now final, delivers Qualcomm a complete legal triumph and blocks Arm from any further recourse in the case. Furthermore, Qualcomm's countersuit against Arm is still pending and expected to go to trial in March 2026.

A green light for Qualcomm's expansion

When Nuvia first introduced its Oryon/Phoenix processor core in 2020, it demonstrated considerably higher performance efficiency than Apple's A13, AMD's Zen 2, and Intel's Sunny Cove, as well as other relevant CPUs over time. After several delays, when the Oryon-based Snapdragon X Elite CPUs hit the market in 2024, they demonstrated competitive performance. However, the Snapdragon X2 Elite processors look considerably more promising, and the launch of the next generation coincides with the legal win.

"This decisive legal victory is monumental for Qualcomm, clearing the path to fully integrate and scale the acquired Nuvia assets under Qualcomm's existing Architecture License Agreement (ALA)," said Neil Shah, VP of Research at Counterpoint Research. "This win provides Qualcomm with significant momentum, enabling them to accelerate the deployment of custom Nuvia-based CPU cores across a much broader spectrum of applications from PCs, smartphones, and automotive to high-performance computing domains like AI servers and even humanoid robotics."

With a better CPU and presumably system-on-chip design, ALA licensing fees, and without legal obstacles and risks, the company can now scale Nuvia cores for a wide range of client (and eventually data center) product categories, including automotive, PCs, and smartphones.

Furthermore, now that it is perfectly legal for Qualcomm to use Nuvia-designed cores, PC OEMs may be more willing to integrate Snapdragon X2 Elite CPUs into their systems, especially considering that Windows on Arm is attempting to gain more traction. Ultimately, this supports Qualcomm’s efforts to challenge the x86 incumbents in the laptop market.

Keeping in mind that Qualcomm also has competitive neural processing units (NPUs) for AI, the company may also introduce new product categories that take advantage of the highly efficient Oryon CPUs and sophisticated NPUs.

And Arm-based custom silicon

Qualcomm's sweeping legal victory against Arm marks an important moment in the semiconductor industry, reaffirming the rights of architecture licensees to develop and scale custom CPU designs without renegotiating terms, which somewhat reshapes the balance of power between IP holders and their partners.

One of the most immediate effects of the ruling is a renewed sense of legal clarity for holders of Arm architecture licenses. Qualcomm’s position that an ALA provides broad design rights, including the freedom to integrate acquired IP, has now been upheld at the highest level. This has implications not just for Qualcomm, but for other major licensees. Companies such as Amazon, Broadcom, Google, MediaTek, and Nvidia have all used custom or semi-custom Arm-based cores in their products.

The precedent set here assures licensees that they can pursue internal development and even acquire CPU startups without facing retroactive restrictions or new licensing demands from Arm. In an era where companies are increasingly seeking to differentiate their silicon at the architectural level, this kind of legal certainty is vital.

Arguably, the verdict also preserves the original intent of the ALA license: to empower chipmakers to innovate freely within the Arm ecosystem and retain Arm as the owner of the ISA. In particular, under a standard ALA license with Arm, licensees can design their own CPU microarchitecture from scratch that runs a specific Arm ISA (e.g. Arm v9), and add internal optimizations, such as specialized execution pipelines, custom data paths, or micro-op fusion techniques, as long as these do not break ISA compatibility. They could even implement custom accelerators or extensions, provided they do not interfere with standard ISA behavior and are not exposed to software that expects strict Arm compliance.

They still will not get as much freedom as they get with RISC-V, as the process of adding to the Arm ISA is complex and takes years. However, companies like Apple canned its custom AMX instruction set with the M4 chipset in favor of ARM's SME.

What about Arm and Qualcomm's relationship?

While Qualcomm has prevailed in the legal fight with Arm, the impact on the latter has yet to be determined. Arm still controls the dominant ISA used across mobile and embedded platforms worldwide, which is gradually expanding into the PC space. Its core IP, software stack, and ecosystem partnerships remain extremely valuable and widely used. But this case has damaged its reputation with one of its largest customers, specifically Qualcomm, which has long been a flagship licensee.

"While Arm's decision to pursue legal action was an understandable measure to protect its business, it was an unfortunate necessity that strained its relationship with a premier customer and partner," Shah said. "Despite this, Arm unequivocally retains the industry's leading low-power architecture, coupled with a robust software and tools ecosystem for computing. Moving forward, Arm must seize the opportunity to mend fences and rebuild trust with Qualcomm.

To avoid a further erosion of trust, Arm may need to shift away from legal enforcement and instead re-establish its relationship with Qualcomm. But the path toward reconciliation remains rocky, as Arm itself is moving toward designing its own CPUs. In fact, this move could make other partners gravitate towards custom designs or alternative architectures.

There's also the question of Qualcomm's countersuit against Arm, which accuses the British company of contract breaches and customer interference. Depending on how that case unfolds in March 2026, the relationship could deteriorate further.

This ruling could define the future of Arm

Qualcomm, along with its subsidiary Nuvia, have finally won the legal battle against Arm Holdings, which had accused the chip-maker of violating license agreements tied to its chip designs. A U.S. District Court judge in Delaware ruled that neither Qualcomm nor its subsidiary breached any of Arm’s architecture license agreement (ALA), dismissing the lone remaining claim in the case and also rejecting Arm's request for a new trial.

The latest ruling follows Qualcomm’s trial victory from December 2024, where the company was found innocent. However, at the time, the jury was unable to agree on whether Nuvia violated its licensing terms with Arm.

The dispute centered on Qualcomm’s use of Oryon cores for its Snapdragon X range of client processors based on Arm’s v8 architecture, created by Nuvia for server-grade chips.

Arm claimed that Qualcomm was supposed to renegotiate licensing terms following its acquisition of Nuvia. Additionally, Arm demanded that the designs be scrapped for allegedly breaching Nuvia’s original Arm licenses. Qualcomm, however, maintained that its existing Architecture License Agreement (ALA) for Arm’s instruction set architecture already extended to designs developed by its subsidiaries, including Nuvia.

Following the verdict, Ann Chaplin, General Counsel and Corporate Secretary at Qualcomm, said “With the Court’s decision today, Qualcomm and its subsidiary Nuvia have achieved a full victory. This decision follows Qualcomm’s December 2024 jury trial win and is a full and final judgment in Qualcomm’s favor. Our right to innovate prevailed in this case and we hope Arm will return to fair and competitive practices in dealing with the Arm ecosystem.”

It is interesting to note that the Qualcomm Oryon general-purpose cores found inside Snapdragon X processors are based on Arm’s Armv8 instruction set architecture (ISA). However, it incorporates “one percent or less” of Arm’s own technology according to Gerard Williams III, one of the lead developers of Oryon and former Apple chip designer.

Williams co-founded Nuvia in 2019 with the aim of building high-performance, energy-efficient custom CPU cores for datacenters, known as Phoenix. To do so, the company secured two licenses from Arm, including a Technology License Agreement (TLA) to modify existing cores and an Architecture License Agreement (ALA) to design custom ones. Since Nuvia’s strategy was to pursue custom designs from the outset, the team developed its cores from scratch, relying minimally on Arm’s physical IP.

Qualcomm has stated that a separate lawsuit against Arm is still ongoing. The case involves claims of breach of contract, interference with customer relationships, and conduct aimed at hindering innovation while promoting Arm’s own products over those of long-standing partners. The company added that it expects the trial to begin in March 2026.

Yesterday, Nvidia and Intel lifted the curtain on a historic collaboration that will see the two chipmakers jointly develop a myriad of CPU and GPU products. While future solutions like the "Intel x86 RTX SoC" were the focus of the announcement, some clarification was also shed on existing projects. Chief among these was Nvidia CEO Jensen Huang saying that the upcoming, long-rumored N1 SoC is essentially the same as the GB10 Superchip that's been out for a while.

For some context, Nvidia has never officially unveiled the N1/N1X SoCs, but speculation sparked from CES 2025's announcement of Project DIGITS, where the company revealed its collaboration with MediaTek. From that came the GB10 "Superchip," which is part of the company's DGX Spark lineup, and multiple vendors have already released their iterations of it. The GB10 is aimed squarely at AI workloads, offering supercomputer-like performance at home. It includes a 20-core ARM-based CPU developed in conjunction with MediaTek, along with a powerful Blackwell-based GPU chiplet.

The N1 SoC shares the same specs, at least according to previous leaks and rumors, featuring 6,144 CUDA cores for its GPU - same as the desktop RTX 5070 - and a 20-core CPU split across two clusters, built using Nvidia's Grace architecture. Back in July, we saw a Geekbench score surface for the N1X, which allegedly confirmed these specs, giving credence to the fact that GB10 and N1 are intrinsically tied. Of course, just because two products are closely linked to each other doesn't mean they're the same, but all signs pointed toward identical chips being used across the board.

That notion has just been legitimized by Jensen Huang, who said the following in a webcast last night, "We also have a new ARM product that's called N1. And that product is - that processor is going to go into the DGX Spark and many other versions of products like that. And so we're super excited about the ARM road map, and this doesn't affect any of that."

According to Nvidia's CEO, the silicon powering the GB10 — which itself is what powers DGX Spark — is identical to the N1/N1X SoC. Especially the part about "many other versions" confirms that N1 could simply be a slightly lower-binned version of the full-fat GB10. After all, the latter is meant for client devices like laptops and desktops, whereas the GB10 targets professionals. The distinction matters because N1 represents Nvidia's first serious attempt at taking their in-house CPU cores mainstream (following Tegra).

Engineer and origami artist Bogdan Ionescu, AKA BogdanTheGeek, has created a web server that runs on a disposable vape. Inspired by his growing collection of these disposables, and prior work he had done on semihosting on any Arm CPU using “a few lines of code,” Bogdan had a lightbulb moment and decided to host “a web server on a vape,” thus creating the VapeServer.

Disposable vapes can contain a surprising amount of computing power/components. Bogdan had been collecting discarded units for ‘future projects’ for a couple of years, with eyes on reusing the batteries. However, he recently became aware of “fancier” units that pack more advanced ICs and microcontrollers. They didn’t just contain PCBs with unknown ‘blob chips.’ He found some with more advanced microcontrollers.

Bogdan says one of the fancier units he disassembled contained an IC marked ‘PUYA C642F15.’ This sparked some research, and the engineer determined that this was actually a PY32F002B, which has the following specs:

• 24 MHz Arm Cortex M0+ processor
• 24KB of Flash Storage
• 3KB of Static RAM
• a few peripheral interfaces

Those are not stellar specs, and perhaps “about 100x slower” than a 10-year-old mobile, by our hero’s estimation. For web serving, though, Bogdan thought the PY32F002B-powered ex-vape with USB-C port could be leveraged to make a “blazingly fast” device.

The method, in brief, would see Bogdan emulate a dial-up modem on the microcontroller using SLIP (Serial Line Internet Protocol) over the USB serial connection. Linux ‘slattach’ and ‘socat’ utilities enabled IP packet transmission, and then the microcontroller leveraged the compact uIP stack to communicate via TCP/IP and set up a web server. The web page, a copy of the linked blog post, just about fit in the remaining 20KB of flash on the PY32F002B.

Cutting page load times from 20s to 160ms

Microsoft has announced an update for its Xbox app for Arm-based Windows 11 devices, allowing users to download and play games natively. Arm users, including those on Qualcomm’s Snapdragon X-powered machines like the Surface Laptop 7, were previously limited to streaming titles through Xbox Cloud Gaming. The latest preview update will now allow select compatible games from the Xbox app to be installed and played locally on these devices.

To access the new functionality, users must be part of both the Windows Insider and Xbox Insider programs. Enrolled Arm-based Windows 11 PCs in the PC Gaming Preview will receive the updated Xbox PC app (version 2508.1001.27.0 or higher) via the Xbox Insider Hub. Those who are not enrolled can join by downloading the Xbox Insider Hub from the Microsoft Store. Simply sign in with your Microsoft account, navigate to the “PC Gaming” option under “Previews,” and select “Join.”

Additionally, Microsoft mentioned in its blog post that it is working on new features to ensure game compatibility on Arm-based Windows PCs and expand the number of playable titles in the coming months. This has been one of the biggest reasons why gaming on Windows PCs with Arm processors has long been a major headache. Since a majority of PC games are designed for x86 chips from Intel or AMD, Arm devices rely on emulation. This usually leads to a drop in performance and sometimes restricts games from launching at all, especially those with strict anti-cheat mechanisms.

On top of that, Arm-based SoCs (System on Chip) typically rely on integrated GPUs, which aren’t exactly performance powerhouses. Yes, Apple has shown promising improvements in gaming on its latest M4 chips, but they’re still a long way from matching the capabilities of discrete GPUs or even the latest integrated graphics solutions from AMD and Intel.

Interestingly, Microsoft’s latest push to improve native gaming on Arm could also be an indication of bigger hardware plans in the future. We all know Nvidia is co-developing an Arm-based SoC for Windows PCs with MediaTek, which many expected to debut at Computex 2025 before its reported delay to late 2026. Last month, an alleged engineering sample of the N1X Arm-based SoC was spotted on Geekbench, suggesting a 20-core CPU and a Blackwell GPU with 48 streaming multiprocessors and 6,144 CUDA cores, which matches the desktop RTX 5070.

If Microsoft is indeed laying the groundwork for improved game compatibility on Arm, it could position Windows to fully support such upcoming hardware from Nvidia, as well as systems from OEMs like Asus, Dell, and Lenovo that are reportedly developing devices around the new chip.

Arm is a company that has existed on the periphery of wholesale dominance for decades. ARM intellectual property has been a part of over 90% of all mobile phones shipped since 2005, currently holding a market share of over 99%, boasting over 250 billion Arm chips shipped since the company's inception. The world’s fastest supercomputer from 2020 to 2022, Fugaku, also ran on ARM CPU cores. The world’s top companies, including Amazon, Google, and Nvidia, have close partnerships with the company and its tech.

And if Arm has anything to say about it, its dominance will only increase over the next decade. The company’s CEO recently claimed that Arm will capture 50% of the Windows PC market by 2029, as well as wishing to penetrate 50% of the data center market by the end of 2025.

The company has made it clear that its scope extends beyond the cellphone, now seeking to become a foundational player in all facets of the computing market. Let’s take a look at the details of this plan, and how Arm is expecting to mount a takeover to surpass Intel, AMD, Nvidia, and the open-standard of RISC-V.

An ARM for Arm — forging an empire without building chips

A crucial note before continuing to write about Arm: Arm is the name of the company and its products, while ARM is the name for its architecture family.

Arm began as Advanced RISC Machines, Ltd., a startup spun off from computer firm Acorn, and funded by Apple. Arm was first known for its invention of the ARM1 Instruction Set Architecture (ISA) and processor, a reduced instruction set computer (RISC) for PCs made in the late 80s. The chip first debuted in the Acorn Archimedes, the first RISC-based PC.

Original developer Sophie Wilson settled on a RISC architecture purely out of budget constraints, with ARM or other RISC ISAs having less power consumption and headroom than ISAs like x86 (the standard for Intel and AMD silicon today). ARM chips were popularized in mobile devices due to their high efficiency and relatively low power draw. The Apple Newton PDA, released in 1993, also ran on an ARM6-based CPU.

Arm’s growth in the market can be attributed to two factors: the highly adaptable nature of the ARM architecture and its unique licensing model. Rather than designing and producing silicon, Arm would license its instruction set and CPU core designs to companies in exchange for an upfront fee and proceeds on eventual silicon sales.

These other companies then receive assistance in turning Arm cores into full CPUs, making Arm a necessary partner with all of its clients. This business model is the same that Arm employs today, allowing it to be the top name in cellphone CPUs without ever producing chips itself.

Arm-designed CPU cores have proliferated widely across the tech landscape. Arm claims it has created the “most pervasive compute footprint in history”, hitting a lifetime 250 billion cores shipped in April of this year.

The company could be content with a win in the relatively young ARMv9 architecture, the newest architecture from the company that generates much higher royalties than the previous-gen ARMv8. But a key window into Arm’s mindset as a company comes from this year’s Arm executive summit at Computex, where SVP and GM Chris Bergey highlighted a graph that differentiates Arm from the “legacy PC” world.

Arm’s current goal, is to overthrow the “legacy PC” giants of Intel, Apple, and AMD in the PC and data center space. But, how are its efforts faring so far?

Arm and Windows — an uphill battle

As entrenched as ARM cores are in the mobile world, Windows and macOS are in the PC operating system status quo. Aside from Linux—which, while commonly thought to be a superior OS, is unlikely to reach mainstream appeal anytime soon—all commercially sold PCs are necessarily forced into Windows or Mac for their operating system. ARM-based chips have been part of Apple’s macOS computers in the Apple M-series silicon chips since 2020, but Windows remains a tough nut to crack.

Windows has primarily run on x86-based processors for decades, with x86 being the complex ISA used in Intel and AMD CPUs for just as long. As a result, most of the programs on Windows are programmed to be run on an x86 architecture, and are, in some way, incompatible with ARM-based architecture.

The key to unlocking market penetration among PC users for ARM chips is a bespoke operating system. This effort failed once before with Windows RT, which some may remember as an OS that looked like Windows 8, with a shockingly small selection of compatible programs. Today, Windows on Arm is thankfully a mostly faithful recreation of the x86-based version of Windows 11. This is achieved through extensive emulation, but it still requires ARM-based computers to run it.

Arm’s partnership with Qualcomm birthed the Snapdragon X-series of desktop processors, a wave of ARM-based CPUs in Windows laptops. Launched under the “Copilot+” badge, the laptops were heavily marketed by Microsoft, Arm, and Qualcomm as the future of Windows and of laptops altogether.

The focus on power efficiency with ARM, plus tons of AI-based features and widgets, was touted as the next biggest development in the PC world. CEOs began making absurdly bullish claims, including that Arm would make up 50% of the Windows PC landscape in five years, or that some OEMs expect Snapdragon chips to make up 60% of their sales within three years. In Q3 2024, penetration of the Snapdragon X Elite processors only reached around 0.8% of the market.

However, this hasn’t stopped corporate interests from cherry-picking favorable statistics to maintain hype—Qualcomm recently celebrated the fact that its ARM devices made up 10% of the consumer Windows PC market for devices priced $800 and above.

Part of the reason for disappointing sales may be that ARM’s lead over x86-based chips was not as great as hoped, particularly after Intel and AMD responded with more efficient chips of their own. While our in-house battery tests of Qualcomm's Snapdragon X Elite Copilot+ laptops showed an impressive 15 hours of battery life, the competition from Intel wasn’t far behind at over 13 hours.

This certainly wasn’t the bulletproof efficiency for Snapdragon X that Arm and Qualcomm were hoping for. However, Qualcomm isn't the only company eyeing up ARM CPUs to ship in 2025, as Nvidia has expectations of its own.

Nvidia & Arm: DGX Spark & N1 / N1X chips loom

Nvidia’s DGX Spark and DGX Station AI workstations, originally known as Project Digits, will be powered by the Nvidia Grace CPU, a design based on Arm’s Cortex-X and Cortex-A CPU cores. The AI-focused workstation stirred up an impressive bit of hype on its announcement at CES 2025 and technical showcase at Computex 2025.

In a press Q&A at Computex 2025, Nvidia CEO Jensen Huang expressed how useful the DGX Spark might be for developers, who currently rely on the cloud for AI-based computing.

"Every single developer can go out and just get one, and just put it next to their desk. You can develop on here, and you want to now scale it out, or test it out on large data sets, it's just like one pull-down menu, point it at a cloud. Exactly the same thing runs there. And so, this is really an ideal AI developer environment," Huang said.

These systems have the potential to serve AI developers and researchers as systems that can handle larger local AI workloads, thanks to large memory pools. Nvidia has heaps of third parties on board, including Asus, Dell, MSI, and more, to ship the DGX Spark, but the chips have yet to arrive on the market. Additionally, the DGX Station has no solid release date, though targeted for a 2025 release.

Nvidia has high expectations for the DGX systems, especially as the company is enjoying a lofty market cap increase, thanks to its efforts in data center AI. But the DGX systems are not the only chip that Nvidia is looking to ship, with additional ARM-based efforts on the horizon.

Nvidia was expected to be more of a ray of hope for Windows on Arm this year, with one of the most anticipated Computex announcements being the MediaTek/Nvidia collaboration “N1” and "N1X" CPUs. This Windows on Arm CPU was considered a lock for a Computex announcement, but an alleged internal delay to 2026 means that the chips have yet to break cover, at least officially. This delay was reportedly due to efforts to align with Microsoft OS updates, as well as to refining chip-level issues. As of the time of writing, the N1X is alleged to begin mass production in October 2026.

However, the N1X was purportedly spotted in the FurMark benchmark database and scored 4,286 points on the 720p stress test, averaging around 71 FPS. This kind of performance lands the chip at around the same level as an RTX 2060. But this might just be a result of an early engineering sample, for now.

Where are second-gen Snapdragon X products?

So, with Nvidia's consumer efforts sidelined until next year, Qualcomm is currently the only provider of Windows on Arm chips in 2025. But, there have not been any iterative chip releases since last year. Snapdragon X’s second generation was not announced at Computex 2025, meaning the 2nd-gen will likely not launch until 2026 at the soonest.

However, it's expected that the Snapdragon X2 Elite will feature up to 50% more CPU cores when compared to the previous generation, sporting the Oryon V3 CPU architecture. These CPUs are also rumored to be pointing toward the desktop or server market, which stems from Qualcomm testing the SC8480XP chip alongside a 120mm AiO, marking a significant generational change. Qualcomm CEO Cristiano Amon has also teased that the company plans to bring the chips to new form factors.

Snapdragon Summit is set to take place in September, and we may see the next generation of Snapdragon X chips debut there, alongside potential products. But until then, we're still waiting to hear more.

Arm's data center efficiency goals

Arm’s current strongest segment beyond consumer electronics is in the enterprise space. ARM-based chip designs are projected to make up 50% of all compute shipped in 2025 to the top hyperscalers, including Amazon AWS, Microsoft Azure, Google Cloud, and Alibaba. Each of these firms has its own bespoke ARM-based CPU family, including Amazon’s Graviton, Google’s Axiom, and Alibaba’s Yitian.

Amazon has been a particularly large customer of ARM-based chips for some years now. 50% of AWS’s new CPU stock has been ARM-based Graviton chips since 2021, with more than 90% of the top 1,000 largest AWS EC2 customers currently running Graviton chips. Arm expects its other major clients/partners to follow suit shortly, and for very good reason.

Arm’s Computex 2025 summary claimed that the ARM-powered chips used by the above hyperscalers are up to 40 percent more energy-efficient than other platforms, taking shots at Intel’s Xeon and AMD’s Epyc enterprise lines. The modern data center is a place where energy efficiency is crucial, as power consumption drives skyrocketing operating costs and leads to negative public opinion against AI data centers.

Elon Musk’s Colossus supercomputer, for example, has been accused of illegally polluting the local community with its 30+ portable diesel and methane generators, used to make up the 100+ MW shortfall between what the site needs and what the power grid can give to it.

A 40% increase in energy efficiency, therefore, seems like a critical win for Arm; certainly a more convincing delta than the 11% in power efficiency that Snapdragon X sees over Intel laptops. Arm works with companies to put their names on their bespoke ARM-based chips, resulting in branding wins for hyperscalers and monetary wins for Arm. However, change may be coming on the horizon for Arm that may carry some exceedingly high risk.

In-house chipmaking ambitions

In February, it became widely reported that Arm had seemingly begun recruiting executives and engineers from its licensee customers, looking to build in-house chips. Multiple industry sources allege that Arm recruiters are sending out letters to potential workers, stating Arm’s intent to begin “selling its own silicon, with a focus on driving AI enablement in the data center." Arm had also reportedly already made a deal with Meta to supply the company with its soon-to-come in-house silicon for Meta’s hyperscaling and web hosting endeavor, and a matching deal with TSMC to fabricate the chips.

Arm has so far declined to comment on this claim, which makes a lot of sense. It directly contradicts the testimony of Arm CEO Rene Haas in court against Qualcomm, stating “we don’t build chips” when asked about Arm’s chipmaking ambitions. More than that, it risks alienating Arm from its customers, who would quickly become competitors. Qualcomm and Nvidia both sell ARM-based enterprise CPUs to clients, but if Arm began selling its own silicon, it would suddenly be at odds with its licensees. What’s more, Arm chips would compete against the ARM-based Graviton, Axiom, and Yitian chips licensed by hyperscalers, potentially scaring these clients off of continuing to license from Arm in these custom chips.

Entering the ring as a silicon maker risks losing a large number of top enterprise clients to Intel or AMD, or perhaps even RISC-V. RISC-V is the open-standard counterpart to the privately owned Arm, itself a RISC ISA family that is on the path the becoming mature enough to entertain business from high-end hyperscalers.

RISC-V has emerged as a key part of China’s computing boom, with the Chinese government urging computing companies blacklisted by the U.S. to develop their own chips on the open standard. RISC-V would be attractive as an alternative to ARM-based chips if Arm scares its clients away by turning into a chipmaker.

The risk of Arm losing customers is not as great as it may seem, however. ARM’s key edge over RISC-V is that ARM is not just an architecture; the company also designs CPU cores for license by customers. If a client were to drop Arm, it would then have to design its own CPU cores from the ground up on the RISC-V architecture. This represents a significant R&D cost, as it involves more than just licensing from and collaborating with Arm at a much later stage in development. A company like Amazon, which already uses Graviton CPUs in around 50% of its servers, would not likely start from scratch on RISC-V.

The future of Arm

Arm, like seemingly every other technology company in the modern day, is betting big on AI and its impact on the tech market. Arm GM Chris Bergey described the current iteration of AI as “the single greatest advancement in the history of computing” at Computex this year. With data center expenditures skyrocketing and enterprise hardware quickly progressing, it’s not surprising that a hardware company would make claims like this to sell more products.

Right now, AI is not making a major material difference in the lives of the average adult. As of April 2025, 66% of American adults claim to have never used an LLM chatbot, and of the 33% who have, 61% of them report the experience to have been only somewhat, or not useful. Nevertheless, 83% of businesses claim that AI is a “top priority” for their business plans.

In such a rapidly evolving segment, everyone is looking to make their share of wealth, and Arm is no different. Through data center products, as well as more consumer-facing ones, like Qualcomm's Snapdragon and Nvidia's DGX Spark and DGX Station.

So, we know that Arm is focusing its efforts on shipping AI-ready PCs, whether that be satellite systems for developers or Copilot Plus laptops. With next-generation Snapdragon chips seemingly just around the corner, in addition to another boon from Nvidia's ARM-based CPU efforts, it paints a new picture for the business after a relatively quiet 2025.

If Arm truly wants to be the thing that turns Intel and AMD into “legacy” technology, it needs to stick the landing on its upcoming, theorized big moves into mainstream consumer processors and in-house chipmaking. If it does, Arm could be powering laptops through its Qualcomm and Nvidia collaboration, breaking into the desktop market, and shipping custom chips in the data center.

Should this all come to fruition as the company hopes, it would likely become one of the greatest silicon empires of the modern day.

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

PC enthusiasts are hopeful that Qualcomm has learnt some lessons from its initial Snapdragon X Elite venture. It had a lot of help from Microsoft to spearhead Copilot PCs using the first Oryon architecture processors last June. However, interest in them fell sharply as AMD and Intel responded rather swiftly with their own power-sipping but potent AI/NPU-enhanced laptop processors.

The last time we heard murmurings about the touted SC8480XP was back in March. At the time, we highlighted that 18 cores would increase the CPU core count by 50% vs the first-gen Snapdragon X Elite. Moreover, the newer architecture is expected to deliver its own performance boost.

Rumors also point to the Snapdragon X2 Elite looking to push into the desktop side of the market, or even servers. This notion comes from chatter about Qualcomm testing chips like the SC8480XP with advanced cooling solutions, including a 120mm radiator toting AiO. That would mark quite an intergenerational change.

Moving on to other SoC specs, like onboard RAM, the last we heard, the SC8480XP was being tested in Qualcomm SiP (system in package) format with 48GB RAM on board, as well as a 1TB SSD storage on board. Now it looks like options with up to 64GB of RAM on board may be offered.

New processors expected in September

Following Huawei and Lenovo's lead in developing self-developed silicon within China, Xiaomi is working on its own XRing 01 SoC. This new chip reportedly features standard Arm Cortex cores and a 3nm-grade process node from TSMC. Per HXL, the XRing 01 carries a beefy decacore configuration, and based on now-delisted Geekbench tests of the chip, shared by leaker Jukanlosreve, Xiaomi's alternative appears to be delivering performance comparable to MediaTek's flagship Dimensity 9400 SoC.

Faced with significant restrictions from the U.S. and motivated by potential cost savings compared to alternatives like Qualcomm and MediaTek, Chinese manufacturers are rapidly transitioning towards in-house chip design and manufacturing. Huawei takes this a step further as its latest Kirin X90 SoCs for the Matebook Pro 2025 family is believed to feature custom Arm-based 'Taishan' cores and is reportedly fabbed in China using SMIC's 7nm process.

Xiaomi’s upcoming 15S Pro mobile devices are rumored to be powered by the XRing 01. Leaked specifications suggest the XRing 01 SoC featured a decacore layout, including two Cortex-X925 prime cores at 3.9 GHz, four Cortex A725/X4 cores running at 3.4 GHz, two Cortex A720/A725 cores at 1.89 GHz, and two efficiency-focused Cortex A520 cores at 1.8 GHz.

While mobile SoCs typically don’t employ four distinct core types, the XRing 01 shares this unusual design configuration with Samsung’s Exynos 2400. The dual Prime core setup is similar to Qualcomm’s Snapdragon 8 Elite and Apple’s A18 Pro, though they use custom Arm designs. MediaTek’s Dimensity 9400 is a better comparison, but even that chip sticks to a more conservative octacore layout with just one Prime Cortex-X925 core.

AMD is allegedly working on an Arm-based SoC, codenamed "Sound Wave", in a bid to power Microsoft's Surface laptops next year, claims Kepler via ITHome. Moving away from traditional x86 designs, Sound Wave is reported to feature the Arm ISA and will likely leverage off-the-shelf Cortex cores. Details on exact specifications, availability, and pricing remain under wraps, so it's wise to approach this leak with caution.

Looking beyond its historical Wintel roots, Microsoft has made a clear push towards the WoA (Windows on Arm) platform. This was put into effect with the firm's partnership with Qualcomm, which yielded the Snapdragon X family. This was likely a significant catalyst that motivated Intel to engineer an efficiency-first alternative: Lunar Lake. In fact, even Nvidia is entering the WoA space with its rumored N1 family of SoCs, developed in partnership with MediaTek.

The strong success of the Nintendo Switch, powered by Nvidia hardware, underlines a lucrative market for Arm-based handhelds. These Sound Wave SoCs, if true, could be a foundation for the Steam Deck 2, but I must emphasize this is highly speculative. Microsoft's current-generation Surface Pro 11 and Surface 7 laptops are powered by chips from Qualcomm and Intel. AMD's existing gap in efficiency compared to Snapdragon X, Lunar Lake, and likely soon-to-launch N1 offerings could be bridged with these Arm-based SoCs.

Ampere Computing on Tuesday quietly added several new processors to its AmpereOne family without a formal announcement or press briefings. The quiet release comes after the company was bought by Softbank. The new AmpereOne M CPUs feature a 12-channel DDR5 memory subsystem and are aimed at applications that require more memory capacity as well as bandwidth. The new CPUs feature from 96 to 192 cores and require new motherboards. 

The AmpereOne M CPU family uses a 7228-pin FCLGA socket and includes six processors with 96, 144, 160, and 192 single-threaded custom Armv8.6+ cores operating at up to 3.60 GHz and equipped with a 2MB L2 cache.

The processors also feature 64 MB of system level cache. The key feature of the new CPUs compared to their predecessors is their 12-channel memory subsystem, which supports a maximum of one DIMM per channel and up to 3TB of addressable DDR5-5600 capacity. The memory subsystem is ECC-protected using SECDED and Symbol ECC to make it suitable for cloud datacenter workloads. 

When it comes to power consumption, AmpereOne M processors consume up to 348W, and to keep their power consumption in check, these CPUs support a combination of dynamic voltage and frequency scaling, adaptive voltage control, and fine-grained thermal sensors.

On the I/O front, the processor supports 96 PCIe 5.0 lanes with bifurcation capabilities down to x4 and has 24 device controllers to connect multiple accelerators, SSDs, network cards, and other high-performance components needed in AI and cloud deployments.

Ampere's AmpereOne M processors are still made on TSMC's N5 process technology, just like their predecessors, so the additional memory channels are indeed the key feature of the new CPUs. These processors can process up to 192 threads per socket, which is lower compared to AMD's 192-core EPYC 9965 CPUs, which support simultaneous multi-threading and therefore can process up to 384 threads simultaneously.

But perhaps, the purpose of AmpereOne M is not only to offer support for 3TB of memory for interested parties in the AI space, but rather to set the stage for the company's next-generation AmperOne MX processors that will feature 256 cores and 12 DDR5 memory channels. This upcoming CPU will be made on TSMC's N3 manufacturing process and therefore will add both features and performance efficiency to its list of advantages.

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.

At its Surface event scheduled for May 6th next week, Microsoft is expected to unveil two new compact offerings under its Surface Pro and Surface Laptop families, according to WinFuture. It's important to note that these smaller variants are not a new generation, but rather expand the existing Surface Pro 11 and Surface Laptop 7 product lines. Information about technical details, pricing, and availability remains in the dark, but Microsoft is expected to reveal more on Tuesday.

Microsoft offers a range of Surface devices, designed for various use cases. The Surface Pro family features 2-in-1 tablets that can function both as a tablet and a laptop with the detachable keyboard (purchased separately). The Surface Laptop series, on the other hand, sticks to the basics with a traditional clamshell laptop form factor.

Starting in June 2024, Microsoft revealed the Surface Laptop 7 (15-inch and 13.8-inch) and Surface Pro 11 (13-inch) families, initially equipped with Qualcomm's Arm-based Snapdragon X chips, for the mainstream. Variants with Intel's Lunar Lake followed, targeted towards business users.

12-inch version of the Surface Pro 11

Microsoft is reportedly extending these product lines with compact offerings, starting with a 12-inch version of the Surface Pro 11. In terms of display, we're looking at a PixelSense Flow LCD panel (likely IPS), though the refresh rate has not been specified. Under the hood, it reportedly features the Snapdragon X Plus X1P-42-100 SoC with 8 Oryon cores, along with a 45 TOPS capable Hexagon-based NPU.

The SoC is connected with 16GB of LPDDR5x memory (likely 8448 MT/s), which is non-expandable. The laptop is expected to be available in 256GB and 512GB storage capacities, supplied by a UFS-based solution, which, unlike SSDs, is not upgradable. Microsoft claims 16 hours of local video playback, but real-world usage will vary. The leaked press slides only mention two USB Type-C ports for data transfer, charging, and hooking up external displays.

13-inch version of the Surface Laptop 7

The first Arm processor was created in Cambridge in 1985. From humble origins and with limited resources, the ARM1 was designed to power the new Acorn Archimedes computer systems. The ARM1 featured 6,000 gates, while modern Arm processors feature more than 100 million. Similarly, the popularity of the "exceptionally power-efficient" Arm architecture has scaled incredible heights. At the latest count there have been more than 250 billion Arm chips shipped, in devices spanning sensors, and smartphones, to datacenters and beyond.

From the ARM Ltd turkey barn

Acorn wanted to create an ambitious follow-up to the BBC Micro, which used an 8-bit processor with a 16-bit address bus, namely the MOS Technology 6502 microprocessor. Just two engineers, Sophie Wilson and Steve Furber, were tasked with this job.

Arm admits that the signature quality of its chips – their energy efficiency – was driven out of frugal necessity. Apparently, the firm couldn't afford ceramic chip packaging at the time of the processor's inception. Thus, the designers had to be very strict about efficiency, so plastic packaging would be acceptable.

The new processor ended up being a pioneering RISC (Reduced Instruction Set Computing) chip, featuring just 6,000 gates and built using 3μm technology. Arm says the result was a "low power, fast, and incredibly energy efficient," processor. Nowadays, Arm chips feature over 100 million gates with graphics acceleration, and multi-core capabilities, and some are built using the latest 3nm fabrication technologies.

Momentum beyond the home computer niche

Arm chips first made waves outside the UK home and school computing niche in the 1990s. The first notable outside interest came when Apple chose an ARM chip to power its ill-fated Newton MessagePad handheld (1993). Nevertheless, this venture opened other device-makers' eyes to ARM chips.

Towards the end of the 1990s ARM chip sales growth would become explosive as we entered the mobile phone era. Nokia's 6110 was probably the key device to illustrate the promise of the ARM architecture, and it went on to become one of the best-selling mobiles ever.

Thankfully for Arm, its advances in processor generations kept pace with the move from dumb- to smartphones – and beyond, propelling popularity at a breakneck pace.

Today, and the future

More than 250 billion Arm chips have shipped, by the latest count, while the world's population is approximately 110 billion. In its anniversary blog post, Arm takes the mathematical opportunity to quip that "there are literally more ARMs than arms in the world."

Arm Holdings hopes to increase its share of the global data center CPU market from 15% to 50% by the end of 2025. Mohamed Awad, senior vice president of infrastructure at Arm, made the claim in an interview with Reuters. The company pins its hopes primarily on AI servers, so consider offerings like Nvidia's GB200 and GB300 machines, custom silicon from large cloud service providers, and Ampere Computing-based systems.

Most servers today run AMD's EPYC processors or Intel's Xeon CPUs that rely on the x86 instruction set architecture, as there is more data center-grade software for x86. However, the situation is changing, and Arm says that some server programs are now developed for Arm-based processors first and then ported to x86. Google and Microsoft have also started designing data center processors with Arm's technology, although their projects are at an earlier stage compared to Amazon.

While x86 dominates the server market and will likely continue for a while, Arm adoption is growing. Arm is used by Amazon Web Services for its Graviton CPUs, which are used for many of its instances instead of AMD's or Intel's processors. Half of the processors used by AWS are indeed Arm-based Graviton CPUs.

In addition, Ampere Computing offers Arm-based CPUs for data centers. Ampere is a chip designer backed initially by Oracle but now owned by SoftBank (which also happens to own Arm Holdings).

In addition to AWS and Ampere, Nvidia is emerging as a major backer of Arm in the datacenter space. The company's Grace CPUs with 144 Arm Neoverse V2 cores power GB200 and GB300 AI servers will likely become popular with large cloud service providers.

But Arm pins its hopes not only on AWS, Ampere, and Nvidia. The company also offers compute subsystems (CSS) based on its Neoverse cores, enabling chipmakers to build their data center-grade CPUs relatively easily. Furthermore, Arm is reportedly developing its own CPUs for large cloud service providers, such as Meta. These CPUs have yet to gain market share, though if Meta deploys them in volume, they will inevitably control a significant part of the server CPU market as Meta is one of the major users of servers globally.

3/24/2025 Update: We have updated the headline and article text to reflect the fact that ASR plugins for Unreal Engine 5.3, 5.4, and 5.5 are already available. Unity plugins are promised later this year.

Months after the announcement, Arm is finally making its ASR (Accuracy Super Resolution) upscaler open to more developers to integrate into their games. At the GDC (Game Developer Conference) 2025, Arm promised pre-built and easy-to-integrate plugins for games developed using the Unity game engine by the end of the year. According to a company representative, a plugin for recent Unreal Engine iterations is already available on Epic’s Fab store. Arm's ASR fills a void in the Android gaming landscape, offering an open-source upscaling solution that was previously absent, thus lagging MetalFX in Apple's dugout.

It's important to understand that while Arm chips are traditionally found in smartphones, they've started to make their way to laptops like Qualcomm's Snapdragon X series SoCs. So it's not far-fetched to believe this technology will support those chips as well. ASR is built on AMD's FidelityFX Super Resolution 2 technology, and at its initial unveiling, Arm claimed 53% more FPS at 2x upscaling with the Immortalis-G720 GPU (Dimensity 8400 and 9300). At the same time, ASR is claimed to deliver power savings of 20% with ASR set to 2x upscaling using the quality preset.

Mobile upscalers can sacrifice visual fidelity for performance and/or battery life. Qualcomm's GSR using spatial upscaling was restricted by its hardware and game support, which limited its impact in the Android space. Since ASR is open-source, developers can tweak and finetune it to match their needs. Moreover, it is likely to be supported on a majority of Arm devices, irrespective of the vendor. In the "Mori" demo, Arm showed off ASR in Unreal Engine 5 delivering up to 30% more FPS, with no compromises to the visuals, according to the company.

As a part of their promotional efforts, Arm has introduced a new chapter in its Mali Manga series featuring ASR and to be honest, I could get used to reading PRs like this. Think of it like DLSS or FSR for your mobile devices, and game integration is expected to be simplified as Arm releases pre-baked plugins for Unity later this year, and Arm plugins for Unreal Engine 5.3, 5.4, and 5.5 available now. Game developers utilizing these engines should find ASR integration relatively straightforward. We anticipate support in leading titles such as Genshin Impact (Unity), Wuthering Waves (Unreal Engine), Fortnite (Unreal Engine), and the list goes on.

Due to its open-source nature, ASR can be integrated into other game engines as well, but that will require increased developer efforts. Sticking to mobile-only SoCs, ASR should work on Qualcomm's Snapdragon, MediaTek's Dimensity, Google's Tensor, Samsung's Exynos, and Apple's A-series/M-series hardware. While we believe ASR should be as scalable as FSR 2 and support virtually all modern-day Arm hardware, the requirements depend on the developers and Arm itself, though they should be pretty lenient.