For greater than 35 years, Arm Holdings has licensed its instruction units to the world’s largest chipmakers and picked up royalties on each processor made with its designs. Now the U.Ok.-based firm is making bodily silicon of its own for the first time.
Arm CEO Rene Haas unveiled his firm’s first in-house chip on Tuesday at an occasion in San Francisco. Arm is looking the brand new knowledge heart central processing unit the AGI CPU. It’s a long-anticipated move that marks a significant change for the so-called Switzerland of chip companies as it enters into contemporary competitors with its clients.
Meta is the first to signal on, as the social media firm builds out a number of gigawatts of AI knowledge facilities and plans to shell out up to $135 billion on capital expenditures this 12 months. In February, Meta secured an enormous quantity of chips from each Nvidia and Advanced Micro Devices.
“In today’s world, you really only have a couple of players,” mentioned Meta software program engineer Paul Saab, who helped with the Arm chip venture since its begin in 2023, in an interview with CNBC. “This adds yet another player to the ecosystem for us.”
Saab added that the Arm deal “allows a lot more flexibility in our software stack and in our supply chain.”
Terms of the settlement weren’t disclosed. For Arm, the deal marks a significant win and a stamp of approval from probably the most worthwhile firms on the planet.
“Let’s say they get 5% of Meta’s $115 to $135 billion capex going into the future,” mentioned chip analyst Patrick Moorhead of Moor Insights. “That is a game changer on the top line for them.”
It’s additionally the newest signal that CPUs are seeing a resurgence in demand. Nvidia, which has established itself as the chief in AI graphics processing items, lately advised CNBC that CPUs are “becoming the bottleneck” as agentic AI modifications compute wants. Futurum Group calls it a “quiet supply crisis,” predicting the CPU market growth rate could exceed GPU progress by 2028.
While GPUs are perfect for coaching and working AI fashions as a result of their 1000’s of cores can carry out many operations concurrently, CPUs have a smaller variety of highly effective cores working sequential general-purpose duties. Agentic AI requires a variety of normal compute energy, with massive quantities of knowledge transferring round throughout a number of brokers.
At Nvidia’s annual GTC conference final week, CEO Jensen Huang unveiled a complete rack crammed solely with Vera CPUs. At the Arm occasion on Tuesday, Huang appeared in a recorded assertion congratulating Arm on its new CPU.
Top leaders from Google, Amazon, Microsoft, Oracle, Broadcom, Micron, Samsung, SK Hynix and Marvell additionally appeared within the video. Arm advised CNBC about 50 companions signaled assist forward of the launch.
“It’s a $1 trillion market, and what we’re seeing over and over again is actually our partners coming out and understanding and realizing this is actually great for the industry,” Mohamed Awad, Arm’s cloud AI head, advised CNBC in an interview.
CNBC acquired an unique first have a look at Arm’s new chip lab the place it is making ready the brand new CPU for full manufacturing later this 12 months.
Arm’s head of cloud AI Mohamed Away gave CNBC’s Katie Tarasov a tour of the chip lab the place it created its first in-house chip, the AGI CPU, in Austin, Texas, on Friday March 6, 2026.
Erin Black | CNBC
‘Create the chip that you really want’
Arm spent $71 million and about 18 months constructing three new lab rooms at its campus in Austin, Texas, the place a once-tiny staff has grown to over 1,000 folks. Inside, engineers “bring up” the chips by placing them via a number of rounds of testing as soon as they arrive off the manufacturing facility line.
Like almost all fabless AI chipmakers, Arm at present manufactures its CPU at Taiwan Semiconductor Manufacturing Company‘s fabrication vegetation. Made on TSMC’s 3-nanometer node, Arm’s CPU is totally manufactured in Taiwan for now. TSMC has a 3nm fab coming to Arizona quickly, and Awad mentioned Arm would “love to manufacture here. It really comes down to what our customers are ultimately looking for.”
Arm is greatest identified as the main structure for cell chips in nearly each smartphone. It acquired into knowledge heart chips in 2018 with the launch of its Neoverse platform. Amazon took Neoverse mainstream in its first customized processor, Graviton, and Google and Microsoft now additionally base their AI chips on Arm.
“If Arm didn’t exist, then all of those companies who have their own processors wouldn’t be able to create their own,” Moorhead mentioned.
Still, most server chips are constructed on conventional x86 structure utilized by Intel and AMD. Moorhead calls x86 “tried and true” and mentioned it could actually “run pretty much anything.”
The advantage of Arm structure is it is “super efficient” as a result of the designs are extra customizable, Moorhead mentioned. “You can just create the chip that you want with nothing else.”
Awad advised CNBC the Arm staff “ruthlessly optimized” its new AGI CPU for synthetic normal intelligence — therefore the title. Up to 64 of the brand new CPUs, a complete of about 8,700 cores, can slot in a single air-cooled rack. It’s a dense configuration Arm is betting will attraction to power-constrained knowledge heart clients around the globe.
“You can get two times the performance-per-watt than you can from an x86 rack,” Awad mentioned. “That means twice as much performance in the same footprint, in the same power.”
Meta’s Saab mentioned wattage is “a very scarce resource.”
“If you have a best in class CPU that’s giving you the best performance-per-watt that you possibly can, that opens up more wattage for other parts of your infrastructure,” he mentioned.
Meta’s 5-gigawatt Hyperion knowledge heart underneath building in Richland Parish, Louisiana, Jan. 9, 2026.
Courtesy of Meta
‘Available to the entire world’
Meta has a giant want for effectivity as it builds out large AI knowledge facilities throughout Louisiana, Ohio and Indiana. The firm can also be reportedly trying to lease area on the big Stargate web site in Texas, the place OpenAI and Oracle scrapped plans to develop as much as 10GW of capability.
Meta’s AI spending spree comes after its Llama 4 mannequin was not well received by builders final 12 months.
“They got behind,” Moorhead mentioned. “They also recognized we don’t have enough compute power to do what we need to do.”
In addition to securing processors from Nvidia and AMD, Meta unveiled four new chips in March inside its own line of Meta Training and Inference Accelerators that it has been making since 2023. Now, it is including CPUs from Arm to the combination.
“It was meant to basically be a full replacement, drop-in replacement, for our current compute CPUs and be transparent to our developers,” Saab mentioned.
Saab was at Facebook in 2011, when the corporate launched the Open Compute Project, a consortium that now has a whole bunch of member firms, together with Arm and Nvidia, dedicated to open {hardware} designs that assist scale back knowledge heart vitality consumption and prices.
“The first conversations we had with Arm were, ‘Hey, if we build this, we don’t want to keep this only within the company,'” Saab mentioned. “We’re not like a chip company that’s trying to build sales channels to sell chips. We wanted it to be available to the whole world.”
Arm would not disclose pricing for the CPU, however Moorhead predicts will probably be within the 1000’s of {dollars}.
Awad advised CNBC it could be “competitively priced,” with an goal to serve as an choice for firms that may’t afford to make their own in-house processor.
“You have to have 1,000 engineers, a $500 million dollar budget to go and create it,” Moorhead mentioned. “So there’s definitely a market need.”
Watch: Inside Arm’s $71 million chip lab where its making its first ever CPU
