For 35 years, Arm made money the same way: design a chip architecture, license it to others, collect royalties. Companies like Apple, Nvidia and Amazon paid to use Arm's blueprint. It was a simple, profitable formula. On Tuesday, Arm ended it.
Arm shifted from only licensing chip architecture to giants like Apple, Nvidia, Amazon and Google to competing with them in physical silicon, unveiled the AGI CPU at its annual conference in San Francisco. 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 bet is enormous. Arm hopes to capture a share of the $1 trillion AI CPU market. CEO Rene Haas predicted in his keynote that the company's datacenter silicon could catapult its datacenter total addressable market to more than $100 billion by the end of the decade. That's a 33-fold increase from the roughly $3 billion in annual royalties Arm currently collects from the datacenter chip business.
What changed? Agentic AI. 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. As AI systems grow more complex and start acting autonomously on behalf of users, they generate unprecedented demand for general-purpose CPU compute. These frameworks need CPU cores and memory to write and execute code, automate tasks, and facilitate the reinforcement learning used to train next gen models. Arm is betting on the proliferation of these agents to drive a four-fold increase in CPU demand, and it's positioning its latest chip to capitalize on this trend.
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. A standard air-cooled 36kW rack holds 30 blades for 8,160 cores total. Other launch partners include Cerebras, Cloudflare, F5, OpenAI, Positron, Rebellions, SAP, and SK Telecom.
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. The company faces stiff competition. Meta is already deploying large numbers of Nvidia's Arm-based Grace CPUs and plans to use the company's Vera chips. The chip will compete directly with Nvidia's standalone Vera CPUs and rack systems.
The CPU market itself is experiencing a genuine supply crunch. AMD CEO Lisa Su said that the increased demand had primarily been driven by the rise in agentic AI applications, and that while this had undoubtedly led to CPU supply chains tightening, AMD was working with customers to ease bottlenecks. "If you talk to our top customers, they're like: 'Wow. Lisa, the demand for CPU compute sitting along AI was perhaps something that was under-forecasted.' We are in the process of catching up," she said.
Yet no single vendor controls sufficient manufacturing capacity to fill that demand. Like nearly all fabless AI chipmakers, Arm currently manufactures its CPU at Taiwan Semiconductor Manufacturing Company's fabrication plants. Made on TSMC's 3-nanometer node, Arm's CPU is entirely manufactured in Taiwan for now. Intel, AMD, Nvidia and Google's custom Axion chips are all chasing the same TSMC production slots.
For investors, the question is whether Arm can execute in hardware, a business it has avoided for decades. The company has already invested heavily. Arm spent $71 million and about 18 months building three new lab rooms at its campus in Austin, Texas, where a once-tiny team has grown to over 1,000 people. Commercial systems are already available for order and the AGI CPU is set to arrive later this year. But arriving late to a crowded market, even with established customers, is not the same as winning it. The real test comes in volume production and sustained competitiveness against rivals with decades of manufacturing experience and their own design advantages. For a company that built its empire on intellectual property, not iron, that's unfamiliar territory.