Arm’s $15 Billion Bet on Agentic AI

Arm Holdings, the company whose chip designs power virtually every smartphone on the planet, has made the most dramatic strategic shift in its 35-year history. On March 24, 2026, Arm announced it will begin selling its own production silicon—a move that transforms it from a neutral architect and licensor into a direct competitor with some of its biggest customers, while positioning it at the center of the AI revolution .

The new chip, dubbed the Arm AGI CPU, represents a bet that the future of computing belongs not just to specialized AI accelerators like GPUs, but to a new class of general-purpose processors optimized for the unique demands of agentic AI . With Meta as its lead partner and co-developer, Arm projects the chip business will generate $15 billion in annual revenue by fiscal 2031—a figure so large it would nearly quadruple the company's current total revenue .

The AGI CPU: Arm's First-Ever In-House Processor

At its "Arm Everywhere" event in San Francisco, the company formally crossed the line from chip designer to chip vendor. The Arm AGI CPU is a 136-core, 3-nanometer data center processor built on the Arm Neoverse V3 platform and manufactured by TSMC on its N3 process node . Designed for a 300-watt envelope, the chip is purpose-built for agentic AI inference workloads—systems that must reason and execute multi-step tasks independently, rather than simply responding to a single prompt .

Arm claims the AGI CPU delivers more than 2x the performance per rack compared with current x86 platforms . The project, according to Arm's Executive Vice President of Cloud AI, Mohamed Awad, began three years ago at Meta's explicit request. Hyperscalers, Awad said, "wanted more than blueprints. They wanted Arm to hand them a product" .

The chip goes into volume manufacturing in the second half of 2026, with first production-shipment revenue expected in the fourth quarter of Arm's fiscal 2026, which runs through March 2027 . This ambitious timeline makes clear that Arm isn't experimenting—it's executing a plan that has been quietly in development for years.

Meta Is Lead Customer—and Co-Developer

The relationship between Arm and Meta on this chip is unusually deep. Meta is not just the first customer; it served as the lead partner and co-developer, working alongside Arm to optimize the AGI CPU for gigawatt-scale infrastructure across Meta's family of apps . The chip is designed to work in tandem with Meta's own custom MTIA accelerators, creating a purpose-built silicon stack for the company's massive AI workloads .

In a statement, Meta confirmed it will deploy the Arm AGI CPU across its data centers and will release board and rack designs through the Open Compute Project later this year, making the platform available to other hyperscalers . Beyond Meta, Arm has named an impressive roster of committed customers including OpenAI, Cloudflare, SAP, Cerebras, SK Telecom, and several others .

Demand Surges: $2 Billion in Near-Term Orders

If there was any doubt about hyperscaler appetite for an Arm-built data center CPU, it vanished during Arm's Q4 FY2026 earnings report on May 6, 2026. Just six weeks after launch, CEO Rene Haas and CFO Jason Child disclosed that customer demand for the AGI CPU now exceeds $2 billion across fiscal 2027 and 2028—more than double the $1 billion figure disclosed at the March launch event .

"Customer response to Arm AGI CPU has been strong," Haas and Child wrote in a shareholder letter. "We now have more than $2 billion of customer demand." However, they also noted a critical constraint: supply . Arm has only secured enough wafer, substrate, memory, and packaging capacity to fulfill $1 billion of that demand, meaning production is effectively sold out at current capacity . Management maintained its $1 billion near-term outlook while working to secure additional supply from TSMC and other ecosystem partners .

Bank of America analysts noted the supply-side tension, writing that "limited supply availability across wafers, substrates, memory, packaging, and other inputs is keeping a lid on near-term upside" despite the explosive demand .

The $15 Billion Target and the Strategic Math

Arm's long-term vision is built on two distinct growth vectors: the traditional IP licensing and royalty business, and the new chip hardware business. Management projects the licensing business will reach $10 billion in annual revenue by fiscal 2031, while the AGI CPU business alone is expected to contribute $15 billion—for a combined $25 billion in total annual revenue .

To put that in perspective, Arm's total revenue for fiscal 2026 was $3.601 billion . The $25 billion target implies a nearly 7x increase in just five years, driven almost entirely by the company's pivot from smartphone IP licensing to data center silicon. CEO Haas has described it as "a very pivotal moment for the company" .

From Smartphone IP to Data Center Platform

Arm's transformation extends well beyond the AGI CPU. The company's traditional royalty business—historically dominated by smartphone chips—is now being eclipsed by cloud and data center deployments. In its Q4 FY2026 report, Arm disclosed that data center royalty revenue more than doubled year-over-year, driven by hyperscaler adoption of Arm-based server chips, networking chips, DPUs, and SmartNICs .

Arm now claims near-100% market share in DPUs and SmartNICs, and its architecture powers every major hyperscaler's custom silicon: AWS Graviton, Google Axion (paired with TPUs), Microsoft Cobalt, and Nvidia's forthcoming Vera CPU . On the Q4 earnings call, management stated flatly that "soon, the data center will be Arm's largest business" .

More than 9,100 accounts now use Arm's Cortex AI platform weekly, a figure the company says continues to grow as enterprises shift from AI experimentation to production deployment .

Haas on Export Controls: "CPUs Are Kind of Like Oil"

Rene Haas has emerged as one of the semiconductor industry's most vocal critics of broad U.S. export controls on AI chips. Speaking in Taipei on June 2, 2026, Haas argued that the U.S. would have "significant difficulty" banning exports of AI-capable CPUs to China because CPUs are so universally used that isolating an "AI CPU" from a general-purpose CPU is practically impossible .

"CPUs are kind of like oil relative to the application space," Haas told Reuters. "That's a pretty hardcore cut" . He warned that overly broad restrictions risk slowing overall technological progress, shrinking the total addressable market, and ultimately harming consumers and companies alike .

The position aligns Haas directly with Nvidia CEO Jensen Huang, who has made similar arguments about the counterproductive nature of broad export bans. Both leaders have argued that such controls risk forcing China to accelerate development of its own independent semiconductor ecosystem, which could ultimately reduce the global market for Western chipmakers .

Haas first raised these concerns publicly in October 2023 at WSJ Tech Live—a position he has reiterated consistently through multiple administration changes .

The Bigger Picture: Arm's Structural Inflection

The AGI CPU represents more than a product launch; it marks a structural inflection point for the global semiconductor industry. The company that for 35 years supplied the architectural blueprints for nearly every mobile device is now a direct silicon vendor—competing with the very licensees that built businesses on its IP.

Analysts at Citi called it "the most significant strategic shift for Arm since its founding" . The launch also reshapes competitive dynamics across AI infrastructure, challenging the x86 duopoly held by Intel and AMD in data centers. Arm-based chips already dominate in edge and mobile; the AGI CPU is designed to extend that dominance to the agentic AI data center—a market Arm estimates at more than $100 billion .

With 50+ ecosystem partners committed, hyperscalers lining up for supply, and a $15 billion revenue target to hit in five years, Arm's pivot is arguably the most consequential business model transformation in the semiconductor industry today.