Riding the global wave of AI agents, silicon intellectual property giant Arm Holdings not only delivered record-breaking earnings but also painted an ambitious blueprint during its earnings call: by 2030, the largest share of the CPU market will belong to Arm architecture, and demand for its first in-house chip designed for AI agents doubled to surpass $20 billion within just six weeks of its unveiling.

Arm reported fiscal fourth-quarter 2026 results after the U.S. market close on May 6, with revenue rising 20% year-over-year to $1.49 billion, beating market expectations. Full-year revenue grew 23% to $4.92 billion, marking the third consecutive year of over 20% revenue growth. However, despite the strong results, Arm’s after-hours stock price saw profit-taking pressure as the market had already priced in the gains and management warned about supply chain bottlenecks.

AGI Chip Demand Surges, Supply Chain Becomes the Biggest Test

The biggest highlight of this earnings report was undoubtedly the first in-house AGI CPU chip that Arm unveiled at its “Arm Everywhere” event in late March. CEO Rene Haas, visibly excited on the conference call, noted that demand for the product has doubled since its launch, with total customer demand exceeding $20 billion—far surpassing the originally agreed supply target of $10 billion.

“Customer response to the Arm AGI CPU has been extremely strong. We now have over $20 billion in customer demand spanning fiscal years 2027 and 2028,” Haas said. He revealed that the team is working intensively to secure supply of wafers, memory, packaging, and testing equipment. Despite the demand surge, CFO Jason Child maintained the product’s annual revenue guidance of $10 billion, primarily due to supply chain capacity constraints. Child confirmed that revenue from initial production shipments will be recognized in the fourth quarter of fiscal 2027.

Responding to market concerns about potential conflicts with existing licensing customers, Haas emphasized that all of the more than 50 industry giants consulted—including AWS, Microsoft, Google, Nvidia, and TSMC (ticker: 2330.TW)—expressed support for the strategy. “The more software that is written and optimized for Arm, the stronger it makes everyone,” he explained.

The Arms Race Shifts from “Chip Count” to “Core Count”

In an era where GPUs dominate AI computing power, the role of CPUs is undergoing a qualitative shift due to the rise of AI agents. Haas believes that AI is moving from human-based queries to continuous, agent-driven workloads, dramatically expanding the CPU’s role in task orchestration, data movement, and memory management. He predicted that as AI agents scale, data centers will require more than four times the current CPU capacity, creating a data center CPU market opportunity exceeding $100 billion by 2030.

Addressing competitor AMD’s $120 billion market estimate, Haas acknowledged the scale could indeed be larger but offered a more precise observation: “In terms of chip count, CPUs may not necessarily outnumber GPUs, but in terms of ‘core count,’ it is highly likely.” He explained that current accelerators are constrained by reticle size, making it difficult to stack more GPU dies, but CPUs can increase core counts to parallel-process large numbers of independent agent tasks. The Arm AGI CPU currently features 136 cores, and a doubling of core count to 500 in the coming years is entirely within expectations.

Ecosystem Fully Aligns, Challenging x86 Dominance

Arm’s footprint in the data center space is rapidly expanding. Top hyperscale platforms including Nvidia’s Vera, Google’s Axion, AWS’s Graviton, and Microsoft’s Cobalt have all adopted Arm-based custom CPUs. Haas stated bluntly during the call: “I am confident that by 2030, the largest market share by CPU type will belong to Arm.”

This confidence stems from significant energy efficiency advantages. Google’s recently unveiled TPU 8t and 8i replaced x86 host processors with custom Arm Axion CPUs, achieving an 80% improvement in overall architecture performance while reducing power consumption by 50%. CFO Child added that the highest-volume accelerators currently shipping—including TPU, Trainium, and Rubin—are all connected to Arm architecture and will be 100% Arm-based going forward, expressing strong confidence in the market share trajectory.

Dual Engines Drive Long-Term Targets

Arm is transitioning from a pure IP licensor to a chip supplier, creating dual growth engines of “IP licensing” and “in-house chips.” The company expects that by fiscal 2031, the AGI CPU business will generate $15 billion in revenue, while the IP business will double to $10 billion, totaling $25 billion in revenue—corresponding to earnings per share exceeding $9.

Looking ahead to the first quarter of fiscal 2027, Arm forecast revenue of approximately $1.26 billion, representing about 20% year-over-year growth. Child indicated that both full-year royalty and licensing revenues are expected to maintain growth rates around 20%. Despite near-term supply chain challenges and after-hours stock price volatility, Arm’s core position in AI infrastructure appears increasingly entrenched.