Demand for AI compute has shifted in 2026, and the biggest beneficiaries are no longer only GPU makers. Hyperscalers that spent heavily on graphics processing units and custom AI accelerators over the past three years are now increasing purchases of central processing units, creating a new growth opportunity for chip designers with strong data center exposure. Among the companies gaining momentum, Arm Holdings has emerged as one of the clearest winners.
Arm has long dominated CPU architecture in mobile devices, but in recent years it has made meaningful inroads into data center processors, a market traditionally led by Intel and, to a lesser extent, AMD. Its energy-efficient designs are increasingly attractive for AI data centers, where power consumption is becoming a major constraint. As AI workloads evolve, especially with the rise of agentic AI, CPUs are expected to play a larger role in managing servers, coordinating GPU activity, moving data, and handling machine-to-machine communication. Industry leaders have suggested that the balance between CPUs and GPUs could shift significantly over time, with more CPU content needed in AI systems.
That shift appears to be benefiting Arm. The company said its market share of CPU compute among the top hyperscalers is now around 50%, reflecting strong adoption from major cloud providers. Nvidia has indicated it sees nearly $20 billion in total CPU revenue this year, while Amazon says its Arm-based Graviton chips are widely used across its customer base and have even been expanded through a new deal with Meta Platforms. Microsoft and Alphabet have also begun deploying Arm-based CPUs in their data centers, adding further evidence of broadening acceptance.
Arm’s management believes this momentum can meaningfully accelerate financial growth. The company has estimated that the total addressable market for data center CPUs could reach $100 billion by 2031, up from $50 billion in 2026. AMD has offered an even higher market estimate for server CPUs, suggesting the opportunity may be larger than Arm’s own forecast. Based on current trends, Arm expects its royalty revenue growth to rise from a 14% compound annual growth rate over the last five years to 20% over the next five years.
The company is also trying to capture more value directly by selling its own chips. Arm has developed the Arm AGI CPU and believes selling first-party chips could generate 10 times the gross profit per chip compared with its licensing model. Management expects the first-party chip business to reach $15 billion in annual sales by 2031, potentially producing $7.5 billion in gross profit. Arm has already reported strong early demand, though supply chain constraints have limited how quickly it can scale production.
Looking further ahead, Arm expects its combined licensing and first-party chip businesses to generate about $25 billion in revenue by 2031, translating to roughly $9 in earnings per share, up sharply from $1.77 last year. However, the stock’s valuation is already very demanding, trading at about 159 times forward earnings. Even with impressive growth ahead, that valuation makes it difficult to argue for another triple in the share price from here.
