Lead
Bernstein updated its outlook for server CPUs on Wednesday, arguing that the rise of agentic artificial intelligence is altering hardware demand inside data centers and creating a large structural market for CPUs. The firm’s analyst, David Dai, increased his estimate of the server CPU total addressable market to $223 billion, marking a substantial upgrade from his previous forecast.
Shift in AI workloads
Bernstein attributes the change to a move beyond generative AI's early chatbot applications toward agentic AI, which the note describes as involving "heavily autonomous task orchestration and execution, which boosts the CPU workload vs. GPU." The firm highlights a material change in the CPU-to-GPU ratio deployed for AI: a shift "from 1:4 or 1:8 to 1:1 or higher." That rebalancing is central to Bernstein's revised demand outlook.
Market sizing scenarios and assumptions
Under Bernstein’s base case, the firm assumes $3.5 trillion in AI data center capital expenditure and a 1:1 CPU-to-GPU pairing ratio for inference. Those assumptions imply a server CPU TAM of $137 billion by 2030, which the firm notes would be roughly six times the 2025 TAM of $37 billion. Bernstein also outlines an upside scenario in which the TAM reaches $330 billion. In contrast, the firm now treats its earlier $137 billion forecast as the bear case given the updated assumptions.
Company implications
Bernstein identifies Arm Holdings as "the structural beneficiary of the renaissance of CPUs for agentic AI." In its note, Dai raised his price target on Arm to $500 from a prior level, representing 21% upside from the firm’s reference price, and projects Arm could reach $22 billion in revenue by 2030, above the company’s own $15 billion target. Bernstein describes Arm’s architecture as power-efficient and "suitable for agentic AI workload."
The firm also adjusted its models for major CPU vendors to reflect stronger and more sustained server CPU demand. Dai increased price targets on AMD and Intel to $600 and $100, respectively. AMD retains an Outperform rating in Bernstein’s view, while Intel is rated Market Perform.
Market and investment context
Bernstein’s revision reflects a structural reappraisal of where processing will be consumed in AI systems, with potential implications across semiconductors, cloud infrastructure, and hyperscaler capital allocation. The analysis ties projected capital spending in data centers directly to server CPU opportunity, and it positions vendors with power-efficient architectures and competitive server CPU roadmaps to capture disproportionate gains if the agentic AI scenario plays out as outlined.
Methodological note
The figures, targets, and scenario assumptions above are drawn from Bernstein’s published note and reflect the firm’s stated base and upside cases.