Bank of America projects $125B CPU market

Bank of America analysts raised their view of the server CPU market to $125 billion by 2030, according to a Yahoo Finance report published on May 20. The bank said that would mark an increase from roughly $43 billion in 2026 and would represent a 31% compound annual growth rate. The call is getting attention because it pushes back on the idea that AI infrastructure growth belongs only to GPUs. (finance.yahoo.com) Bank of America’s argument, as reported by Yahoo Finance, is that CPUs are set to capture a larger share of data-center spending as AI systems move toward more agentic workloads. (amd.com) ### Why are Bank of America analysts suddenly talking about CPUs again? Bank of America’s forecast is tied to a broader agentic AI thesis that the firm has been developing across hardware and software. In a Bank of America Institute report published in September 2025, BofA Global Research said agentic AI’s total addressable market could reach $155 billion by 2030. (finance.yahoo.com) Agentic AI systems do more than generate text or images from a prompt. The Bank of America report describes them as systems that plan, monitor, invoke tools, update data and coordinate actions with limited supervision. (finance.yahoo.com) Those functions require general-purpose compute for orchestration and system control, which is the part of the stack CPUs are built to handle. ### What jobs are CPUs expected to do in an AI data center? Nvidia executive Dion Harris told CNBC in March that “CPUs are becoming the bottleneck” as AI and agentic workflows expand. (institute.bankofamerica.com) CNBC reported that the shift is being driven by a change from chatbot-style AI to task-oriented applications that move large amounts of data and coordinate across multiple agents. CNBC said GPUs remain central for training and running models, while CPUs are used for sequential, general-purpose tasks. In practical terms, that means CPUs sit closer to memory management, orchestration, data handling and head-node functions that support larger AI systems. (institute.bankofamerica.com) ### Which companies are positioned to benefit if that forecast holds? AMD has already been making the same case. AMD said on May 5 that first-quarter 2026 data center revenue rose 57% year over year to $5.8 billion. CNBC reported the same figure from the company’s earnings release. (cnbc.com) Lisa Su said on AMD’s first-quarter earnings call that the server CPU market could exceed $120 billion by 2030, according to reports summarizing the call. That places AMD’s own view close to Bank of America’s new $125 billion figure. (cnbc.com) Nvidia is also pushing harder into CPUs. Nvidia’s Vera CPU page says the chip has 88 Olympus cores, and CNBC reported in March that Vera had entered production as Nvidia prepared to expand CPU deployments in AI systems. ### Does this mean GPUs matter less? (amd.com) Bank of America’s forecast does not say GPUs are being displaced. Yahoo Finance’s report framed the CPU upgrade as part of a broader bullish view on AI chip spending, not a replacement of one category by another. Nvidia’s own product strategy points the same way. (msn.com) CNBC reported that Grace and Vera CPUs are typically deployed alongside Nvidia’s GPU systems, and Bank of America has separately raised its broader AI data-center systems forecast in recent weeks. That suggests the firm sees CPUs taking a larger share inside a growing overall market. (nvidia.com) ### What should readers watch next? July 22-23 is AMD’s scheduled Advancing AI event, where the company is expected to discuss next-generation data-center products and customers. Nvidia, for its part, has already begun detailing Vera as part of its AI data-center roadmap. (finance.yahoo.com) By 2030, the benchmark in this debate is now explicit: Bank of America’s latest estimate puts the server CPU market at $125 billion, with AMD and Nvidia both building product roadmaps around the same agentic AI buildout. (finance.yahoo.com) (amd.com) (cnbc.com)