As artificial intelligence evolves from simply answering questions to performing tasks autonomously, the role of CPUs inside data centres could change dramatically.
BofA Securities expects this shift toward agentic AI to create a massive new opportunity for server CPUs, potentially benefiting Advanced Micro Devices (AMD), Intel (INTC) and Arm Holdings (ARM).
The brokerage has raised its estimate for the global server CPU total addressable market to approximately $210.6 billion by 2030, from around $170 billion previously. That would represent roughly 36% annual growth between 2026 and 2030.
Why CPUs Could Become More Important in the AI Era
Today’s AI infrastructure is heavily associated with GPUs. GPUs handle the massive parallel computations required to train and run large AI models.
But GPUs do not operate alone.
In an AI server, CPUs coordinate workloads, manage operating systems and applications, handle networking and storage, and feed data to accelerators.
As AI systems become more autonomous, BofA believes the CPU’s role could expand significantly.
The next stage could be the rise of the agentic server.
AI agents can plan tasks, retrieve information, execute code and perform multiple steps with limited human intervention. These workloads involve significant amounts of orchestration, decision-making and data movement.
That makes CPUs particularly relevant because many of these tasks are latency-sensitive, sequential and I/O-intensive, rather than purely parallel mathematical workloads.
The $210.6 Billion CPU Opportunity
BofA’s latest estimate represents a substantial increase from its earlier projections.
The bank previously raised its 2030 server CPU opportunity to more than $170 billion, up from $125 billion. The latest estimate of approximately $210.6 billion represents another significant increase.
BofA sees the market broadly divided into three areas:
- Traditional cloud and on-premise CPUs: approximately $30 billion
- AI cluster and head-node CPUs: approximately $90 billion
- Standalone CPUs for agentic AI workloads: approximately $90 billion
The third category is particularly important because standalone processors dedicated to AI-agent workloads represent a market that was relatively small or nonexistent just a few years ago.
BofA also expects CPUs to account for around 10% of total data-centre systems spending by 2030, compared with roughly 7% during the 2024–25 AI training era.
1. AMD — BofA’s Top CPU Pick
AMD appears to be BofA’s preferred way to play the CPU opportunity.
Analyst Vivek Arya has described AMD as the firm’s top CPU pick, citing its server CPU position, product pipeline and ability to participate in both CPU and GPU markets.
BofA previously raised its AMD price target to $560 from $500, reflecting higher CPU and GPU estimates. The firm estimated AMD could maintain approximately 25%–27% server CPU market share through 2030.
AMD’s advantage is that it does not have to choose between traditional CPUs and AI accelerators. The company has been building a broader data-centre portfolio covering CPUs, GPUs and related compute technologies.
The company has also been developing high-core-count server processors designed to deliver greater throughput within power constraints—an important consideration as data centres become increasingly power-limited.
2. Intel — A Potential Turnaround Opportunity
Intel remains the largest traditional x86 CPU player and could also benefit substantially if the server CPU market expands as BofA expects.
The opportunity is particularly interesting because Intel is simultaneously working on its CPU business and foundry strategy.
BofA has previously raised its Intel price target to $160 from $135, citing improving opportunities in server CPUs and the company’s long-term foundry potential.
For Intel, the agentic-AI thesis could provide an additional growth driver at a time when the company is attempting to regain technological and competitive momentum.
However, Intel’s opportunity comes with greater execution risk than AMD’s. The company must successfully execute its product roadmap while improving manufacturing economics and competing against AMD and increasingly against Arm-based solutions.
3. Arm Holdings — The Architecture Challenger
Arm represents a different way to participate in the CPU opportunity.
Rather than primarily selling processors like AMD and Intel, Arm licenses its CPU architecture and related intellectual property to chipmakers.
The rise of AI infrastructure could accelerate demand for Arm-based server processors, particularly where energy efficiency, scalability and customisation are important.
BofA previously raised its Arm price target to $335 from $245, citing the company’s longer-term chiplet opportunity and the growing adoption of Arm-based processors.
Arm-based CPUs could therefore capture a growing share of the server market as hyperscalers and other large customers increasingly design customised silicon.
Agentic AI Could Change the CPU-GPU Balance
The key point in BofA’s thesis is not that CPUs will replace GPUs.
Instead, AI data centres may need more CPUs alongside their GPUs.
