Bank of America raised its forecast for the 2030 server CPU market to more than $210 billion, driven by rising demand from AI agents that require processors to orchestrate GPU workloads in data centers.
The new forecast reflects a shift in the CPU-to-GPU ratio from about 1:4 during the training era toward roughly 1:1 for agentic AI, positioning CPUs as complementary to rather than competing with GPUs.
What happened
Bank of America analyst Vivek Arya lifted the 2030 server CPU total addressable market forecast to more than $210 billion from around $170 billion, implying nearly 5x growth off the ~$35bn CY25 level and a 36% compound annual growth rate (up from 30% previously).
Why it matters
The shift reflects AI agents' rising demand for CPUs in data centers. BofA argues CPUs are becoming essential as the orchestration control plane alongside GPUs rather than being substituted by them, meaning chip makers face expanding rather than shrinking opportunities. CPUs are now expected to account for about 10% of the overall $2.2 trillion data center systems market by 2030, versus roughly 7% in the training era.
What to watch
BofA names AMD as its top CPU pick on dual leadership in frequency and core count, flags Nvidia as the overall sector top pick, highlights Intel's foundry optionality, and notes Arm as the fastest share gainer.
Bank of America analyst Vivek Arya lifted the firm's 2030 server CPU total addressable market forecast to more than $210 billion, up from around $170 billion, citing strong AI compute and memory demand trends observed in second-quarter earnings reports. The new figure implies nearly 5x growth off the ~$35bn CY25 level, with the market now expected to grow at a 36% compound annual rate, compared to the prior 30% forecast.
Central to the upgrade is a structural shift in how data centers allocate compute resources for AI workloads. BofA noted that the rise of agentic inference—AI systems designed to autonomously complete tasks—is reshaping the CPU-to-GPU ratio from roughly 1:4 during the training era toward approximately 1:1 for agentic AI. The key change is that CPUs are becoming the orchestration control plane, managing scheduling and coordination while GPUs handle intensive inference computation. As a result, CPUs are expected to account for about 10% of the overall $2.2 trillion data center systems market by 2030, versus roughly 7% during the training era.
BofA framed the debate around substitution versus expansion, arguing that CPUs are additive to overall system TAM rather than cannibalizing GPU demand. The firm cited near-record GPU rental and memory spot prices as evidence of a broader compute shortage that extends to CPUs, supporting the view that data center operators will purchase more of both processor types to meet agentic AI demands. AMD remains the firm's top CPU pick on the basis of dual leadership in frequency and core count, while Nvidia retained its position as the overall sector top pick. BofA also flagged Intel's foundry optionality as a potential opportunity and noted Arm as the fastest share gainer among CPU vendors.
Bank of America's CPU market upgrade reflects a fundamental reassessment of how AI infrastructure will evolve beyond the initial GPU-centric training phase. The core insight is that agentic AI—systems that act autonomously to complete tasks—requires a different hardware balance than large language model training. Where training workloads are GPU-dominated (reflected in the historical 1:4 CPU-to-GPU ratio), agentic inference spreads the load: CPUs handle orchestration, scheduling, and control logic while GPUs execute compute-heavy inference kernels. This shift is not a zero-sum replacement but an expansion of total system demand.
The firmness of near-record GPU rental and memory spot prices, which BofA cites as evidence of an across-the-board compute shortage, suggests the market is capacity-constrained rather than saturated. Under this reading, CPUs are additive—companies building agentic systems will need more total processors, not swap GPUs for CPUs. The move from ~7% to ~10% CPU share of the $2.2 trillion data center systems market by 2030 is modest in percentage terms but represents a doubling in absolute opportunity within a vastly larger pie. AMD's positioning on frequency and core count aligns with orchestration workloads, which benefit from single-thread performance and lower core counts than GPU-heavy setups might require.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
Quanta Computer announced at its August 13 earnings call that it is raising its full-year 2026 capital expendi…
Salesforce spent $27 billion on stock buybacks in its first quarter ending in April—a record for the company—e…
Nebius Group and CoreWeave both reported Q2 2026 earnings this week showing strong AI compute demand
Nebius, an AI cloud company listed on Nasdaq, has agreed to lease high-density NVIDIA-powered AI infrastructur…
Cerebras Systems fell over 18% in premarket trading after its second earnings report as a public company showe…
Anthropic's Fable 5, billed as the most capable AI model available, captured only about 6 percent of tokens pu…
The AI news that matters, in one minute each morning.
Sign up free