Nvidia (NVDA) has been the undisputed star of the AI chip boom, with its GPUs powering the training and inference of large language models. But as the industry shifts from simple chatbots to autonomous AI agents, a different kind of chip is stepping into the spotlight: the humble CPU.

BofA Securities sees a massive $210 billion opportunity in server CPUs by 2030, and three companies are positioned to ride that wave: Advanced Micro Devices (AMD), Intel (INTC), and Arm Holdings (ARM).

Here's the deal: in a traditional server, the CPU is the brain that coordinates everything. It manages the operating system, runs applications, handles storage, and juggles network connections. GPUs, on the other hand, are the muscle, doing the heavy mathematical lifting for AI models. In AI servers, GPUs do the intense calculations, but CPUs still play a crucial role in feeding them data and managing the overall workflow.

Now, BofA Securities is looking ahead to what it calls the "agentic server." AI agents can perform multi-step tasks without needing a human to spell out every instruction. In this new architecture, the CPU becomes an orchestration layer, coordinating the complex dance of data and tasks. That shift could dramatically increase the demand for CPUs.

Why AI Agents Could Change The CPU Market

The key distinction is between today's AI models and tomorrow's AI agents. Today's models are like calculators: you give them a prompt, they give you an answer. Tomorrow's agents will be more like employees: you give them a goal, and they figure out the steps to achieve it.

That requires a lot more coordination, which is where CPUs come in. According to BofA Securities semiconductor analyst Vivek Arya, this could change the balance between CPUs and GPUs inside AI data centers.

The CPU Market Could Nearly Quadruple

BofA Securities has raised its 2030 server CPU market estimate to $210.6 billion, up from about $170 billion previously. That's a roughly 36% compound annual growth rate from 2026 through 2030. The bank estimates CPUs could account for about 10% of the total data-center systems market by 2030, versus roughly 7% during the 2024-2025 training era.

Inside that $210 billion, the bank splits the market three ways. Traditional cloud and on-premise CPUs account for about $30 billion. AI cluster and head-node CPUs, which supervise racks of accelerators, account for about $90 billion. And a third bucket worth about $90 billion didn't meaningfully exist two years ago: standalone processors running AI agents.

The crucial insight is that CPUs are additive to GPUs, not a replacement. Agentic workloads layer new CPU-dense infrastructure alongside existing accelerator clusters. BofA expects the CPU-to-GPU ratio to move from roughly 1:4 during the AI training era toward 1:2 for inference and eventually close to 1:1 as agentic AI expands. In other words, the AI boom could require more CPUs precisely because it requires more GPUs.

Investors have already caught on. Since the start of the year, the three major CPU producers have seen their stocks rally by triple digits:

ARM Could Be the Biggest Share Winner

The market-share forecast is where the story gets particularly interesting. BofA expects Intel's share of overall server CPU revenue to fall from roughly 34% in 2026 to 22% in 2030. AMD is expected to remain relatively stable, moving from about 28% to 31%.

ARM is the major share winner. BofA expects ARM-based CPUs to represent 47% of server CPU value by 2030, including roughly 38% from merchant products and 9% from custom designs. The merchant opportunity includes new products such as Nvidia's Vera CPUs and Qualcomm solutions. The custom opportunity includes hyperscaler chips such as Amazon's Graviton, Google's Axion, and Microsoft's Cobalt.

But that doesn't mean ARM captures all the economics. Much of the value can accrue to companies licensing its architecture or building chips around it. AMD, by contrast, sells the chip and keeps the margin.

Why AMD Still Matters

Despite ARM's share gains, BofA Securities names AMD its "top CPU pick."

"AMD remains our top CPU pick on its wide breadth of portfolio and dual leadership – highest-frequency (compute/head node rack) and highest core/thread count (agentic AI)," Arya said.

The bank expects AMD to remain strong in high-end AI server workloads, helped by its core-count advantage. Arya points to AMD's Zen 6 Venice, which is specified at up to 256 cores, against 192 for Intel's Diamond Rapids and 88 for Nvidia's Vera.

The investment takeaway is simple: the AI infrastructure boom may be creating a second chip market alongside GPUs.

What Wall Street Says on AMD

According to analyst ratings, AMD carries a consensus Buy rating and an average price target of $550.86, roughly 12% above a recent $493.38. The range runs from $248 to $730.

Most of the recent moves clustered around second-quarter earnings:

So while Nvidia may be the face of the AI boom, the next wave could belong to the chips that make it all work together. And that's a story worth watching.