Agentic AI CPU Demand: Mizuho Lifts AMD, Intel, SanDisk Targets

What happened

Mizuho has raised price targets on a group of semiconductor and hardware stocks. Its argument is that agentic AI, meaning software that plans and carries out multi-step tasks rather than simply answering prompts, is creating a new wave of demand for general-purpose processors. The firm expects this to lift CPU shipments and benefit CPU-based servers and memory suppliers. 1

Analyst Vijay Rakesh kept his Buy rating on Advanced Micro Devices and raised his AMD price target to $705 from $580. 2 He also reaffirmed a Buy on SanDisk and lifted that target to $2,050 from $1,750. 2 Targets also went up for Dell Technologies, Intel and Super Micro Computer, all of which Rakesh expects to gain from agentic workloads. 2 The new levels for those three names were not detailed in the available coverage. Rakesh tied the changes to recent product launches, specifically Meta Platforms' Muse and OpenAI's Dots. 2

The numbers behind the call

The core of the thesis is a forecast for the size of the market. Mizuho expects the agentic CPU total addressable market to reach $80 billion by 2030. That implies a 123% compound annual growth rate over the next four years. 2 By then, the firm projects that agentic workloads will make up roughly 30% of all server CPU shipments. 2

Several company-level estimates sit on top of that:

• AMD: Rakesh sees a CPU opportunity that could reach $220 billion by 2030. 2
• Nvidia: Its CPU business is expected to generate about $20 billion in revenue and more than double by 2027. 2
• Arm Holdings: Named as a beneficiary because of its AGI-focused CPU roadmap. 2

One inconsistency is worth flagging. AMD's $220 billion figure is far larger than the $80 billion agentic CPU market. The most likely explanation is that the AMD number covers a wider CPU opportunity, not only agentic workloads. Readers should not treat the two figures as parts of the same total.

How the coverage differs

The two reports agree on the basic story but frame it differently. The Seeking Alpha piece leads with AMD and Intel and describes the call at the level of the industry: more agentic use cases, more CPU demand, and gains for server and memory makers. 1 TipRanks leads with AMD and SanDisk and supplies most of the hard figures, including the target changes, the market forecast and the company-specific opportunity estimates. 2

The headline mention of Intel is notable mainly as a signal. Intel has spent recent years losing data-center share, so even an unquantified target increase indicates that Mizuho thinks a rising tide in server CPUs could help the incumbent too. 12 The size of that boost is not clear from the coverage.

Why it matters

For most of the generative AI boom, the investment story has centered on accelerators, mainly GPUs, with CPUs treated as supporting hardware. Mizuho's note reflects a growing view that agentic systems change that balance. A chatbot answering one query is mostly an accelerator problem. An agent works differently. It calls tools, queries databases, runs code and coordinates sub-tasks. Much of that orchestration happens on conventional processors and depends on large amounts of memory and storage.

That logic explains why the beneficiary list goes beyond chip designers to include:

• Server builders: Dell and Super Micro.
• A flash storage supplier: SanDisk.
• Arm: whose architecture underpins a growing share of custom data-center CPUs.

It also explains why Nvidia's own CPU business features in the forecast. Even the dominant GPU vendor stands to gain from more CPU-heavy deployments. 2

The caveats

A 123% compound growth rate is aggressive. It assumes agentic AI moves from early launches to mainstream enterprise deployment within a few years. 2 Pointing to individual product releases such as Muse and Dots shows momentum, but it does not prove that spending will last. Several things could shrink the forecast market:

• adoption that is slower than expected;
• efficiency gains that let agents run on fewer cores;
• a shift of orchestration work onto accelerators.

Price targets are also opinions, not guarantees. The steep jumps here, especially for SanDisk, build in considerable optimism.

The takeaway

The most useful reading of Mizuho's call is less about any single target and more about where AI infrastructure spending may go next. If agents become a standard way businesses use AI, the bottleneck could widen from GPUs to the whole server: CPUs, memory and storage. That would spread the gains across a broader set of suppliers, with AMD best placed among CPU makers. Whether the $80 billion forecast holds will depend on how quickly agentic tools move from demos into daily workloads. That makes the next several quarters of server CPU shipment data the figure to watch.