Why Intel Could Meet Only Half of CPU Demand—and What AI Agents Have to Do With It

Intel CEO Lip-Bu Tan said at Splunk’s .conf26 that Intel could meet only about half of the CPU demand customers were requesting. He pointed to AI’s expanding inference and agent workloads as a source of new demand, while supply constraints—including packaging substrates—limited how quickly the industry could respond. The shift adds to CPU demand alongside GPUs; it does not mean GPUs are no longer needed. 1418

Why AI agents can increase CPU demand

AI training and AI inference are different stages of a workload. GPUs are central to many training and model-running tasks, but AI systems also need general-purpose computing to coordinate work: managing orchestration, control flow and other tasks around inference. As applications move toward agents that take actions across multiple steps, that coordination can add to CPU requirements. 2718

That is the demand-side explanation Tan gave for the squeeze. It is not evidence that every AI agent uses the same amount of CPU capacity, or that CPU demand has replaced demand for GPUs. The two types of processors play different roles in AI systems. 718

Supply is constrained beyond Intel’s factories

Tan’s account also pointed to limits in the wider supply chain. Packaging substrates were identified as one pressure point, alongside the challenge of expanding CPU availability quickly enough to meet customer requests. 425

Intel had also described supply constraints in its second-quarter earnings discussion. The company reported second-quarter revenue of $16.1 billion, up 25% year over year, while its earnings-call transcript said supply constraints persisted amid strong demand. Those results give context to the demand story, but revenue growth alone does not show how much of the shortfall came from any one component or production limit. 2

Tan’s foundry pitch is about diversification—not an immediate fix

Tan argued that the industry should spread manufacturing orders across more foundries rather than rely so heavily on one dominant supplier. He presented Intel’s foundry business as a potential alternative to TSMC as demand for AI chips grows. 20

That argument addresses the structure of chip manufacturing supply. It does not establish that Intel’s foundry can immediately resolve the CPU shortage or the substrate constraint. The reported gap between customer requests and available CPUs remains a capacity challenge, not proof that Intel has already closed it. 12025

How the demand story showed up in stocks

Investors had already responded to the CPU-demand narrative before October 2. Intel shares rose 12.1% on September 21 amid reports linking Meta’s Muse AI agent to expectations for stronger CPU demand. That connection describes the market’s interpretation; it does not show that Muse caused Intel’s reported supply shortfall. 9

On October 2, available market snapshots showed Intel up 4.17% at 2:36 p.m. Eastern, while AMD closed up 2.95% and Arm closed up 5.18%. These are not identical time snapshots, so they should not be treated as a directly comparable closing-price table. 404246

Analyst price targets offered a separate, and variable, point of comparison. One October 2 compilation reported an average Intel target of $114.54, with estimates ranging from $75 to $200 among the analysts it tracked. Price targets are estimates, not guarantees, and the range underscores that the market’s enthusiasm for AI-related CPU demand did not erase uncertainty about Intel’s prospects. 44

The takeaway

Tan’s explanation joins a demand shift with a supply-chain constraint: agentic AI and inference may add to CPU workloads, but Intel cannot currently fulfill all the requests it receives. Substrate availability and broader manufacturing capacity complicate efforts to expand supply. Intel’s foundry pitch is a longer-term argument for more manufacturing options—not evidence that the immediate gap is solved. 141820