$INTC has a problem. The kind of problem most companies would kill for.
CEO Lip-Bu Tan told the Cisco AI Forum that "almost every CEO is calling me up" begging for more chips. Buffer inventory? Gone. Sold through in Q4. New supply doesn't come online until late Q1, and even then it might not be enough.
After years of watching Nvidia eat its lunch, Intel is suddenly the bottleneck everyone's trying to unclog.
The Agentic Flip
AI moved past the "throw GPUs at training" phase into agentic inference — deploying thousands of AI agents handling varied, sequential tasks across enterprises. These workloads play to high-core-count server CPUs with large caches and mature software ecosystems, not pure parallel throughput. Intel's Xeon lineup is the direct beneficiary.
Tan put it bluntly: customers are "crying for more products." Q1 server CPU guidance is down sequentially, and CFO David Zinsner confirmed it's entirely a supply constraint, not a demand one.
The demand is structural. The fabs just can't keep up yet.
Pricing Power Returns
When you can't make enough of something everyone wants, you raise prices. Intel is doing exactly that — Reuters reported a 10% price hike on server CPUs for Chinese customers. That's confirmed across multiple sources. You don't make that move from weakness.
Intel's Data Center and AI segment pulled in $16.9 billion in 2025. The prior peak was $26.1 billion in 2020 — before Fab 34 in Ireland, before Granite Rapids on Intel 3, and before 18A existed.
With Diamond Rapids on 18A expected late 2026 or early 2027, Intel has a clear path past that high-water mark if agentic demand keeps compounding.
18A Ships
The "five nodes in four years" skepticism has quieted. Panther Lake (Core Ultra Series 3) is shipping on 18A. Tan called it an "over-delivery." That's a massive de-risking event for both Intel's IDM model and its foundry ambitions.
The foundry play is building momentum too. Tan hinted at "a couple of customers knocking on my door" for 18A manufacturing. The U.S. government, Nvidia, and SoftBank have all invested in Intel's foundry future. External fab revenue was never going to happen overnight, but the pipeline is real.
The stock reflects the shift — up 110%+ from 2024 lows to $46-47, with an 84% run in 2025 and another 26% in January alone.
The Window Has a Timer
Intel's supply crunch has a shelf life. TSMC is expanding capacity. Samsung is pushing Gate-All-Around at 2nm. AMD keeps executing on EPYC server share gains.
Customers don't wait forever. They redesign around whatever's available. The agentic AI tailwind is real, but it's not exclusive to Intel. Every chipmaker with server-grade silicon is seeing the same demand.
Then there's China. A 10% price hike on Chinese customers works until trade policy shifts wipe out that premium overnight. Geopolitical friction remains the wildcard nobody can model.
Bottom Line
Intel at $46 is pricing in a turnaround that's actually happening. Lip-Bu Tan inherited a mess and turned it into a company where CEOs fight for allocation. 18A works and ships. CPU demand from agentic AI is structural, not cyclical. Pricing power is back.
Much of this is in the stock after a 110%+ run. The open question is how much sustained execution can widen the moat before competitors close the gap.
For a company that spent three years being written off as dead, having too much demand and not enough supply is the best problem in a decade.
INTC