A record data-center quarter, a 20% PC guide-down, and a $23 billion equity cushion—Intel is growing and shrinking at the same time.
INTC · Earnings Call · 2026-10-29
A Good Quarter, a Bad Guide
Intel’s Q3 was a tale of two businesses. Revenue came in at $16.5 billion, non-GAAP EPS was $0.52 against $0.38 guidance, and gross margin was 44.1%, about 200 basis points above Q2 and guidance. Data center and AI revenue was $7.3 billion, up 17% sequentially and nearly 80% year over year, with $3 billion of operating profit—a 42% margin. Lip-Bu Tan called it the “strongest revenue growth for the segment in 15 years” — Lip-Bu Tan, Chief Executive Officer · 2026-10-29, then immediately flagged the offset: “weakness in CCPG driven by the industry-wide impacts from higher memory prices and limited IT budgets” — Lip-Bu Tan, Chief Executive Officer · 2026-10-29. Client revenue was $8.1 billion, down 9% sequentially; Q4 CCPG is guided down more than 20%. Total Q4 revenue is guided to $15–16 billion, with gross margin of 38% and EPS of $0.25.
That is the change. Three months ago, Dave Zinsner was describing client as a relative bright spot—“client will be flattish” — David Zinsner, Chief Financial Officer · 2026-07-23—with only a softer underlying market. Now Intel is planning for the PC market unit TAM to fall about 12% in 2026 and another 10% in 2027, with second-half 2026 units down more than 20%. The inventory correction is no longer a quarterly nuisance; it is the base case for the next two quarters. Management expects customers to keep flushing inventory through Q4 and Q1, maybe into Q2. That is a different shape than the “supply improves every quarter” message from the July call.
Memory Is the New Macro
The culprit is memory. Intel had already noted memory cost and availability as a client headwind in prior quarters; in Q3 it became the organizing principle of the guide. Lip-Bu described memory shortages as part of an industry navigating one of the most significant supply shortages in history, driven by demand for superintelligence. The company now expects the PC unit TAM to decline roughly 10% in 2027, and Zinsner was unusually blunt: “we are not optimistic about next year in terms of memory. We think that it is going to impact the TAM. We think TAM will be down 10% on a units basis next year.” — David Zinsner, Executive Vice President and Chief Financial Officer · 2026-10-29 He added that Intel does not expect a sustained PC recovery until memory availability improves and pricing normalizes—and that they do not expect improvement in 2027.
The industry continues to navigate one of the most significant supply shortages in the history, driven by the unprecedented demand for superintelligence.
This matters because Intel is not just a server story. Client still generated $8.1 billion last quarter, roughly half of total revenue. The memory price spike is transferring purchasing power away from PC OEMs and toward memory suppliers; Intel is caught in the middle, with a fixed-cost fab base and a product mix that is shifting toward lower-margin cost-optimized parts. The Q4 gross margin guide of 38% reflects lower revenue against that fixed cost base and the Fab 28 accident in Israel, which management sized at roughly $300 million of revenue and 200 basis points of gross margin. The Israel impact is expected to be similar in Q1 revenue, though the wafer scrapping is mostly a Q4 event. That is a temporary hit, but it lands on top of a structural PC reset.
The Server Boom Is Real—and Constrained
The bull case is equally clear. DCAI demand is running ahead of supply, and Intel now expects industry server CPU units to grow more than 30% in both 2026 and 2027. Tan argued the server CPU TAM could rise from about $40 billion today to more than $200 billion over five years as agentic workloads put more CPUs into AI racks. That is a huge claim, but it is directionally consistent with the agentic AI narrative that has been building across the market. The problem is execution and capacity: Intel 3 yields and cycle times are improving, but substrates for EMIB packaging are tight, and the company is still rationing supply. Zinsner said the industry will remain supply constrained “throughout 2027 and into 2028” — David Zinsner, Executive Vice President and Chief Financial Officer · 2026-10-29. Intel will not catch up this year; it already told investors in July that it would “not catch up” — David Zinsner, Chief Financial Officer · 2026-07-23 in Q4 and would be behind.
The mix is also a margin trap. Purpose-built silicon is growing fast—more than 50% sequentially and more than doubling year over year—but it carries margins below the corporate average. Zinsner effectively asked investors to focus on gross profit dollars rather than margin percentage: “And so we’re just going to grow the gross profit. That’ll grow the earnings. It may put a little pressure on the operating margin percentage through next year.” — David Zinsner, Executive Vice President and Chief Financial Officer · 2026-10-29 That is the right economic argument if the growth is real, but it means the P&L will look messier even as the top line improves.
Foundry: Still a Promise, Not a Number
The foundry option remains the most valuable and least proven part of the story. Intel Foundry revenue was $6.7 billion, up 16% sequentially, but external foundry revenue was only $411 million—up $118 million on a non-recurring customer program—and the segment lost $1.9 billion. The company is making genuine technical progress: Intel 3 costs improved about 15% from Q1, 18A costs are down about 40%, and 18A client CPUs have gone from less than 5% of client CPU revenue to nearly one-third. Intel 18A-P is on track for year-end, and 14A’s 0.9 PDK was released to select customers in October. But the commercial proof is still missing. Tan would only say Intel is engaged with “a couple of highly strategic customers” — Lip-Bu Tan, Chief Executive Officer · 2026-10-29 and hopes to announce packaging customers in the next two quarters. That is the same “wait for the customer to announce” posture as prior calls: in April, he said Intel had “no plan to announce the customer unless the customer wants to announce it” — Lip Bu Tan, Chief Executive Officer · 2026-04-23. The market has heard that for several quarters. The equity offering in August—$23 billion—bought Intel time and credibility, but it also diluted shareholders and raised the bar for external foundry wins. Gross capex is still $20 billion for 2026, and 2027 will be up strongly, with firm purchase orders to equipment suppliers up nearly 40% sequentially.
The Tape Is Voting on the Cycle
Intel’s stock is down about 30% over the last 90 days and roughly 35% from its June peak. That is not a verdict on Q3’s revenue beat; it is a verdict on the forward shape. The global AI data center theme has also cooled: the tape shows the cluster down about 5% over 90 days with far more decliners than advancers. Intel is therefore fighting two de-ratings at once—a broad AI hardware pullback and a company-specific PC reset. The valuation still assumes a successful turnaround: at the last filed quarter, Intel traded at 12.3x revenue against a business with negative operating income and effective net cash of -$13.5 billion. The $54 billion cash pile and $15 billion revolver are real, but 2027 is now explicitly an investment year with “probably not a dynamic where we’re going to have positive free cash flow next year” — David Zinsner, Executive Vice President and Chief Financial Officer · 2026-10-29.
The story is not that Intel is broken. It is that the AI server opportunity arrived faster than Intel could build supply, while the PC business—the cash engine that funds the turnaround—is being squeezed by a memory cycle Intel does not control. Management is choosing to invest through it, protect the foundry roadmap, and accept negative free cash flow. That can be the right long-term call. But after an enormous 2026 rally that left the stock up roughly 170% year-to-date as recently as mid-October, the shares are now 35% below the June peak. The market is no longer paying for the promise alone. It is asking for external foundry customers, normalized PC demand, and positive free cash flow—none of which management was willing to put a date on.