Intel's Q3 in Two Halves: Memory Tightness Drags PC Demand as Data Center Logs Fastest Revenue Growth in 15 Years
Two opposing supply curves shaped Intel's third quarter. On the earnings call held October 30, Chief Executive Lip-Bu Tan told analysts that memory pricing and allocation conditions have started to reach customers directly, with the PC market flashing deceleration signals in recent weeks. The counterweight came from the data center and AI portfolio, which stayed resilient and, by his description, delivered the fastest revenue growth in 15 years.
Tan framed the near-term picture as a stack of three linked pressures — manufacturing supply shortages, an unprecedented memory crunch, and the softer PC demand that trails from them — while still expressing confidence in the company's longer trajectory. For industry watchers, the more useful reading is structural: component economics, process capacity and AI server demand are now coupled tightly enough that one quarter can tell two different stories at once.
Snapshot: The Metrics Behind the Split Narrative
| Measure | Third-quarter result | Change or context |
|---|---|---|
| Total revenue | $16.51 billion | Up 21% year over year |
| Non-GAAP earnings per share | $0.52 | Above the $0.38 the company had previously guided |
| Non-GAAP gross margin | 44.1% | Roughly 2 percentage points higher than both the second quarter and prior guidance |
| Data center and AI revenue | $7.3 billion | Up 78% year over year |
| Data center and AI operating profit | $3.0 billion | About 42% of segment revenue; up roughly $600 million sequentially |
| Client computing and physical AI revenue | $8.1 billion | Down 5% year over year |
| AI PC revenue | Growing line within client | Up 35% in the same period |
| Wildcat Lake (cost-optimized client silicon) | Shipments up | Up about 70% quarter over quarter; sold out currently |
| Fourth-quarter revenue outlook | $15 billion to $16 billion | Company guidance |
| 2026 capital expenditure | $20 billion | Planned; a substantial further increase is expected in 2027 |
"Non-GAAP" above refers to figures presented outside U.S. generally accepted accounting principles, which is how the company reported the earnings and margin results.
Where the Growth Actually Came From
Segment mix explains most of the upside. The data center and AI business supplied $7.3 billion of revenue and, unusually for a growth push, converted it at scale: $3.0 billion of operating profit, or about 42% of that segment's revenue. The sequential gain of roughly $600 million in operating profit suggests the expansion is not purely a volume story.
- Granite Rapids: built on Intel 3; Tan described it as one of the company's fastest-ramping server products.
- Clearwater Forest: Intel's first server product on Intel 18A. It has begun shipping and has started to confirm revenue, with considerable demand anticipated in 2027.
The Client Line: A Supply Problem With a Demand Echo
Client computing and physical AI generated $8.1 billion, down 5% year over year — the number that gives the quarter its split character. Tan attributed the weakness to the knock-on effect of memory cost and availability rather than to a single product misstep, and Chief Financial Officer David Zinsner pointed to pockets of momentum inside the same segment.
- AI PC revenue rose 35% in the same period, even as total client revenue declined.
- Wildcat Lake, positioned for cost-sensitive configurations, saw shipments climb about 70% quarter over quarter and is currently sold out — a shortage signal as much as a demand signal.
- Panther Lake, on Intel 18A, has completed its ramp to volume production, according to Tan.
- Nova Lake, the first client product slated for Intel 18A-P, is running ahead of schedule; at launch it is expected to carry a design-win count that is two times that of Panther Lake.
Manufacturing: Yield, 18A-P Timing and the 14A PDK
Process execution was the other area where management offered concrete updates. Tan said yields improved in the third quarter on both Intel 3 and Intel 18A, with the gain on Intel 18A measured in double digits, and wafer output on both processes came in above expectations.
- Intel 18A-P is expected to be used in Intel's own internal products before year-end.
- Intel 14A remains on plan; in October the company released a version 0.9 process design kit to selected customers.
- Beyond its core x86 franchise, Intel continues to fund specialized custom silicon, advanced packaging and external wafer foundry work.
Zinsner quantified one part of that broader push: specialized custom silicon revenue grew more than 50% sequentially and more than doubled year over year. He also noted that all three businesses — custom silicon, advanced packaging and external foundry — remain relatively small in revenue terms.
Guidance, Capex and the Watch List
Forward-looking commitments were as notable as the print itself. Management put fourth-quarter revenue guidance at $15 billion to $16 billion, total 2026 capital spending at $20 billion, and flagged that the figure will rise substantially in 2027. For analysts, the next few disclosure cycles will likely turn on a short list:
- Whether memory pricing and allocation stabilize enough to let PC demand recover rather than merely repress it.
- How long data center and AI momentum holds while the client segment absorbs cost pressure.
- Whether Intel 18A and 18A-P ramp gains translate into external foundry traction, starting with the early Intel 14A design-kit conversations.
- How the widening capex plan is funded while gross margin sits near the 44.1% level just reported.
FAQ
Why did client revenue fall while AI PC revenue rose?
The two can coexist because the constraint sits upstream. When memory is scarce and expensive, finished PC configurations get repriced or deferred, which drags total unit economics; meanwhile the premium, AI-capable tier grew 35% in the same period, indicating buyers are shifting mix rather than exiting the market outright.
How much of Intel's profit is now carried by data center?
On the numbers disclosed, the segment produced $3.0 billion of operating profit on $7.3 billion of revenue — about 42% — which is the clearest evidence that AI-driven server demand is doing more than offsetting softness elsewhere.
What does "sold out" tell us about Wildcat Lake?
Roughly 70% sequential shipment growth plus a sold-out inventory position currently suggests supply, not interest, is the limiting factor for that cost-optimized client line.
Why should a chip designer's outlook depend on memory at all?
Tan's account makes the dependency explicit: manufacturing shortages and the memory squeeze travel together, and the resulting PC demand weakness is the visible output. A processor vendor can ship parts and still see revenue compress if the bill of materials around them becomes unaffordable.
What is the closest thing to a scoreboard for 2026?
Two disclosures frame it: fourth-quarter revenue guidance of $15 billion to $16 billion, and the 2026 capital expenditure plan of $20 billion, with a marked increase signalled for 2027.
The Takeaway
Strip away the single headline number and the quarter reads as a resource-allocation story: one business compounding quickly on AI server demand, another held back by input costs outside Intel's direct control, and a fab roadmap being used as the bridge between them. The next test is whether capacity and yield gains let Intel feed the data center surge without continuing to ration the client franchise.
Statements and figures as presented on Intel's third-quarter earnings call dated October 30; original market reporting reference: wsj.com.