英特尔第三季度AI需求强劲但代工业务仍存隐忧
英特尔第三季度AI需求强劲但代工业务仍存隐忧
Conclusion
Intel’s latest results show real returns from AI-related server demand at the product level, but not yet convincing returns on the larger foundry-capacity investment. Data-center growth, improving process costs and positive quarterly adjusted free cash flow are encouraging. The counterweight is that foundry remains loss-making, external foundry revenue is still limited, and management expects capital spending to rise again in 2027. The selloff reflects that gap between strong demand and uncertain, delayed returns—amplified by a weak near-term outlook for PCs.
Data as of November 1, 2026. Intel’s Q3 report, for the quarter ended , was filed on . The latest regular-session quote is the : INTC finished at , down that session.
What is working: AI demand is translating into product revenue
Intel reported Q3 revenue of $16.5 billion, up 21% year over year. Its Data Center and AI group generated , with revenue up . Management said the segment’s operating profit was , and that demand for its data-center products still exceeded available supply. Intel attributed the demand to server CPUs’ role in AI infrastructure and agentic workloads, while warning that supply constraints could persist into 2027 and 2028. These are company statements and forecasts, not proof that all planned capacity will earn attractive returns.
There are signs of improved operating performance as well. Q3 non-GAAP gross margin was , and non-GAAP diluted EPS was . Intel said 18A costs had fallen as yields and cycle times improved. It also said 18A-based products had grown from less than to nearly of client CPU revenue. That is meaningful progress, though it is Intel’s own assessment and does not establish the return on the capital invested.
The unresolved question: are foundry investments earning a return?
The difference between Intel’s product momentum and its foundry economics is stark. Intel Foundry reported in Q3 revenue but an operating loss of . Of that foundry revenue, external customers contributed $411 million; Intel said the increase was primarily from a non-recurring customer program. The larger segment revenue therefore should not be read as equivalent to broad third-party adoption: much of Intel Foundry’s activity supports Intel’s own products.
There are customer footholds, but they do not yet settle the leading-edge question. Fortinet named Intel Foundry for its next-generation security chip, which CNBC reported would be made on Intel 4, an older process than 18A or 14A. Intel’s CEO said on the Q3 call that customer announcements could come in the next two quarters, but until those deals are identified and scale, leading-edge external demand remains a key uncertainty. CNBC’s Fortinet report and Q3 coverage provide the customer context.
Spending is large, and the payback is not immediate
Intel said 2026 gross capital expenditures remained on track at , and that 2027 gross capital spending would be “up strongly.” Equipment purchase orders had increased nearly , according to management. In July, Reuters reported that Intel had raised its 2026 capex forecast from $18 billion to $20 billion as AI-driven server demand strengthened. The spending supports both immediate capacity needs and longer-term foundry ambitions, so its returns will depend on utilization, yields, product mix and whether external customers arrive—not just on current AI demand.
Q3 cash flow offers some reassurance, but not a full return-on-investment answer. Intel reported in operating cash flow and adjusted free cash flow of . The cash-flow statement showed gross additions to property, plant and equipment of in the quarter and over the first nine months. These figures indicate that Intel can currently generate cash while investing, but they do not show whether the incremental fab projects will earn returns commensurate with their cost. Intel also completed a equity offering; that bolsters funding capacity but increases dilution risk. Management forecast a Q4 fully diluted share count of .
Why investors sold: weaker near-term guidance and execution risks
The sharp decline followed a strong Q3 print but a weaker outlook. Intel guided to Q4 revenue of and non-GAAP EPS of . CNBC reported analyst expectations of revenue and EPS. Intel’s CFO said the PC slowdown could continue through the first half of 2027. On the call, management forecast a roughly in the 2026 PC market and planned for roughly a in 2027. The Q4 outlook also included an early estimate of about in revenue impact and of gross-margin impact from an incident at Fab 28 in Israel.
Broader market skepticism is another factor: investors have begun asking whether AI infrastructure spending will produce profits quickly enough to justify the scale of investment, as CNBC described in its October 30 coverage. For Intel, that concern is especially consequential because its capacity spending must be funded before the expected demand fully converts into shipments, cash flow and sustainable foundry profits.
Investor read-through
The constructive case is that server CPU demand is real, Intel’s data-center business is growing rapidly, and the company reports better yields and costs on newer processes. If supply expands into committed demand, product margins hold up, and external foundry customers adopt 18A or 14A at scale, higher investment could support materially better earnings and cash generation.
The cautious case is that current product strength does not yet validate the return on all planned factory investment. Foundry losses, limited external revenue, a larger capital budget, share dilution, PC weakness and execution or supply disruptions leave substantial risk. The evidence supports improving product returns, but an unproven return on the broader foundry build-out. The most useful confirmation points are recurring external-foundry revenue, customer wins on leading-edge processes, improving foundry losses, and cash generation that keeps pace with rising capex.
