On its third-quarter call, Intel Corporation (NASDAQ: INTC) told investors something it has rarely said so plainly: a couple of foundry customers are ready to be announced within the next two quarters. The market’s reply on October 30 suggested it no longer accepts “soon” as an answer. For Intel’s foundry, everything now turns on one question: who is actually building on 18A and 18A-P, and when will they say so?
Key numbers
- External foundry revenue, Q3 2026: $411 million, up $118 million on the quarter, largely from a one-off program
- Intel Foundry revenue, Q3 2026: $6.7 billion, mostly from Intel’s own products
- Intel 18A cost per wafer: down about 40% since Q1
- 18A share of client CPU revenue: nearly one third, from under 5% in Q1
- Intel 14A: 0.9 PDK released in October; 1.0 PDK targeted for Q1 2027
What Intel has said
Lip-Bu Tan has been consistent on one point: Intel will not name foundry customers itself. If a customer wants to announce, Intel will support it. That is normal foundry etiquette, and TSMC follows the same rule. But it leaves investors reading signals rather than contracts.
The signals from the October 29 call were the strongest yet. Tan said Intel is engaged with a couple of highly strategic customers on 18A-P, that inbound interest from external customers is rising, and that those customers are pleased with the yield. The current work, he said, is on die size and making sure the necessary IP is ready. On 14A, the 0.9 process design kit went to select customers in October, with the 1.0 kit planned for the first quarter of 2027. Then came the line the market will hold him to: a couple of customers ready to announce in the next two quarters.
Manufacturing data supports the confidence. Intel 18A yields improved at a double-digit rate in the quarter, wafer costs on 18A have fallen roughly 40% since the first quarter, and Intel’s own 18A client chips (Panther Lake and Wildcat Lake) now make up nearly a third of client CPU revenue. Clearwater Forest, the first 18A server chip, is shipping for revenue. Intel has also opened Fab 62 in Arizona, its second high-volume 18A fab.
Who is public on Intel 18A and 18A-P
Some names are already on the record:
- Socionext. In August 2026 the Japanese chip designer said it will use Intel 18A-P for next-generation custom SoCs, adding Intel alongside its existing TSMC work. It is the clearest third-party 18A-P commitment so far.
- Microsoft and Amazon. Both announced 18A designs in 2024: Microsoft a custom chip, Amazon Web Services an AI fabric chip. Neither has since given volumes.
- US government programs. Intel completed the Defense Department’s RAMP-C program this quarter, which funded prototypes and early customer engagement on its leading-edge nodes.
- Apple (reported, not confirmed). The Wall Street Journal reported in May 2026 that Apple and Intel signed a preliminary chipmaking agreement, with 18A-P widely reported as the node. Neither company has confirmed it. We cover what is known in a separate article.
Intel’s own timeline adds context. 18A-P’s 1.0 PDK was delivered at the end of 2025, risk production was reported in June 2026, and Intel says 18A-P is on track to launch across its internal products by year-end. Diamond Rapids (server) and Nova Lake (client) are its first 18A-P products. Customers follow the internal ramp: they want to see a node yielding in Intel’s own chips before committing theirs.
Why an 18A customer matters more than a 14A customer
Intel’s long-term foundry story rests on 14A, which enters risk production in the second half of 2027 and high volume in 2028. But a customer on 18A or 18A-P matters more right now, for three reasons.
It fills fabs that already exist. Intel’s CFO admitted on the call that 18A will be somewhat underloaded next year, which is why part of the Oregon 18A line is being converted to Intel 3. An external 18A-P customer would put wafers into capacity that has already been paid for, which goes straight to margins.
It proves the process to outsiders. Intel’s foundry still makes almost all its revenue from Intel. External revenue of $411 million in Q3 was helped by a non-recurring program. A recurring outside customer on a leading node would be the first proof that Intel Foundry can be a true foundry, not an internal factory.
It de-risks 14A. Intel has said it will only build 14A at scale with clear evidence of customer demand. A customer that has shipped on 18A-P is the most likely candidate to commit to 14A.
What to watch for INTC
The next two quarters set a clear test. Intel has said a couple of customer announcements are coming; the Q4 results in late January and the Q1 call in April are the windows. Watch customer press releases rather than Intel’s, since Intel will not name them first. Watch the 14A 1.0 PDK date in Q1 2027. And watch external foundry revenue: a rise without a one-off program would be the quietest, and most convincing, sign that outside wafers are flowing.
Until a name appears, Intel 18A is a process that works for Intel. One credible outside customer would make it a foundry business, and for INTC that is worth more than any single quarter’s earnings.
Sources: Intel Q3 2026 earnings call (Oct 29, 2026) and earnings release; Intel Q4 2025 earnings call; Socionext announcement as reported by TrendForce (Aug 20, 2026); The Wall Street Journal via SiliconANGLE (May 8, 2026); Intel announcements with Microsoft (Feb 2024) and AWS (Sep 2024).
Leave a Reply