Breaking
Loading news...

Intel Signed Its First Customer for Its Most Advanced Chips...

Intel signs first outside customer for its 18A chips

Intel has spent five years telling the world it can make chips for other companies as well as Taiwan's TSMC does. On Wednesday, chief executive Lip-Bu Tan said someone finally signed up. Speaking at the TomorrowX Summit in Austin, Texas, Tan told the audience Intel had "just signed the first customer" for its 18A manufacturing process, the most advanced node the company has. He did not name the customer, say how much the deal is worth, or say when production starts. The whole announcement was one sentence on a conference stage, with no press release or regulatory filing behind it.

A milestone with a blank where the name should be

18A is the process Intel's entire foundry turnaround depends on. It uses RibbonFET gate-all-around transistors and PowerVia backside power delivery, two technologies Intel built first for its own chips. Intel technology chief Pushkar Ranade later clarified on X that the customer signed for 18A-P, a refined variant of the node that entered risk production in June 2026, the early low-volume manufacturing runs used to iron out defects before full production. Tan also gave the number that foundry customers actually care about: he said 18A yields have climbed to roughly 70 percent, up from figures he described as horrible when he took over as chief executive in March 2025. Yield is the share of chips on a wafer that come out working, and it decides whether a foundry can make money on a deal at all.

As for who signed, the guessing started immediately. Reports in May suggested Apple was nearing a foundry arrangement with Intel, and trade reporting from October 2025 claimed Microsoft's Maia 2 AI chip would be built on 18A or 18A-P, a report Intel never confirmed or denied. Neither name has been confirmed, and Intel is not saying whether the customer is a household name or a company most readers have never heard of. That gap is the whole story in a way, because a foundry business exists to prove to outside chipmakers that it can match TSMC on yield and reliability, and a named flagship client would carry far more weight than an anonymous one.

Microsoft signed something too, and this one has a name

The same day, Microsoft and Intel put out a joint announcement that is formal where the 18A news is not. Microsoft is the first company to announce it will use Intel Foundry's EMIB advanced packaging technology for its own Azure silicon. EMIB, short for Embedded Multi-die Interconnect Bridge, joins separate pieces of silicon with small bridges set into the base of the chip package, which lets companies build bigger chips out of smaller chiplets. The agreement also confirms multi-year supply of Intel Xeon server processors to Azure, including the next generation of chips, codenamed Diamond Rapids, which Intel has said will arrive in 2027 built on 18A-P.

Two foundry wins announced hours apart, one named and formal and one anonymous and spoken, will obviously invite the question of whether they are more closely related than Intel is letting on. The companies did not link the two, and it is worth keeping them separate: EMIB is packaging, not manufacturing a chip on the 18A process, and the release named no Azure chip, gave no volumes or dates, and disclosed no financial terms. The Microsoft deal is real and documented. The 18A customer is real only as far as one sentence from a chief executive on a stage.

The market liked it, with reservations

Intel shares closed at $57.45 on Wednesday, up 5.09 percent on the day, according to consolidated exchange data. That is a genuine bounce, and it came after a rough stretch for the stock. But a share price is a mood, not a measurement, and the measurement that will settle this story is volume: how many wafers the unnamed customer actually orders, and whether the chips coming off the line work well enough to keep them coming back. Intel's next quarterly filing is the natural moment when a named customer, or confirmation that none is coming, might finally surface.

For now Intel has what it has wanted for five years, a first outside customer for its best process, and it has it in the weakest possible form, a claim with no name attached. The yield number is the part that lends the announcement credibility, because 70 percent is the kind of operational detail you do not volunteer if the line is not working. The customer name is the part that would turn a milestone into proof. Until one of those two things changes, Intel's foundry story is still a promise, just a better-supported one than it was on Tuesday.

---------------
Author: Alan Ward
Seattle News Desk