Intel pushes 18A-P into risk production to prove its manufacturing pitch

Intel said on Tuesday that the next iteration of its 18A manufacturing process, designated 18A-P, has entered risk production, the stage at which a chipmaker runs the process on real hardware to prove it works before committing to mass volumes.

The announcement is less about a single chip than about a promise Intel has spent years struggling to keep: that it can still build leading-edge silicon on schedule, on American soil, and sell that capability to other companies.

The figures Intel put on 18A-P are incremental rather than dramatic, which is what you would expect from a refined version of an existing node. Compared with the base 18A process, the company says 18A-P delivers 9% higher performance at the same power, or 18% lower power at the same speed, along with better thermals and more design flexibility.

Those are the trade-offs that matter to the customers Intel most wants, the ones deciding whether to trust it with their own designs.

That is the real audience here. Intel has poured money and credibility into turning its foundry arm into a credible rival to TSMC, with limited success and heavy losses, and 18A is the node meant to change the story.

Moving its successor into risk production is the kind of milestone that lets Intel argue it is executing rather than promising, a distinction external customers have learned to scrutinise after years of delays.

Alongside the process news, Intel introduced the Core Ultra Series 3, which it described as its first AI PC platform built on the 18A node and designed and manufactured in the United States.

The domestic-manufacturing line is doing deliberate work. Intel operates the only leading-edge fabrication on American soil, a position that has drawn both a US government equity stake and the interest of would-be customers looking for an alternative to relying entirely on Taiwan.

The courtship has been visible all year. Apple has held early talks with Intel and Samsung about a second source for some of its processors, Google and Nvidia have been reported eyeing Intel as a TSMC backup, and the company has signed on as foundry partner for Elon Musk’s $25bn Terafab project.

None of that converts to revenue until the manufacturing delivers, which is exactly what Tuesday’s announcement is meant to demonstrate.

The wider backdrop is a company that has been substantially reshaped over the past year. The US government took an equity stake worth billions, Nvidia has invested, and Intel restructured its foundry operation into a separate subsidiary, concentrating engineering on 18A as the first leading-edge logic process built entirely in the United States.

Each of those moves was, in part, an attempt to give external customers confidence that Intel’s manufacturing arm would still exist, and still be advancing, by the time their own chips were due to tape out.

The caveat sits one node further out. Intel has signalled it could pause or abandon its pursuit of 14A and other next-generation processes if it cannot secure enough committed demand, a reminder that the foundry strategy is contingent, not guaranteed.

18A-P entering risk production is a genuine step. If it persuades a marquee customer to sign is the question the chip itself cannot answer.