Intel Stakes Its Comeback on 18A-P
The day belonged to Intel, which moved its most advanced manufacturing process — called 18A-P — into "risk production," the early phase that precedes full-scale manufacturing. It's a milestone the chipmaker is positioning as proof that its turnaround plan is on track, and one it hopes will reassert its standing in an industry increasingly shaped by AI.
The technical headline is a transistor design no one has shipped before, debuting alongside early numbers that point to meaningful gains: more performance, less energy. Intel is leaning on those figures to court foundry customers — the outside companies that would pay Intel to build their chips. And there's a tantalizing subplot: multiple reports raise the prospect that Apple could be watching 18A-P closely.
Investors liked what they saw. Intel shares climbed as the 18A-P progress boosted hopes for the company's foundry ambitions, with the limited production start read as a tangible sign of momentum rather than another promise.
Nvidia Builds Out the Physical Layer of AI
While Intel fought for relevance, Nvidia kept extending its reach into the infrastructure behind artificial intelligence. The company broke ground on a new "AI factory" and expanded its data-center platform alongside Hewlett Packard Enterprise — with CEO Jensen Huang framing the buildout as a bet on more jobs, not fewer.
Nvidia's gravity was visible across the day's other stories. Optical-components maker Coherent broke ground on an expanded plant in Sherman, Texas, aiming to quadruple output of the optical chips that move data with light inside AI systems. Nvidia was on hand for the ceremony, and the facility — backed by up to $50 million in funding — is tied tightly to Nvidia's roadmap. In Europe, computing firm Bull and manufacturer Foxconn teamed up to build AI systems around Nvidia's Vera Rubin NVL72 platform, pushing the company's next-generation hardware onto the continent.
Cracks in the Throne
Nvidia's dominance isn't unchallenged. The cloud giants' homegrown chips are quietly chipping away at the company that has long been the undisputed king of AI hardware, powering everything from chatbots to scientific research. It's a slower-moving threat than any single product launch, but a structural one — the hyperscalers designing their own silicon to reduce dependence on a single supplier.
Packaging and the Next Frontier
Some of the most important work happens after the chip is made. TSMC is speeding up a next-generation packaging technology called CoPoS, according to research firm TrendForce. Packaging — the often-overlooked stage where finished chips are assembled and connected — has become a competitive battleground in the AI era, where stitching components together efficiently can matter as much as the transistors themselves.
A Bet Beyond the Data Center
Finally, Qualcomm looked past the server racks. At the Augmented World Expo, it unveiled the Snapdragon Reality Elite, a flagship chipset built to power the next wave of augmented reality headsets and smart glasses — a wager that AI silicon's future is on our faces as well as in the cloud.
The through-line: the AI chip race is no longer just about who designs the fastest processor. It's about who can manufacture, package, power, and deploy them — and Intel, Nvidia, TSMC, and Qualcomm each spent today pressing a different edge of that fight.