First floated in 2021, the 2nm-class node was meant for top-shelf CPUs and GPUs, but with delays to its follow-up 14A node, Intel is now rejigging 18A to cover more ground.

The x86 outfit unveiled 18A-P for mass-market chips with an extra eight per cent performance-per-watt over base 18A, and 18A-PT, which adds through-silicon vias (TSVs) for high-stacked, AI-heavy multi-die packages.

At Foundry Direct Connect 2025, Intel Foundry SVP Kevin O’Buckley (pictured) said: “This is what our customers are telling us they need us to deliver for them. It's a 3D construction with multiple fully-utilised base reticle dies with many compute dies stacked on top… surrounded by an extraordinary massive memory capacity… both electrical and optical interfaces with tens of terabits of bandwidth.”

It is a silicon sandwich and looking remarkably like AMD’s MI300X, built using Intel’s EMIB and Foveros packaging tech, with a nod to affordability. That bit matters, as Intel Foundry posted a $2.3 billion operating loss last quarter.

Chief global ops officer Naga Chandrasekaran added that moving from base 18A to the newer 18A-P variant should be “seamless” for customers already evaluating designs.

Backing the plan, Intel is linking arms with Amkor for packaging, mirroring TSMC’s own US CoWoS plays. The bigger goal: ride Washington’s pro-US semiconductor push. As CEO Lip-Bu Tan put it:

“We are the only company that does advanced, leading-edge semiconductor R&D and manufacturing in the United States.”

Tan, who’s trying to finish what Pat Gelsinger started, was keen to show he’s all about pragmatism now. No more moonshots. Just steady trust-building in a “service business.”

“This is truly a service business, and that is built on the foundational principle of trust and you have to be patient to earn your trust,” Tan said.

He’s also not hiding the axe calling it “De-laborising” and it will remove 20 per cent of Chipzilla’s workforce.