Every decade or so, Intel's future comes down to a manufacturing node. In the 2010s the company lost its process lead through the long 10-nanometer delay, and the recovery plan that followed — five nodes in four years — was designed to end with 18A, the process Intel said would return it to the front of the industry. That node is now the hinge on which two things swing: Intel's product competitiveness, and the viability of Intel Foundry as a contract manufacturer for other companies' chips.
18A is genuinely new technology, not a shrink. It combines two major changes at once, which is part of why the industry watches it with both interest and skepticism.
Why it matters
The geopolitical stakes are explicit. The United States has committed billions through the CHIPS Act to ensure leading-edge logic manufacturing exists on American soil, and Intel is the only US-headquartered company attempting it. If 18A fails to reach competitive yields and win external customers, the US remains dependent on Taiwan for the most advanced chips regardless of how much fab construction happens in Arizona — TSMC's US fabs still lag its most advanced nodes at home by design.
For the chip market, a working Intel Foundry would end a decade in which any company wanting leading-edge silicon — Apple, AMD, Nvidia, Qualcomm — had exactly one realistic supplier. Nvidia and others have publicly evaluated Intel Foundry; converting evaluations into orders is the metric that counts.
How it works
RibbonFET is Intel's gate-all-around transistor: the gate wraps the channel on all sides, giving better electrostatic control and lower leakage than the FinFET design the industry has used since 2011. PowerVia is the more unusual move — power delivery wiring is relocated from the front of the wafer to the back, freeing the front side for signal routing and reducing voltage drop. Intel calls the combination industry-first backside power at production scale, ahead of TSMC's equivalent, which is scheduled for a later node.
The risk is that integrating both changes simultaneously is hard, and yield — the share of working dies per wafer — is everything in leading-edge economics. Intel's first 18A product, the Panther Lake laptop processor family, serves as the in-house proof case before external customers commit.
Evidence
Intel's process technology pages describe 18A's RibbonFET and PowerVia architecture and its production status with Panther Lake. The Department of Commerce's CHIPS award to Intel ties billions in funding to milestones across Intel's US fabs, and the US government's subsequent equity stake in Intel, announced in 2025, made the company's manufacturing success a matter of direct federal exposure. Intel's quarterly filings disclose foundry segment results, including the operating losses that quantify the cost of the build-out.
The competing read
Skeptics point out that Intel has announced foundry ambitions before and withdrawn them, that its foundry segment has posted multi-billion-dollar annual losses, and that designing for a new foundry's process is a multi-year commitment customers make only on demonstrated yield, not roadmaps. The counterargument is that Panther Lake shipping on 18A at volume is exactly the demonstration that was missing, and that geopolitical anxiety gives customers a reason to want a second leading-edge source that transcends pure cost math.
What happens next
The sequence to watch: Panther Lake volume and reviews, the 18A-P variant aimed at external customers, and any announced design win from a marquee fabless company. Yield numbers leak through the supply chain before they appear in earnings calls; a sustained report of competitive 18A yields would be the single most consequential data point in American semiconductors this year.