The promise behind Intel stake

The promise - that’s what @benthompson talks about in the new piece “U.S. Intel” - a 10% 🇺🇸 stake in @Intel is a commitment device to break the standoff: buyers won’t risk 14A without certainty, and Intel won’t fund 14A without them. 👇

Under the hood, this is about yield learning and irreversible capex and not ideology.

14A lives in the High NA era and follows 18A with backside power (PowerVia). That stack only pays with boring basics: PDKs that hit 1.0, IP, MPW shuttles, solid D0, and fast cycle time.

Without external tapeouts, High NA scanners turn into very expensive museum pieces. The equity moves bargaining power so big silicon buyers believe Intel will keep the node alive long enough to learn.

Still, policy can’t fix execution: PowerVia, 18A PT, and hybrid bonding sound great. They only matter if defect density and reliability match TSMC’s early ramps. The bigger trap is chasing flashy backside power wins while customers need the basics: clean PDKs, libraries that close, and design rules that do not churn.

There is a market risk too: Government ownership can skew procurement and trigger foreign subsidy rules, which makes non U.S. sales harder.

Context helps: TSMC’s N2 to A16 cadence brings backside power at scale and a customer service flywheel built over decades. Intel needs guaranteed demand and time to learn. That is the bet.

If it works, the U.S. gets a real onshore anchor at the leading edge and less single point of failure risk. If it flops, we burn scarce political capital and still depend on offshore capacity.

What I am watching next: 14A PDK hitting 1.0 without churn. How many MPW shuttles and who shows up. First external 14A tapeouts with binding supply deals. D0 and cycle time vs N2 and A16. Packaging capacity onshore for AI stacks. Clear rules on governance spillover.

My read: Risky, but a coherent attempt to fix incentives.

Here’s the article: https://stratechery.com/2025/u-s-intel/