Lots of people still use older processor nodes. For example Groq[1] uses 14nm chips for their inference chips. Older fab tech is cheaper to use (less demand and higher yields). Most chips made are on older processes, it's only the bleeding-edge CPU/GPU chips that use bleeding-edge fab tech.

[1] https://groq.com/wp-content/uploads/2024/08/GroqChip%E2%84%A...

I'm assuming TSMC has new and old lines. A 555 chip isn't done on the same line as an AMD Zen 5 chip. Permits, buildout, infrastructure, suppliers, you already get a big warehouse that makes chips and you can bring in your own people to optimize it on day one not year 4.

Intel has a couple of euv machines that tsmc could use and they could take Intel's current nodes and sell them as US made chips without threatening their own top nodes.

Ignoring their newer fabs for a minute (which do exist), because there's still a load of money to be made on reliable older nodes.

This data is a couple years out of date now but it still shows how relevant older nodes are: https://stockdividendscreener.com/technology/semiconductor/t...

> Why would TSMC want to buy factories from Intel?

I think the reason is obvious...it has no choice

because Trump said TSMC is stealing US chip biz, and threaten 100% tariff [1]. I think this is TSMC's attempt to avoid tariff

[1]https://www.pcmag.com/news/trump-intel-was-great-until-taiwa...