Comment by petu

He said/implied that we would have shortage either way (without data center buildout, but everyone who wants access to LLM does private/local inference), which doesn't sound far fetched.