I think they were really planning to do that. They wouldn't need $100B otherwise. Fabless chip company would cost 10-100x less, even if they order millions of chips right-away. You just wouldn't be able to spend THAT amount of money. Cutting-edge lithography on the other hand...

If you’re pitching a deal of epic scale (“AI is here and we’ll be at the center of its hardware”), you’d include roadmaps to complete independence and vertical integration. People want to see numbers for the big win before they place their bets.

So you’re probably not launching there, but of course you use it as the first horizon you’ll try sailing towards.