[an_opabinia]

> First, find several dozen billion dollars.

Some electric car companies have not had trouble tapping into capital markets. Even Transmeta, a fabless IC maker, could tap $275m in 2000.

If you were a super talented engineering firm with the ability to spend $2b on a fab, you could tap finance for it.

> More seriously, find a low-performance niche application and make chips on old technology for it.

This seems like the opposite of any winning strategy, and the opposite of what the government should subsidize. TSMC is so successfully because it relentlessly focuses on performance chips. If you are enjoying American subsidies it better be for the cutting edge, what the hell else are we paying for?

> Differentiate yourself on the basis of being easy for your customers to work with

The only customer that matters is Apple, and your only competition is Samsung and TSMC. Everyone else is using "foreign chip manufacturers" as a way to launder low Chinese labor costs, not because there is inadequate supply or poor competition in specifically semiconductor manufacturing.

Perhaps that's what people want - a manufacturing hub, i.e., assembly of chips, not the chips themselves. I don't know if that's super important.

[qbasic_forever]

Starting from zero and competing directly with TSMC's top end processes right now is basically being at the "President Kennedy said we're going to the Moon" stage of landing on the moon. You are going to need many, many billions and billions of dollars and years of engineering and manufacturing (and failures) to get there. This is not something you just turn on the money spigot and solve the problem overnight. It is spectacular engineering at the smallest atomic level to produce in mass quantity the chips from TSMC today.

[javajosh]

The US has an overabundance of brilliant physics PhDs most of which are probably pretty sick of academia's bullshit at this point. Let's give 2000 of them a 100k/year job plus $5B, cheap/free access to federal land, energy, and water, and see what kind of fab(s) they could make. I bet they'd get to TSMC's level pretty quick. If we offered Taiwan a mutual defense pact, we might even get their help to get there.

[qbasic_forever]

They'll turn it down because they can make $250k a year writing code for Facebook/Google/etc.

And it's really more talent in process engineering that you need, not in the physics of chip building. We already know what has to be done to make transistors. It's the how to do it at scale and with high enough yield that's impressively difficult.

[javajosh]

Haha. No way. I don't think anyone who gets a physics PhD is excited by the prospect of solving ad market problems. Or financial market problems (because you forgot to mention the $500k quant jobs). Physics PhDs aren't, in my experience, particularly motivated by money. How could they be?

A project like this would a) let them use far more of their training than coding ever would and b) it is for a good cause, effectively a peaceful version of the Manhattan project.

[mikebike]

You might want to check the bio of Antonio García Martí­nez, who has a physics background, and subsequently worked in both finance and adtech.

[an_opabinia]

This is the right idea. But like I said, most of the value is in exploiting cheap Chinese labor, not the chips, which is why really none of that would work accounting wise - like you’re talking about a valuable economic enterprise but not one that would produce a lot of profits for shareholders.

[Ekaros]

They have spend decades to fine tune the process to be just right. Just money isn't enough, and then we can also look Intel and their 10nm and that is company with experience, history and some workforce in place. Reaching that level from fresh isn't simple thing.

[ajmadesc]

>What the hell else are we paying for?

Are chips for industrial equipment or cars? When you say cutting edge I think desktop / phone / servers.