If nVidia thinks TSMC isn't investing enough, they are free to give TSMC money to invest in fabs. But I'll be honest, TSMC seems to be the only sane person right now. "Doubling capacity" in 10 years, which would likely cost $1T or so, seems like a risky bet on an industry that can't figure out how to turn a profit.
Discussion
Loading...
Not only could they, there is industry precedent. When Apple started moving from microdrives to flash for the iPods, they realised that they were vastly increasing the demand for flash. They made a deal with their foundry that they would pay the capex for a new fab and, in exchange, get 100% of the output for the first two (I think) years, then discounted rates for a few more years. The foundry ate the opex for a couple of years but then got a free fab.
When the iPod Shuffle launched, it was cheaper than many equal-sized flash drives which weren’t also music players because this arrangement gave Apple a much lower wholesale price for flash than anyone else.
A company that believes it is going to increase demand for a particular kind of IC enough that suppliers need to scale up faster is always able to make this kind of deal. Especially if they’re running at such a profit that they have large cash reserves. You do have large cash reserves from making the profits you reported, right NVDA? You’re not misleading investors with circular financing that is massively inflating your revenue numbers, are you?
This is just more prelude to blaming TSMC for their stupid ideas not working.
I'll leave it to you as to whether racism plays a role in all these American companies choosing a legendary fab to blame, when I'm pretty sure they don't understand how electron holes work.