How will this possible government action affect American AI development if a case is brought against Nvidia? Wouldn't a case kill AI development by years, if not decades?
A successful case would presumably reduce Nvidia’s markup on GPUs. That by itself would make GPUs more affordable and more widely available to people working on AI.
The problem is if development of those GPUs has such a high fixed cost that in a more competitive environment revenue does not pay for it. But that seems unlikely to me at a first glance. AMD, Apple, Google, and Facebook all have their own chips. They just don’t have CUDA to go with it, or don’t sell their chips to customers.
Nvidia is dominating with high prices; if you force them to lower their prices, wouldn't it make it even harder for the also-rans to compete? What you are asking for is simply to save money, but that is very different than antitrust issues.
An antitrust case against Nvidia would not lower prices by forcing Nvidia to sell at lower prices. In fact, selling below cost is itself a practice that can be subject to antitrust scrutiny.[1]
Rather, a case like this would lower prices by increasing competition. Other commenters have mentioned that if CUDA opened up, that would help companies like AMD become more competitive. Right now, if you want CUDA, you are forced to buy a GPU from Nvidia. So broader CUDA interoperability would benefit "also-rans" and consumers too.
Other than pricing and any anti-competitive deals or collusion that DOJ digs up, CUDA seems like the obvious place you would attack by forcing Nvidia to open it up. But I don't know if that's likely to be an angle the government would pursue.
Kind of, all the other advantages they have are a result of being in such a dominant position in the market, which they gained more or less through that superior toolkit (CUDA).
It'll probably actually help AI development everywhere.
It'd reduce the US advantage over Europe-- since an AI startup could actually do useful stuff without breaking the bank on training and running the models, but it'd basically be a rebalancing where hardware companies get less and actual AI companies get more.
The problem is if development of those GPUs has such a high fixed cost that in a more competitive environment revenue does not pay for it. But that seems unlikely to me at a first glance. AMD, Apple, Google, and Facebook all have their own chips. They just don’t have CUDA to go with it, or don’t sell their chips to customers.