Nvidia Influence Weakens US Efforts to Restrict AI Chips to China
AI chips and upstream semiconductor equipment is America's most powerful and durable moat, and if the US was seriously focused on maintaining a competitive advantage, it's the easiest thing to protect (yes some background rate of smuggling, but can be reduced from present).
But Nvidia are the world's biggest company and a powerful influence in DC. They make money by selling chips into China. A lot of money. Money that gets turned into ballrooms. They also make money by selling chips to the US AI companies. The US AI companies make more money and can buy more chips with the 'boost' of the securitised 'race with china' narrative (through favourable regulation that saves them money, federal investment, infrastructure buildout support). They can even float industry backstops (as was briefly tried).
So you can play both sides. Everyone's stock goes up. Everyone wins. It all makes sense as long as you're not modelling it as a rational competitive strategy for the US in relation to its closest technological rival.
Now I'm not cynical enough to say anyone's doing the following. And I don't think anyone's cynical enough to do it. But to be spicy for a sec: one can't help but observe the obvious. The more a tight AI race actually came into reality, the stronger the financial and deregulatory support for companies (especially as AI gets adopted into defense & security). Maybe the backstop proposals would fly. The more chips Nvidia would sell. And the best way to make that happen?
Well, yes, exactly.
So as ludicrous as this situation is, it's not really in anyone's interests to oppose this super strongly is it?(1) You're making an enemy of the company whose chips are your absolute critical lifeblood. And (2) you're undermining your own commercial interests. If you're stubborn and idealistic (*cough, Anthropic) you might do it anyway. But if you're not, it really doesn't make much sense.
Why is US policy in such a mess on this? Incentives are at least a part of it. Maybe a big part. But lots of money will be made. And yes, the US lead might suffer. But the biggest loser?
Safety of these unprecedentedly powerful models. Robust oversight. The interests of the Public, and their right to benefit from AI and be protected from its harms. And let's face it, who cares about that?