The US has recently escalated its efforts to block Chinese access to AI chips, and of course, Nvidia is at the center of it all. Can you give us the broader context here? What's the goal of the US export bans, and why is Nvidia so central to the strategy?

Look, Nvidia runs and manufactures all the critical chips that people need for training and inference on AI models. They design the chips, those chips get built by TSMC in Taiwan, and they get distributed globally. There's an immense amount of concern in Washington, D.C. these days that those chips are empowering China, and specifically China's military.

So if you really want to unpack it, yes, there's a Chinese commercial part of this that is important, but it's not what is really pissing off policymakers in D.C. It's really the fact of civil-military fusion in China. The companies that are building out the commercial technologies are also using those for the People's Liberation Army, the navy, etc. And so there's huge concern with Nvidia being able to export best-in-class chips to the Middle Kingdom.

You were just at Hill and Valley last week in D.C., which has become a pretty large gathering of tech and policymakers in the US over the past few years. Did this come up at all? What was the topic of conversation?

At the Hill and Valley Forum, it was just all over the place. People are very concerned about the chips. The tension here has always been that Nvidia makes an immense amount of profit from its Chinese sales, right? And this is true not just of Nvidia. It's true of Intel, it's true of Qualcomm. China is one of the largest import markets for semiconductors in the world.

That means American companies that have already done the R&D, the design, and the development are looking at the Chinese market and saying, "Look, this is just pure revenue. This is just great margin to be able to fund back into the R&D and stay competitive." And so there's a fundamental tension: you don't want the Chinese to have access on the military side, but at the same time, we would really like their money to continue subsidizing American supremacy in these technologies. I think there were 10 or 12 panels throughout the day, and China was the unifying theme across almost all of them.

Do people really care about the other players right now, the Intels of the world, or is it really just Nvidia because they're at the forefront of the technology?

I think they're at the epicenter in the sense that they have the most performing chips on the thing most people are concerned about, which is AI models. Now, when you look at those export chips, they do cohesively cover everyone. They don't just cover Nvidia. They cover Intel, they cover AMD. In some cases, they're also covering some of the EDM software, so the Cadences and others in that world. So D.C. has a much more expansive view, but from a headline perspective, Nvidia is definitely the one everyone's concerned about.

Is the ban actually working, from what you've been able to glean?

The ban is porous. One of the challenges is that when you think about compute, things are just getting calculated. These days, thanks to the internet, we can calculate anywhere. So what you've seen over the last two years, even under the Biden administration when some of these rules started getting put in place, is that Nvidia's sales to China have shrunk or slowed down. Meanwhile, sales to Singapore have jumped. Singapore is a small nation of 5.6 million people. We wouldn't expect them to suddenly buy tens of billions of chips.

So there's clearly a sign of either a gray-zone market or a smuggling market, and Singapore and Malaysia are epicenters for this. In some cases, you're just re-importing and re-exporting to China, and you're just handing it over. In other cases, you're holding those chips in Singapore, and Chinese companies are using them in the cloud to process data in Singapore, which is arguably equivalent.

I don't think the law and the rules have really caught up to the fact that it's very hard, the same way that controlling money globally is hard. These are very liquid categories. Being able to just say, "Look, China just can't compute things," is a really hard proposition to enforce.

Can the US realistically plug these holes, or are we just sitting ducks to some extent?

It's always a question of to what degree Nvidia has a trusted supplier set, right? It sells to resellers, and to the degree it actually wants to inspect those resellers and figure out who the ultimate buyer is, it can do so. It doesn't have the incentive to do so. If you're Nvidia, you don't really care. You would just like your chips sold, and to make as much money as possible within the context of the law.

So unfortunately, it really is incumbent on the US government to say, "Look, what are the rules? Are you going to fine some of our most important companies if some of these chips leak into countries of concern?" If you do, now you're fining some of the most important companies at the competitive edge of these spaces, and that's not a really great situation either. So it's sort of a glass half full, half empty situation.

Obviously, Nvidia is playing both sides to some extent. At the end of the day, it is an American company, and a lot of folks feel pretty strongly that Nvidia should be taking the US side here, which I think it is to some extent. It's following all the protocols. But Jensen is also in Beijing, so there's definitely a dance going on. What do you think is appropriate for the position they're in right now?

I tend to be relatively in the center of, look, the most important way to stay ahead of the curve and make sure American companies dominate is to make as much money as possible. You want Chinese dollars, or I guess yuan, but Chinese yuan moving into dollars to buy US chips to continue to fund US R&D in US laboratories. And cutting that off means that, essentially, the Chinese government says, "Look, we have to buy more Huawei. Even though those chips are not particularly good, we're going to keep it domestic, we're going to invest in our own infrastructure, we're going to help their R&D."

Every time we get a dollar from China going into a US company is a good thing. And so I think you see this tension with Jensen and Nvidia. You see this tension with Elon Musk and Tesla, where you have a Shanghai Gigafactory and a massive EV market. Now, that market is in massive turbulence right now, but he is someone else who understands that every dollar that comes here is good for the company. It accelerates the flywheel. It keeps you at the cutting edge of society.

What you don't want to be is Ford, which was essentially kicked out of the Chinese market. You had the CEO going there in September last year and saying, "Wow, the world has passed us by, and we're completely left in the dust."

If we look ahead over the next, call it, one, two, three years, what, if anything, should we be watching for to know whether the bans are working?

I always just focus on outputs. It's the same thing with venture investing, right? You look at the output. What we're talking about here is an input, compute going into the ecosystem, but what you're really asking is what capabilities come out of that ecosystem.

There are a lot of concerns around AI benchmarking today and people cheating on them, etc. But I would be looking very closely at the degree to which US AI models outcompete their Chinese peers, whether that's with the same chips or a different set of chips, which doesn't matter. From the consumer's perspective, the end user's, or the enterprise's, all you care about is the performance: what you're actually paying and what you're actually getting.

So I would look at that feature parity and ask: Are we ahead? Do we have a sustainable lead, so that as long as we continue to invest, we're always ahead? Or has China caught up? Regardless of whatever the policy was, and whether we give them underpowered H20 chips instead of A100s, the reality is that if their AI is just as good, then we have failed.