At a private dinner a few months ago, Jensen Huang said frankly What I’ve been thinking about for some time. America is far behind China in AI development. Here are some reasons.
Huang starts with the ratio of AI developers in China (he estimates 1 million) to AI developers in the US (20,000). This is a ratio of 50:1. While I think he is overstating China and underestimating the US, I use a different metric that gives the same general results. When you’re reading academic papers about the latest developments in AI, count authors with Asian names1; Count numbers with European names. For now, forget where the authors live or work: maybe MIT, maybe Alibaba. Asian names (including South Asia) will be a significant majority.
Now remember where the author might have lived, and consider the fact that America hung a big “not welcome” sign for immigrants. Forget about “we only want good immigrants”; This is incredibly condescending, and no one of any nationality will believe it, or believe that they will be treated fairly after they arrive. Every immigrant worker living in America or considering coming to America has to consider the possibility that he or she will be in the wrong place at the wrong time, with the wrong skin color, and will end up in a death camp. Are we surprised there are international employees? except? Are we surprised that immigrants are coming in fewer numbers? The $100,000 price tag on H1B visa applications says, “We will admit you only if you make it worth our time.” This is a matter of gangsters, not of a responsible government. America’s ability to train high-quality engineers and programmers and provide them with a high standard of living after graduation has historically been one of its greatest strengths. But given current policies, are we surprised? Fewer international students are coming to America? China has built an impressive network of colleges and universities, especially for engineering and science. Students can get a first-class education without the risks of coming to America, which includes saying the wrong thing on social media and being sent back to the border.
Want radar delivered straight to your inbox? Join us on Substack. Sign up here.
I read somewhere (maybe it was Kissinger) that no empire has survived without the import of intellectual capital. This is true of the Romans and Greeks. This is true of the British. And this is true of America and its huge immigrant workforce. (Unfortunately, providing the necessary skills and expertise never saved immigrants from racism.) law of joy is about companies rather than nations, but it still applies: “No matter who you are, most of the smartest people work for someone else” is almost a restatement of the same idea. If you want to work with the smartest people, you have to bring in as many people from the outside as possible – and even if you’re bringing in outside workers, the supply of good talent that doesn’t work for you will always outweigh the talent you’ve managed to acquire. But now, think about what education means for this talent flow: China no longer needs to export students to the US and hopefully some of them will return. And they are in a position to do so attract talent from elsewhere.
Huang is also correct that restrictions on semiconductor exports have not only failed, but they are also driving China to develop its own technology. China’s domestic GPU industry has almost reached the US level and will no doubt surpass it. The restrictions have had another impact on semiconductor sales that Huang did not mention. What do you do when you don’t have the fastest hardware? You make your software more efficient. You optimize it so that it runs faster and draws less energy. you make it Run efficiently on older hardware Using techniques such as quantization. This is evident in all the recent models from DeepSeek to the recently released ones from China quen-3-max-thinking Or km K2.5. America is stuck in the belief that bigger is better2; China is playing the “smaller is better, more efficient and open” card. Guess who wins?
Electric power also tells a similar story. Top domestic AI companies are talking about building data centers that will require several gigawatts of power capacity. It looks like the plan is to build that generation capacity with coal, gas and nuclear power – good luck. These are the most expensive and inflexible ways of building capacity. The current administration has disrupted the development of solar energy, wind energy, and battery backup. Along with coal production, China also leads the world in solar capacity building. It is also a world leader in the development of solar and wind technology, only partly due to its rare-earth resources. It’s possible that the cost of coal production could drop – after all, coal is a commodity that very few countries want, and a lack of demand could cause the price to decline. But cheap coal and free solar power are not equivalent. If AI powers in the US plan to build large-scale data centers, they will need to come up with better, less expensive sources of power. This is another area in which China is far ahead.
After reports of Huang’s dinner talk were leaked, he retracted comments he made in some posts on X. But the question arises: which version do you believe? I know which version I believe; The evidence—employees, chips, efficiency, power—all points in the same direction.
footnote
- I include names that appear to be Indian and Arabic; India and Arab countries are often ignored. By my count a surprising number of names are Spanish: European, of course, but also the focus of anti-immigrant sentiment in America.
- I see it as a cultural burden, but that’s another argument.
