
What China gets from its AI honeypots and smuggled chips
You might want to start casting the film right now: Jensen Huang, the Taiwanese-American CEO of Nvidia, the world’s largest technology company, is drawn into an illicit affair with a Chinese spy. The world’s most valuable technology – Huang’s super-powerful computer chips – is smuggled into China or installed in data centers connected to America’s biggest rival. Above all, the possibility of super-intelligent AI, a truly artificial mind, looms – and its creator will be able to rule the world.
At least that’s the simple version of the story. But the real picture of the great geopolitical AI race is both more complex and more incredible than the story of spies and smuggling currently playing out in the press. China wants US tokens – that much is clear. But why? And what can our only true rival power really do with it? To answer this, rolling stone spoke with two experts in China’s tech sector, who lifted the lid on what the superpower really wants when it comes to AI – and why it’s willing to resort to sex, smuggling and lies to achieve it.
Broadly speaking, there are two ways to think about AI, says Kyle Chan, an expert on Chinese technology development at the Brookings Institution.
The first is a direct race towards a single goal: to create a truly intelligent AI that outperforms human intelligence in all general tasks and can improve autonomously. In the industry, this is called AGI, or Artificial General Intelligence. This view holds that the first country or company to achieve true AGI will have an earth-shattering advantage over all competition in economics, warfare, medicine, and technology. Basically, this presents the world of AI as a free-for-all. Many US officials – including at times Trump himself – have made this argument, alongside many private sector CEOs currently working on this technology.
“If we start from this assumption, then efforts to slow down China are almost existential,” Chan says.
America’s biggest advantage is Nvidia’s chips. The world’s most powerful computer chips are made in a single factory in Taiwan, using incredibly complex and rare machines made by a single company in the Netherlands. This is a process so detailed and scientifically complex that it might as well be magic. Sure, other companies can make chips, but none of them can make chips as well as the sole manufacturing company in Taiwan that Nvidia and most US tech giants use. AI companies need these chips, and these chips only, to run the most advanced AI models — which is why the United States has done much to limit China’s access to these chips, establishing export controls on who Nvidia can sell its chips to and which countries they can ship to.
Because of this constraint, the Chinese don’t really see AI in the same way as most American manufacturers.
Instead of an existential threat, Chan says, the second way to look at AI is as a very powerful technology that won’t necessarily lead to a massive strategic advantage for the first team after the AGI position. This makes the AI race a longer-term game, decided more by how power can holistically integrate advances in supercomputing into its society and economy, something China has largely focused on.
“What we see developing in the future are two different markets,” said Kenton Thibaut, a China expert at the Atlantic Council’s Digital Forensic Research Lab. The United States focuses on “frontier” systems – absolute cutting-edge AI, powered by its chip advantage – while China provides “near-frontier” technology that touches different parts of the AI technology market. In China, Thibaut said, AI is seen as a “general purpose” technology that should be harnessed for practical use as quickly as possible, rather than a more conceptual race to the farthest frontiers.
“All (Chinese) policies, major political statements and planning around AI have to do with the question: ‘How can we make this useful in society?’ “, declared Thibaut. “I haven’t seen anything like this in the United States.”
“The (Chinese) version of victory is to be able to benefit as much as possible from AI,” says Chan. “Integrate it into their economy, automate their factories, accelerate drug discovery, et cetera. (They want it) to help them win in all these other areas, rather than just focusing on achieving AGI.”
This is a philosophical difference fueled by the physical constraints China’s tech sector faces. The United States has access to more computing power than any other country, allowing it to push the boundaries. Companies like Anthropic have called on the U.S. government to regulate them more, fearing their AI models could spiral out of control. In China, these discussions are rare, Thibaut says, because the material they work with makes them basically moot — and the Chinese government is predisposed to preventing any disruption to society. Instead, Chinese AI companies generally strive to do more with less – creating functional and practical AI products and applications that make the most of the computing power they can harness.
And in some cases, that’s where the spies come in. The Nvidia honeypot scandal was widely controversial because it involved a woman who ran a Singaporean shell company that set up data centers in Southeast Asia with superchips made in Taiwan, then leased the use of those chips to Chinese companies. Chan refers to this rift, and the broader Chinese plan to get its hands on American technology, as a “bridge” to the country’s true goal: separating itself from the Western “tech stack” while retaining competitive computing power. Eventually, Chan says, China wants to have its own version of Taiwan’s super-manufacturers, its own version of the best chips — but until then, it must use more subtle methods to get what it wants.
“If they want to keep up today, they still need access to Nvidia chips,” Chan says.
Parts of the U.S. government could help them get there. Commerce Secretary Howard Lutnik was very lenient toward Nvidia and Huang personally, allowing the company to continue selling to companies that the United States knew were linked to China. While Trump has said he views AI as a winner-takes-all race, Reuters reported that his administration also approved the sale of 900,000 AI chips to Chinese companies over a six-month period, and his own AI czar, David Sacks, has explicitly advocated selling more chips directly to China.
So China has allies to go along with its spies – and it seems that one way or another, it’s getting all the chips it needs.
Gn bussni