Samuel Hammond U.S. Companies Are Helping China Win the AI Race America must strengthen its export controls on chips.
https://www.city-journal.org/article/artificial-intelligence-china-deepseek-nvidia-broadcom-openai
How secure is America’s lead in artificial intelligence? It wasn’t long ago that the conventional wisdom put China years behind the curve. This was partly because the United States introduced export controls in 2022 and 2023 that notionally embargoed Chinese companies from the most advanced AI chips.
Today, however, most experts I speak with believe China trails the U.S. on AI by at most six to nine months—if at all. Take the sudden rise of China’s DeepSeek, the impressive AI startup that began as a hedge-fund CEO’s side project. DeepSeek shocked the markets earlier this year by releasing R1, a “reasoning model” that replicated top AI firm OpenAI’s breakthrough o1 model only a month after the latter’s unveiling, and seemingly at a fraction of the cost
OpenAI’s latest models still top the leaderboards, but only because they have access to vastly more computing resources. What DeepSeek nonetheless demonstrated is just how few technical barriers stand in the way of competing at the AI frontier. As DeepSeek’s CEO Liang Wenfeng put it in an interview last year, “money has never been the problem for us; bans on shipments of advanced chips are the problem.”
In practice, though, U.S. chip export controls have been incredibly leaky. Chinese buyers have circumvented the controls on a mass scale by routing chips through third parties in nearby regions, the Wall Street Journal recently reported. In other words, East Asian allies and U.S. firms are enabling the transmission of chips to a geopolitical adversary. America’s economic and military advantage depends on government stepping up to stop the flow.
The evidence for mass export-control evasion was always right under our nose. When the chip controls were first introduced in 2022, Nvidia responded by releasing a new, slightly inferior chip tailor-made for China—the H800—within weeks. Controls were then tightened in 2023 to address the workaround, prompting Nvidia to develop another China-specific chip, the H20. In the interim, Chinese firms hoarded as many H800s as possible, while Nvidia’s sales to Singapore skyrocketed. Singapore now accounts for over 20 percent of Nvidia’s total revenue, while the broader Sinosphere—including Taiwan and Malaysia—accounts for just under half. Many of these chips get diverted directly into China through networks of resellers, while the remainder go to cloud providers that provide Chinese companies with cross-border access to the hardware remotely.
Earlier this year, the U.S. government launched a probe to investigate whether DeepSeek acquired controlled chips through Singapore; the Singaporean government recently followed suit. Yet to industry observers, the complicity of American multinationals in export control violations is an open secret. Thus, when the CEO of the American semiconductor giant Broadcom was asked on the firm’s latest earnings call whether its products were being diverted into China, he gave a knowing laugh, before replying simply, “no comment.”
The stakes from weak export-control enforcement couldn’t be higher. China’s military relies on advanced AI hardware to conduct cyber-attacks, train autonomous weapons systems, and process intelligence at scale. And with Artificial General Intelligence—AIs that match or exceed human performance in every domain—just around the corner, access to large amounts of processing power may soon translate into transformative rates of economic growth and technological innovation, giving the leading AI superpower a decisive geopolitical advantage.
China’s government understands this, which is why it recently announced a $1 trillion yuan ($138 billion USD) state-backed VC fund to finance a build-out of AI data centers—Beijing’s answer to OpenAI’s Stargate project. The project’s chips have got to come from somewhere!
As it stands, the agency charged with enforcing export controls, the Bureau of Industry and Security, operates with limited resources and antiquated technology. There are only two export enforcement officers for the whole of China, and only one for the rest of Southeast Asia, making it virtually impossible to verify the final destination and end-users of U.S. technology.
Short of hiring more enforcement officers, various policy options exist for making their job easier. Chip makers could be required to export through certified logistics providers, for instance, making exports easier to track. Better yet, chips destined for countries like Singapore, Taiwan, or Malaysia could be required to enable ping-based location verification, deterring chip-smuggling operations in the first place.
Despite enforcement gaps, export controls remain America’s best tool for maintaining its strategic edge in AI. The reason is simple. Modern AI systems require just four main ingredients to produce: algorithms, data, chips, and energy. China is at parity or ahead of the U.S. in each of these areas but the chips, with the next biggest bottleneck being energy. And while the U.S. power grid is at or near capacity, China added a record 429 gigawatts of new energy generation to its grid in 2024, a 21 percent year-on-year increase. Were it not for export controls on AI hardware, in other words, China would be able to import leading-edge chips by the millions, with no shortage of energy to power them.
America may still have the upper hand, but without swift action, we are currently on track to surrendering AI leadership to China—and with it, economic and military superiority.
Comments are closed.
