Will blocking China from accessing relatively high-performing AI chips help the U.S. maintain global AI leadership? Perhaps counterintuitively, the answer is no. Starving China’s supply of U.S.-designed AI chips will have the opposite effect, as it will push China to more effectively develop and deploy its own AI chip capacity and ecosystem. This, in turn, will weaken America’s leadership in AI and related technologies.
Seesawing policy decisions on AI chip exports
The competing views regarding U.S. AI chip exports to China are reflected in a set of seesawing policy decisions regarding an Nvidia graphics processing unit (GPU) known as the H20. GPUs are critical to the computations used to train and run large AI models. And Nvidia, a Silicon Valley company with a market capitalization of over $4 trillion, is by far the world’s leading supplier of high-performance GPUs.
Nvidia designed the H20 several years ago to comply with Biden-era rules on exporting advanced chips to China. Its performance places it below the threshold that would have run afoul of the Biden administration’s limits.
In January 2025, Chinese company DeepSeek released an AI model that shook U.S. stock markets, demonstrating that Chinese AI capabilities were significantly more advanced than many people had suspected. In response, the Commerce Department imposed new restrictions on sales of the H20 to China. Nvidia explained in a regulatory filing that the government said “the license requirement will be in effect for the indefinite future” and took a charge of over $5 billion to its quarterly earnings due to the negative impact on sales.
Then, in July, after Nvidia CEO Jensen Huang met with President Donald Trump, the company announced that “NVIDIA is filing applications to sell the NVIDIA H20 GPU again” and that the “U.S. government has assured NVIDIA that licenses will be granted, and NVIDIA hopes to start deliveries soon.”
But the government saying that approval will be granted is different from actually granting approval—and leaves the question of what level of H20 sales to China the U.S. government will permit up for debate. One side of the argument is that keeping sales to a trickle will leave China without access to the AI data center computing power it needs to compete effectively, thereby helping to cement America’s status as the global AI leader. But that is naïve and short-sighted for several reasons.
U.S.-imposed scarcity will spur greater AI investment within China
First and most obviously, the more challenges that China faces in obtaining advanced AI chips from the U.S., the more it will invest in growing its own capacity. Huawei has made enormous strides in recent years and has plenty of human and financial capital that it can direct towards AI. Of course, Huawei is already making investments in AI. But those investments will increase in response to the expanded market opportunities within China, globally created by a U.S.-imposed shortage of AI chips.
The overlooked importance of the network edge
Second, and more subtly, the long game in AI isn’t only about high-performance chips in AI data centers. AI is an ecosystem that also includes billions of “edge” devices—everything from mobile phones to laptop computers, connected vehicles, and factory automation solutions that sit at the edge of networks. The AI solutions of the future will increasingly involve approaches that use a combination of computing at both the edge device and the data center.
Huawei has an enormous presence in many markets outside the U.S., a fact reflected in the ubiquitous consumer-facing advertising of Huawei products in regions including Latin America, the Middle East, and Africa. In more technical terms, Huawei has extraordinary penetration at the network edge in much of the world. This large and rapidly growing global installed base of Huawei edge devices confers an advantage to Huawei when companies in non-U.S. jurisdictions are making decisions about how to build their AI data centers.
It’s not an insurmountable advantage. If, as is the case today, America’s AI chips continue to outperform Huawei’s chips, the U.S. will remain a supplier of choice on the global stage. But if America’s best chips are only slightly better than Huawei’s, or not better at all, Huawei’s large installed base of edge devices provides a strong incentive outside the U.S. to build what amounts to all-Huawei AI ecosystems. And if that happens, regardless of how strong the domestic U.S. AI chip market is, the U.S. will have lost the global AI race.
Boosting the incentives for a U.S.-focused AI ecosystem
The best way to ensure that doesn’t happen is to limit the economic incentives for China to further accelerate its AI chip advancements. That means letting U.S. companies such as Nvidia and AMD sell highly capable (though not the most capable) chips into the Chinese market. And more broadly, it requires policy attention to ways to better position the U.S. to be the world’s supplier of choice for many different components of the AI ecosystem—including both AI data center chips and edge devices.