How to Make China Build Better Chips, Faster
China has apparently decided that if the global AI arms race is going to be a farce, it might as well be a high-budget one with customs officials as the punchline. Nvidia’s H200 AI chips, recently cleared for export by the US government after much strategic throat-clearing, have been stopped at the Chinese border. Production lines have paused, suppliers are confused, and everyone involved is insisting they have no idea what’s going on, which is generally how you can tell geopolitics is working exactly as designed.
According to reports, Chinese customs authorities have quietly told agents that the H200s are not allowed in. No formal ban, no written explanation, just a firm bureaucratic shrug. Domestic tech firms have reportedly been summoned and warned not to buy the chips unless absolutely necessary, a phrase that in authoritarian capitalism usually means “we will decide what necessary means later.” Nvidia, meanwhile, has said nothing, presumably because saying anything would only add another layer to the already Byzantine trade ritual surrounding these processors.
This is the same H200 that Washington, after months of hand-wringing about national security, suddenly decided was safe enough to sell to China, provided it takes a scenic route through the United States first. The chips are designed in the US, manufactured in Taiwan, shipped to an American lab for “testing,” taxed at 25% as they pass through, and only then allowed to head toward Chinese buyers. This Rube Goldberg machine of trade policy allows the US government to both claim it is protecting national security and skim a cut of the profits, which is a neat trick if you can keep a straight face while doing it.
The stated logic for allowing the sale is familiar and deeply unserious. Some experts argue that selling advanced US chips to China will slow China’s progress by keeping it dependent on American technology. This is the same argument that has been used, unsuccessfully, for decades in sectors ranging from aerospace to telecommunications. It rests on the charming assumption that a country with enormous state capacity, vast industrial scale, and a long memory for humiliation will respond to restrictions by shrugging and giving up rather than, say, throwing money, talent, and national pride at the problem until it goes away.
History, of course, suggests the opposite. Restricting access to technology has an uncanny habit of accelerating domestic alternatives. Japan did not stop building cars because Americans told it to. South Korea did not abandon semiconductors because someone else got there first. China itself did not decide to dominate solar panels, batteries, or high-speed rail because the West made it easy. But this time, we are told, it will be different. This time, China will surely remain permanently dependent on Nvidia, as long as we periodically fine-tune the tariff spreadsheet.
Meanwhile, those warning against the sale of the H200 argue that the chip is powerful enough to be used in weapons systems that could one day threaten the US or its allies. This concern might land harder if the US were not simultaneously encouraging the sale, taxing it, and pocketing the revenue. Apparently the chip is dangerous enough to worry about, but not dangerous enough to leave money on the table. National security, like everything else, is flexible when quarterly earnings are involved.
China’s move to block the chips may be temporary, tactical, or simply a signal that it is not impressed by Washington’s attempt to have it both ways. It may also be a reminder that Beijing has its own leverage and is perfectly capable of introducing uncertainty into a supply chain that Silicon Valley treats as sacred. If the goal of US policy was clarity, stability, or strategic coherence, it has failed on all three counts.
What is most striking is how little imagination seems to exist in this supposedly forward-looking debate. The idea that “slowing down China” through selective tech exports will somehow freeze a rival economy in place is not just naïve, it is willfully ignorant. The more likely outcome is the one we have already seen: China treats the restriction as a deadline, not a deterrent, and responds by powering up its own chip industry at speed. Not that this has ever happened before, of course. That would require learning from history, which is famously unprofitable.
So the chips sit at the border, the factories pause, and the analysts argue. The AI arms race continues, slightly more expensive, slightly more absurd, and no closer to the comforting fantasy that one side can simply regulate the future out of existence.
