SMIC just posted record revenue and credited US export restrictions as a growth driver. Zhipu, a Chinese AI startup, doubled its revenue and watched its shares surge 35% on the news. I am not surprised. You cannot embargo an idea, and the history of trying is not encouraging.
The closest parallel is the encryption wars of the 1990s. The US government classified strong encryption as a munition, restricted its export, and spent a decade trying to ensure adversaries could not access cryptographic tools that US agencies could not break. The policy achieved the following: it hobbled US software companies competing in international markets, it chilled open cryptographic research at a critical period, and it gave us the Clipper Chip proposal, which the security community rightly treated as a backdoor masquerading as a standard. Meanwhile, strong encryption proliferated anyway, because the mathematics was public and the talent was global. By the time the restrictions were quietly wound back, the only lasting damage was to the competitiveness of the US companies that had been forced to ship weakened products.
The AI chip situation has the same structure, and it carries an additional risk that the encryption analogy does not fully capture. Restricting China's access to advanced chips does not make those chips less strategically valuable. It makes them more so. Taiwan produces the chips China cannot have, which raises the question of what China might eventually do about that. The export controls are not just failing to contain Chinese AI capability; they may be increasing the military pressure on the very facilities producing the chips. That is an outcome the containment strategy was presumably meant to avoid.
There is also something worth noting about who is framing this as a contest between good and bad actors. The US has this year conducted military operations in Venezuela and Iran. It has moved to ban Chinese cars from the American market, not because they are unsafe or technically inferior, but because they are cheap and good and American manufacturers cannot currently match them. The same logic driving the chip restrictions is driving the EV bans: protect domestic incumbents, frame the competition as a threat, and present self-interest as principle. I am based in New Zealand. I have no particular brief for Washington or Beijing. From here, both look like empires pursuing advantage and dressing it up as something else.
What is actually happening in Chinese AI is worth looking at on its own terms, without the frame. China is producing capable models. It is releasing some of them as open weights, which is more than several US labs are doing. Its chip industry, denied access to the leading edge, is building domestic alternatives at a pace accelerated precisely by the restrictions. Its EV industry built cars that consumers around the world want to buy, and is being locked out of markets by tariffs rather than beaten by competition. The picture that emerges is not of a country being contained. It is of a country being motivated.
None of this requires pretending that China's government has a clean record on human rights or that its AI ambitions are purely civilian. It does not. But the argument for restrictions rests on the idea that the US and its allies occupy higher moral ground, and that ground has been eroding visibly. Bombing Iran and Venezuela while restricting Chinese access to chips and EVs is not a human rights policy. It is a power competition, and calling it something else does not make the restrictions more effective or the reasoning more honest.
The chip export controls will continue to add rocket fuel to SMIC's revenue and urgency to China's domestic semiconductor programme. They will continue to raise the strategic value of Taiwan's fabs. They will continue to cost US chipmakers a large and profitable market. And they will continue to be described as a necessary defence of the rules-based international order, by a government that has spent this year demonstrating a flexible relationship with both rules and order. The encryption precedent ended with the restrictions abandoned and the technology proliferated. This one appears to be tracking the same curve.