global geopolitics

Decoding Power. Defying Narratives.


America’s Chip Control Campaign Is Winning the Wrong Race

Washington Keeps Closing Loopholes. Beijing Keeps Building Fabs

Opinion Analysis | November 2026

The remote-access rule published on 19 November 2026 is the fifth or sixth iteration of a policy that began in October 2022, and whether it will work can only be judged by asking why the previous iterations didn’t. That is the honest way in. The rule itself, a licence requirement for data centres providing significant computing capacity to firms headquartered in China or Macau, a 90-day window to obtain authorisation, five additional countries brought under licensing for advanced chips is a technical instrument. The interesting question is what it reveals about a coercion campaign that has now run for four years without producing the outcome it was designed to produce.

Walk back to where the escalation actually began. Before October 2022 there was a decade of targeted measures, entity lists, the Huawei campaign, the ZTE case, but those were sanctions against specific firms for specific conduct. The October 2022 rule was categorically different. It was the first time the United States attempted to deny an entire class of technology to an entire country, not because of anything that country had done but because of what it might become. That was a break in the post-Cold War consensus, and everyone involved understood it as such.

The 2022 package restricted exports of advanced computing chips above a performance threshold, along with semiconductor manufacturing equipment and the American software and personnel needed to operate it. The theory was chokepoint leverage. The United States controlled the design tools, the lithography, the design IP and the most advanced fabs, and if you control every chokepoint in a supply chain you can strangle the downstream. It was a plausible theory and it was almost immediately undermined by its own transparency.

Within months Nvidia had designed the H800 and A800 to sit just below the performance thresholds, legally compliant, functionally near-identical for training large models. The rule had been written against a published specification, and the specification was public. The target adjusted. October 2023 tightened the thresholds, moved to performance-density metrics and added entities. Nvidia designed further compliant parts. December 2024 added high-bandwidth memory controls and expanded the entity list again. January 2025 brought the AI Diffusion Rule, the most ambitious attempt of the sequence, a three-tier global architecture that would have governed where American compute could go and under what conditions. It was rescinded in May 2025, four months later, under pressure from allies, industry and internal disagreement about whether it was enforceable at all.

That rescission is the most instructive data point in the whole sequence, and the architects of the current rule will have studied it closely. The AI Diffusion Rule failed not because its objectives were wrong but because it tried to do too much, too fast, with too many actors whose cooperation it needed and whose consent it had not obtained. The current rule is narrower, more targeted, and attaches the control to a service rather than a good, which means it can be enforced through licensing rather than through border interdiction. That is a real design improvement. Whether it is enough is a different question.

The honest assessment of whether any of this has worked has to start with Huawei, because it has been running longest and it is the clearest case. The 2019 and 2020 measures did substantial damage. Huawei’s handset business collapsed in Western markets, its 5G position was rolled back in Europe and Australia, and its revenue fell sharply. By any tactical measure that was a success. The strategic result was the opposite of what was intended. Huawei rebuilt around domestic fabrication, developed its own operating system, reoriented to the home market, and by 2023 was shipping a 7nm chip in the Mate 60, achieved without EUV, using DUV multipatterning, at lower yield and higher cost, but achieved. The campaign did not eliminate Huawei. It forced Huawei to become the thing the campaign was meant to prevent: a self-sufficient Chinese technology champion with captive domestic demand and state backing.

The chip controls have followed the same pattern at a larger scale. China’s imports of advanced chips fell after 2022, but stockpiling had preceded the controls by more than a year, and transshipment networks through Southeast Asia and the Middle East emerged within months. The controls imposed real costs, lower performance per watt, higher prices, delays in deployment, but they did not stop the development of domestic capability. SMIC’s 7nm process is the proof of concept. It is not competitive with TSMC’s leading edge on cost or yield, and it does not need to be. It exists because the alternative was to stop, and stopping was not an option the Chinese system was prepared to accept.

So the pattern across four years is consistent. Each round imposes costs, the target absorbs them, the target finds a workaround, the next round closes the workaround, and the cost of the workaround rises. The question is whether that ratchet eventually reaches a point where the target cannot absorb the cost, or whether the target’s capacity to substitute grows faster than the controls’ capacity to constrain. That is a game-theoretic question, and it is where the analysis gets genuinely interesting.

The campaign is a repeated game with incomplete information. The United States is playing a chokepoint strategy: leverage control of critical nodes to impose costs, with the implicit threat of escalation. China is playing a substitution-and-time strategy: absorb the costs, build domestic capacity, wait for the chokepoints to erode. The outcome depends on three variables, and none of them favours the coercing party as cleanly as the policy’s architects assume.

The first is the discount rate. The American political system operates on a two-to-four-year cycle. The Chinese system operates on a generational one. A coercion campaign that requires ten years to produce its intended effect is structurally vulnerable to a political system that evaluates it every two years. This is not a partisan observation; it is structural. Every administration inherits the previous administration’s controls and faces pressure either to show results or to change course. The AI Diffusion Rule lasted four months. The current rule has a 90-day window. The tempo of the policy is set by the political calendar, not by the technology, and the technology does not care about the political calendar.

The second is credibility. For coercion to work, the target has to believe that compliance will be rewarded, that if it meets the current demands, the pressure will stop. The ratchet structure destroys that belief. Each round is explicitly justified by the failure of the last, and each round is announced as necessary precisely because the target found a workaround. A rational target observing this sequence concludes that compliance is pointless, because the next round will come regardless of what it concedes. So the target builds for the next round, which produces the workaround that justifies the next round. This is a commitment failure, and it is built into the design of the policy rather than incidental to it. You cannot run a ratchet and credibly promise it will stop ratcheting.

The third is the behaviour of third parties, and this is where the current rule is most exposed. Indonesia, Malaysia, Singapore, Thailand and Brazil are being conscripted into a compliance regime they did not write. They are being asked to choose between access to American technology and access to Chinese investment and markets, and the choice is being made for them rather than by them. Some will comply because the American technology is more valuable in the short term. Some will hedge, building dual supply chains and tolerating leakage. Some will resist quietly. The more coercive the enforcement, the more the relationship looks like extraction rather than alliance, and the more those countries will price that extraction into every future negotiation on trade, basing, standards and everything else. This is the classic overreach dynamic in hegemonic coercion: enforcement costs rise faster than compliance gains, because each new demand requires more monitoring, more enforcement and more political capital to sustain.

Now, will the current rule succeed? It depends entirely on what success is defined to mean. If success means preventing China from deploying advanced AI compute at scale, the answer is almost certainly no. The domestic substitution trajectory is too far advanced, the state funding is too large and the demand is too strong. If success means slowing deployment by two to five years in the most advanced applications, the answer is probably yes — the controls are imposing real costs and the costs are rising. If success means preventing military application, the question is largely irrelevant. The military does not need cutting-edge commercial chips. It needs volume, reliability and domestic supply, and China is building all three.

The more important question is what the controls will produce that was not intended. The most likely outcomes are these. A bifurcated global compute market, with an American sphere governing access and a Chinese sphere governing deployment, and a contested middle where countries hedge and arbitrage. An acceleration of Chinese domestic capability at a cost premium, because the controls have functioned as a subsidy to the Chinese semiconductor industry by guaranteeing it captive demand and a protected home market. A more expensive and less efficient global AI build-out, because fragmentation raises costs for everyone and the costs fall hardest on the countries with the least capacity to absorb them. And an erosion of American soft power among the states being conscripted, which will show up in trade negotiations, standards bodies and basing agreements for years.

There is also the boomerang effect that the policy’s architects rarely discuss. Nvidia’s China revenue has fallen sharply. That is revenue that funded research and development for the next generation of American chips. Meanwhile Chinese firms have captive domestic demand and state funding, which means they can iterate without needing to win in the global market. The controls are, in effect, a transfer of competitive advantage from the American ecosystem to the Chinese one, paid for by American firms and justified by a strategic logic that may or may not hold. The lobbying against the equipment ban by Amazon, Google and Meta is not merely rent-seeking. It is a signal that the firms bearing the cost of the campaign do not believe the campaign’s benefits will accrue to them.

None of this means the controls are irrational. There is a real strategic problem. The deployment layer of artificial intelligence is where the industrialisation of intelligence actually happens, and if that layer is dominated by a strategic competitor, the consequences are real and the chokepoints are genuine. The question is whether the instrument, a licensing regime enforced through third-party compliance, layered on a ratchet that cannot credibly commit to stopping, can achieve the objective without producing costs that exceed the benefits.

The honest answer is that nobody knows, including the people imposing the controls. What we do know is the pattern. Four years of escalation, each round justified by the failure of the last, costs rising on both sides, no sign of convergence, and a target that has responded to every round by building the capacity the round was meant to deny. That is what a ratchet looks like when it is working, and it is also what a ratchet looks like when it is stuck.

The endgame, on current trajectory, is a world with two technology spheres and a contested middle. That is not the outcome anyone in Washington set out to produce. It is the outcome the structure of the policy makes most likely the ratchet, the commitment failure, the conscription of third parties, the boomerang on American firms. Whether that world is better or worse than the alternative is a question history will answer, not the people writing the rules. The one thing that can be said with confidence is that the instrument will shape the outcome as much as the objective, and the instrument chosen here is coercion exercised through the compliance departments of foreign companies. That is a way of ordering the world, and the world will remember it was ordered that way.

Authored By: Global GeoPolitics

If you prefer to make a one time donation in support of my work, you can do so by clicking any link below:

https://donate.stripe.com/7sY8wPfvi864don0k3awo07 |

https://buymeacoffee.com/ggtv |

https://ko-fi.com/globalgeopolitics |

Bitcoin: 3NiK8BoRZnkwJSHZSekuXKFizGPopkE7ns



Leave a Reply