When the Bureau of Industry and Security published its October 7, 2022 rules, the framing was unusually candid for trade policy: the goal was not to protect American chip companies' profits but to constrain a rival's capability in artificial intelligence and advanced computing. Three years and several rounds of tightening later, there is enough evidence to evaluate the strategy on its own terms — and the evaluation is genuinely mixed.
What the controls have clearly done is impose friction. What they have not done is freeze anything, and the gap between those two outcomes defines the current policy debate in Washington.
Why it matters
This is the most consequential technology policy the United States has pursued in a generation, and its results shape three things Americans care about: the trajectory of AI capability worldwide, the revenue of the US semiconductor industry — which forfeits sales every time the rules tighten — and the precedent for using supply-chain chokepoints as instruments of national power.
The stakes extend past chips. If chokepoint controls work, expect the model applied to biotechnology, quantum computing and energy technology. If they visibly fail — if the targeted capability advances anyway while domestic industry loses the market that funds its R&D — the opposite lesson gets learned. Both parties in Washington are watching the same evidence and drawing opposite conclusions.
How it works
The controls operate at three layers. The first restricts the chips themselves: accelerators above defined performance thresholds — total processing power and interconnect bandwidth — require licenses for export to China, with a presumption of denial. The second restricts the tools: Dutch and Japanese cooperation extends US rules to the lithography and deposition equipment needed to manufacture advanced chips, most critically ASML's extreme-ultraviolet lithography systems, none of which have ever shipped to China.
The third layer restricts people and services — US persons are limited in supporting advanced Chinese fabs — and closes loopholes as they are found: the 2023 update tightened thresholds to capture chips redesigned to duck under the original limits, and subsequent rules targeted the cloud-computing workaround of renting controlled chips remotely. Each closure reveals the core difficulty: the controls target a moving technical frontier administered through a static regulatory process.
Evidence
The primary record is the Federal Register itself: the October 2022 interim final rule, the October 2023 updates, and the December 2024 package that expanded controls on manufacturing equipment and high-bandwidth memory. BIS publishes these with unusual candor about intent. On effects, the public evidence shows friction without stoppage: Nvidia disclosed billions in charges and lost sales from rules affecting its China-specific chips; Chinese labs have continued releasing competitive models, with developers publicly acknowledging compute constraints while demonstrating workarounds like aggressive model-efficiency engineering; and SMIC has progressed to 7-nanometer-class production using older deep-ultraviolet tools, as confirmed by independent teardowns.
Market data shows the adaptation pattern: China has become the largest market for semiconductor manufacturing equipment overall — buying the mature-node tools the controls permit — while a documented gray market moves controlled accelerators through third countries, prompting the 2025 diffusion rule's country-tier system. ASML's own financial disclosures quantify the shift: China's share of its sales surged as Chinese fabs stockpiled DUV lithography ahead of restrictions.
The competing read
Supporters of tightening argue the controls are working as designed: frontier capability is what matters, frontier capability needs frontier chips at scale, and scale is exactly what the rules deny. On this view, Chinese model releases that impress are evidence of talented engineers squeezed by real constraints — and the constraints compound each generation as the frontier moves.
Critics, including much of the US semiconductor industry, argue the controls sacrifice American revenue — which funds the R&D that maintains the lead — while accelerating the creation of a self-sufficient Chinese ecosystem that would otherwise have remained dependent on US suppliers. On this view the policy trades a durable commercial advantage for a temporary capability gap, and the gray market means the gap is smaller than the regulations suggest. Both arguments are coherent; the evidence so far supports pieces of each.
What happens next
Three developments will test the strategy. First, enforcement: whether the United States can police the third-country transshipment and cloud-rental channels without alienating the partners it needs. Second, Chinese lithography: whether domestic lithography programs can produce advanced chips at commercially viable yields, which would retire the equipment chokepoint entirely. Third, the negotiation track: periodic proposals to trade chip access for other concessions test whether Washington treats the controls as permanent architecture or as leverage. Watch the Federal Register, ASML's quarterly China-revenue share, and teardown reports on Chinese flagship phones — between them, they report the policy's real score.