US semiconductor export controls on China are a set of Commerce Department rules, imposed in stages since October 7, 2022, that block the sale of the most advanced chips, chipmaking equipment, and related expertise to Chinese buyers — and reach foreign-made products built with American technology through the extraterritorial foreign direct product rule. The December 2024 update alone added about 140 Chinese entities to the restricted Entity List and extended controls to high-bandwidth memory.
This explainer covers the machinery: which tools the government uses, how the thresholds work, where the documented loopholes were, and what the enforcement record shows. It relies on Federal Register rules, Bureau of Industry and Security statements, and congressional research, through the first half of 2025.
Which legal tools does the government use?
Three do most of the work. First, license requirements: the Export Administration Regulations require US government approval before shipping advanced computing chips, semiconductor manufacturing equipment, and supercomputer components to China, and licenses are presumed denied. Second, the Entity List: named Chinese companies — SMIC since 2020, YMTC since 2022, and dozens of equipment makers and AI firms added in 2023 and 2024 — are cut off from US-origin technology absent a license. Third, the foreign direct product rule: the most powerful lever, which extends US jurisdiction to foreign-manufactured chips if their production used American software, equipment, or design tools, which in practice means nearly every leading-edge chip anywhere. A related rule restricts US persons — American engineers and executives — from supporting advanced Chinese fabs without authorization.
What do the technical thresholds cover?
The 2022 rules drew lines by performance: chips exceeding defined computing thresholds, later refined in the October 2023 update with a "performance density" metric that closed the first workaround — chips like the Nvidia A800 and H800, which Nvidia had created in 2022 to sell into China just under the original limits. The December 2024 package tightened equipment controls and captured high-bandwidth memory, the component that makes AI accelerators fast. The Netherlands and Japan aligned from 2023 onward, restricting the lithography and deposition tools their firms — ASML, Tokyo Electron, Nikon — sell to China, which matters because the United States cannot restrict that equipment alone. A January 2025 rule attempted a global licensing framework for AI chips by country tier; the incoming administration rescinded it in May 2025, leaving country-specific controls as the operative policy.
Where have the controls leaked?
The documented record shows three patterns. Loophole-and-patch: Nvidia built China-specific chips around each rule, and each round of rulemaking plugged them — a cycle visible in 2022, 2023, and again with 2024-2025 products. Third-country transshipment: congressional investigators and Commerce enforcement actions have traced restricted chips reaching Chinese buyers through intermediaries in Malaysia, Singapore, and the Middle East, which is why entity-list designations expanded to distributors. And indigenous progress: SMIC's 7-nanometer processor, found in Huawei's August 2023 smartphone, demonstrated that controls slow but do not stop Chinese fabrication — at lower yield and higher cost, per industry analysis, but not at zero. The honest assessment from CRS researchers is that controls raise China's costs and delay timelines; nothing in the record shows they have halted capability development.
| Rule action | Date | What it did |
|---|---|---|
| Foundational advanced-computing rules | Oct 2022 | First sweeping chip and equipment controls on China |
| First update | Oct 2023 | Closed A800/H800 gap; added performance density test |
| Second update | Dec 2024 | HBM controls; about 140 Entity List additions |
| AI diffusion framework | Jan 2025 | Global tiered licensing — rescinded May 2025 |
Related stories: Space Force and the Orbital Traffic Jam · Authorized vs Appropriated: Two Budgets.
How does enforcement work on the ground?
The rules are only as strong as their enforcement, which runs through the Commerce Department's Office of Export Enforcement — agents who investigate unlicensed shipments, and administrative or criminal cases pursued with the Justice Department. The statute behind the controls authorizes substantial penalties, including criminal sentences of up to 20 years for willful violations and civil fines that are adjusted upward annually, which is why even mid-sized freight forwarders have appeared in enforcement actions. The documented enforcement pattern since 2022 has three strands: prosecutions of distributors who diverted restricted chips to Chinese buyers through intermediaries; voluntary self-disclosures by US companies that received suspicious orders; and entity-list designations aimed at known transshipment hubs, added in the 2023 and 2024 rule packages. Congressional investigators have publicly traced restricted Nvidia chips reaching Chinese entities despite the bans — findings the company said it prohibits and investigates — which frames the structural problem BIS itself acknowledges: a small enforcement agency policing a global flow of commodities, where the marginal unit of enforcement attention matters more than any single rule's text.
How do allied controls fit in?
The regime is trilateral at its core because the tools themselves are concentrated. The most advanced lithography machines come from one Dutch company, ASML, whose extreme-ultraviolet systems have never been licensed to China; deep-ultraviolet lithography, etch, and deposition tools come from the Netherlands, Japan, and the United States. Beginning in 2023, the Dutch and Japanese governments imposed their own license requirements on advanced semiconductor equipment exports to China, and the December 2024 US package was coordinated to minimize substitution — the scenario where a Chinese buyer simply shifts orders from an American firm to a foreign competitor. The coordination has documented seams: allied measures are unilateral national rules rather than a treaty regime, coverage differs on older-generation equipment, and Chinese retaliation has so far concentrated on critical minerals and customs pressure rather than chips themselves. Congressional researchers characterize the arrangement as an unprecedented alignment of three equipment-supplying governments, held together by regular negotiation — durable enough to matter, and fragile enough that every annual rule cycle doubles as an alliance-management exercise.
What does this change?
The controls have redrawn the semiconductor industry's map: supply chains re-routed through allied countries, Chinese firms investing in domestic toolmakers, and US firms absorbing billions in China revenue loss disclosed in their filings. What the next marker looks like is procedural: each annual rule update is a negotiation among national-security agencies, the industry's compliance lawyers, and allied governments, and each gap discovered — by auditors, journalists, or congressional investigators — forces a patch. The debate the record frames is not whether to control but at what cost curve: defense officials argue the restrictions buy years at the frontier; industry disclosures and some economists counter that the revenue lost to restrictions funds the very research capacity the controls aim to preserve.