US Chip Export Controls

What it is

The US semiconductor export control regime is administered by the Bureau of Industry and Security (BIS), an agency within the US Department of Commerce. BIS restricts exports of chips and chipmaking equipment through two main instruments: the Commerce Control List (CCL), which assigns Export Control Classification Numbers (ECCNs) to controlled items, and the Entity List, a blacklist of companies and institutions barred from receiving US-origin technology without a BIS licence. A third mechanism, the Foreign Direct Product Rule (FDPR), extends US jurisdiction beyond US-made goods: any product built anywhere using US-origin manufacturing equipment or design software can be pulled into the control framework. That is what gives the rules reach over Dutch manufacturer ASML's extreme-ultraviolet lithography machines and wafers processed at TSMC fabs in Taiwan.

History

The first major escalation came on October 7, 2022, when the Biden administration imposed sweeping restrictions on advanced computing GPUs above a defined threshold of interconnect bandwidth density, along with semiconductor manufacturing equipment critical to sub-14nm production. A second wave in October 2023 expanded controls to additional equipment categories, high-bandwidth memory, and added entities including YMTC to the Entity List. A December 2, 2024, interim final rule added 140 more entities, extended controls to advanced packaging equipment, and tightened high-bandwidth memory restrictions. On January 15, 2025, BIS issued foundry due-diligence requirements compelling downstream customers to verify transistor counts and conduct formal risk assessments. The Trump administration, taking office in January 2025, rescinded the three-tier Diffusion Rule Biden issued on January 13, 2025, replacing it with a bilateral deal framework while maintaining core controls on China.

Current state

As of mid-2026, China is subject to licence requirements for advanced chips above roughly 100 tera-operations per second of compute density. A US-China agreement reached in May and June 2026 opened a pathway for Chinese cloud firms to purchase up to 75,000 Nvidia H200 units each under a bilateral licence. Beijing simultaneously instructed those firms to pause purchases and favour domestic alternatives from Huawei, Cambricon, and Alibaba, leaving deliveries at near zero, as documented in the Washington cleared the H200 for China; Beijing won't let anyone buy it. Nvidia's share of the Chinese AI-accelerator market reportedly fell from roughly 95 percent to 55 percent over 2025. Allied alignment remains partial: the Netherlands has restricted ASML EUV exports under US pressure, and Japan imposed its own semiconductor equipment restrictions in 2023, but enforcement gaps persist in third-country hubs such as Singapore and Malaysia. China's domestic chip industry, led by SMIC and Huawei's HiSilicon design arm, is now producing at an estimated 7nm to 5nm equivalent process node.

Relationships

The controls form one pillar of a wider US semiconductor strategy. The August 2022 CHIPS and Science Act authorised US$52.7 billion in domestic manufacturing subsidies and research funding, explicitly prohibiting recipients from expanding fabrication capacity in China for ten years. Multilaterally, the Wassenaar Arrangement on Export Controls for Conventional Arms and Dual-Use Goods and Technologies, with 42 member states, provides a broader framework, but China is not a member and Wassenaar requires consensus to tighten rules. China has responded with its own export controls on critical minerals, rare-earth processing compounds, and, in June 2026, by extending its dual-use control regime to Japanese defence entities, documented in China blacklists 40 Japanese defence entities in second wave of dual-use export crackdown, partly mirroring the US regulatory playbook. The chip controls are also inseparable from the broader US-China trade fracture: semiconductors account for roughly US$75 billion in annual US exports to Asia and remain the bilateral relationship's most contested technology.

What to watch

• Whether Beijing formalises a prohibition on US AI chip purchases or allows the H200 licensing pathway to reopen under a renegotiated bilateral framework before existing Chinese cloud inventories run down.
• Compliance at downstream fabs, particularly TSMC in Taiwan: any discovered diversion of advanced-node wafer capacity to Chinese customers would test the FDPR's extraterritorial reach and force US enforcement against a strategic ally's flagship company.
• Allied alignment: whether South Korea (covering Samsung and SK Hynix high-bandwidth memory exports) and the Netherlands (ASML EUV shipments) converge fully with US controls, or negotiate national carve-outs under commercial pressure.
• China's indigenous progress: SMIC's 5nm yield rate and Huawei's next GPU generation are the benchmarks that determine how long the controls hold a meaningful technology gap.