29 August 2026

Are US export controls on tech failing?

IISS  |  Chris Clague

Huawei unveiled its LogicFolding chip design approach in May 2026, attempting to bypass United States export restrictions on ASML extreme ultraviolet lithography equipment. Shortly after the Bureau of Industry and Security restricted Anthropic’s Fable 5 and Mythos 5 artificial intelligence models in June, Chinese start-up Z.ai released a cheaper competitor with comparable capability.

Industrial adaptation has systematically undermined American sanctions designed around a small yard and high fence policy. Beijing leveraged threats against rare earth exports to secure a temporary suspension of the affiliates rule until November 2026 following a summit between Donald Trump and Xi Jinping. Enforcement is hampered by bureaucratic underfunding. Meanwhile, smuggling and intellectual property theft further erode trade barriers, forcing Washington to pressure Southeast Asian trade partners into mirroring domestic enforcement capabilities. Ultimately, US policy has spurred Chinese state-directed innovation around chokepoint technologies while reducing overall strategic visibility into Beijing's advancing technological stack.

Comment

Unilateral trade restrictions on dual-use technology often accelerate target-state industrial substitution rather than preserving asymmetric denial. When the Foreign Direct Product Rule targeted Huawei's access to ASML extreme ultraviolet lithography systems, Chinese semiconductor research pivoted from acquiring foreign capital equipment to architectural redesigns like LogicFolding. This transition shifts industrial vulnerability from external supply chain chokepoints to domestic fab capacity and yield efficiency.

Specifically, bypassing lithographic barriers requires replacing single-exposure EUV capabilities with multi-patterning techniques on existing deep ultraviolet systems manufactured by Shanghai Micro Electronics Equipment. This structural workaround increases mask steps and lowers wafer yield per hour, raising unit costs for advanced silicon. Consequently, SMIC and Huawei absorb these higher operational scrap rates through state capital subsidies designed to maintain domestic wafer throughput.

Strategic Question for Discussion
If SMIC relies on multi-patterning with Shanghai Micro Electronics Equipment DUV machines to bypass ASML restrictions, at what point do rising unit scrap costs outweigh state subsidies in sustaining advanced silicon production?
The trajectory indicates that state subsidies can absorb yield inefficiencies for national security applications where unit economics are secondary to raw access. However, as high-volume commercial AI deployment demands millions of accelerators, exponential scrap rates on deep ultraviolet multi-patterning will eventually strain state fiscal allocation relative to global competitors. My assessment is that Beijing will maintain non-commercial funding indefinitely for defense-adjacent computing, while dual-use commercial applications face severe scaling bottlenecks.
Share your assessment in the comments below.

No comments: