US takes step to halt Nvidia AI chip shipments to Chinese overseas subsidiaries - South China Morning Post

The U.S. Department of Commerce took action on June 1, 2026, to prevent the export of advanced AI chips to Chinese overseas subsidiaries, closing a loophole that may have allowed such shipments under ... Read original →

The U.S. Commerce Department’s move to block Chinese firms’ overseas subsidiaries from accessing advanced AI chips marks a shift from product-based bans to entity-level, jurisdiction-piercing controls. Technically, this accelerates China’s push for sub-3nm architectures and EUV alternatives, yet creates an acute compute gap that cripples large-model training in the near term. Compliance burdens surge: NVIDIA and AMD must now vet every customer’s ultimate ownership, triggering licensing reviews that delay shipments and inflate supply-chain insurance costs. Strategically, AMD may temporarily gain share with MI350X among clients abandoned by NVIDIA, but sustained decoupling will funnel Chinese state capital into domestic GPU ecosystems like Ascend and Cambricon. Over the next 18 months, expect China to pivot toward Southeast Asian joint ventures for mature-node chips and double down on chiplet and in-memory computing to bypass lithography restrictions. This isn’t just another export rule—it’s the institutionalization of a semiconductor iron curtain.