Nvidia tightens Asia distributor checks to curb China chip diversion

Nvidia has reduced its approved distributor list in Asia by more than 50% and introduced stricter vetting in Japan, Singapore, and Malaysia to prevent its AI chips from reaching China through intermediaries. The restrictions have mostly hit neoclouds, custom cloud platforms built for AI computing, though rejected firms can reapply after changing their compliance practices.

The new checks include data center audits, contract verification, user interviews, and more physical inspections, with support and oversight from the US Department of Commerce. Washington is targeting broker networks after years of restrictions failed to fully stop advanced chips from moving into China. In March, a Supermicro co-founder was indicted over a $2.5 billion scheme allegedly using a Southeast Asian intermediary to ship Nvidia-powered servers into China.

US Commerce Department guidance in May also requires export licenses for top-tier AI processors when the parent company is based in China, regardless of shipment location, closing a loophole involving Singapore and Malaysia branches. Beijing separately continues to bar Nvidia’s H200 chips to support domestic suppliers, though insiders said Alibaba Group Holding, ByteDance, and DeepSeek may gain limited access. Any approval would restrict H200 use to AI training on public data, with only about 200,000 chips allowed in, less than 50% of what companies sought earlier this year.