Nvidia rejects claims that China’s DeepSeek is using ‘smuggled Blackwell chips’

Nvidia

Nvidia has dismissed allegations that Chinese AI startup DeepSeek is training its next-generation model using smuggled Blackwell GPUs. The denial follows a report from The Information claiming DeepSeek had secretly obtained the restricted hardware despite US export controls.

Speaking to CNBC, an Nvidia spokesperson said the company has not seen any credible evidence supporting the claims. “We haven’t seen any substantiation or received tips of ‘phantom datacenters’ constructed to deceive us and our OEM partners, then deconstructed, smuggled, and reconstructed somewhere else,” the spokesperson said. “While such smuggling seems farfetched, we pursue any tip we receive.”

DeepSeek drew attention earlier this year after announcing its R1 model, which it claimed was trained at a fraction of the cost of comparable models developed in the US. Because the US government bans export of Nvidia’s most advanced Blackwell chips to China, any unauthorised use has attracted scrutiny from regulators and competitors.

Nvidia, meanwhile, is building new systems to address potential smuggling. Reports earlier in the day revealed that the company is testing location verification technology that lets customers determine the country in which a chip is operating. The optional software agent measures chip performance and communication latency to infer physical location and can be used to monitor inventory across an entire GPU fleet.

“We’re in the process of implementing a new software service that empowers data center operators to monitor the health and inventory of their entire AI GPU fleet,” the company said.

The combination of export bans, escalating demand for top-tier AI chips and China’s rapid model development has heightened geopolitical pressure around the supply chain. For now, Nvidia maintains that there is no evidence DeepSeek has acquired Blackwell hardware through illicit channels.