Nvidia Denies DeepSeek Is Using Its Banned Blackwell Chips

• Nvidia (NVDA) firmly denied reports that Chinese AI startup DeepSeek is using smuggled Blackwell chips, stating it has seen no evidence of elaborate “phantom datacenter” smuggling schemes.
• U.S. export controls continue to fully ban Nvidia’s flagship Blackwell GPUs to China, while President Trump proposed allowing H200 shipments to approved Chinese customers in exchange for a 25% U.S. revenue share, a plan that faced immediate Republican pushback.
• DeepSeek previously rattled the industry with its low-cost, high-performing R1 model released in January and has signaled that China expects domestic next-generation AI chips soon, intensifying focus on Nvidia’s restricted hardware reaching Chinese developers.

Nvidia (NVDA) publicly dismissed claims that Chinese AI startup DeepSeek has gained access to its restricted Blackwell-series GPUs through smuggling operations. The denial follows a report from The Information alleging that DeepSeek has been training its next-generation models on Blackwell chips that entered China without U.S. authorization, despite strict export controls imposed to prevent the transfer of America’s most advanced AI accelerators to Chinese entities.

A Nvidia spokesperson stated that the company has received no credible evidence of elaborate smuggling schemes involving “phantom datacenters” built for inspection purposes and later disassembled, transported, and rebuilt elsewhere. While acknowledging that such scenarios are theoretically possible, Nvidia emphasized that it actively investigates every credible tip it receives regarding potential violations of export restrictions.

The Blackwell architecture represents Nvidia’s current flagship offering for large-scale AI training and inference, delivering substantially higher performance and efficiency than the H100 and H200 series that preceded it. The U.S. Commerce Department added the entire Blackwell family to its export-control list, effectively barring sales to China unless individually licensed – an outcome that has become exceedingly rare for cutting-edge AI hardware.

The issue has taken on heightened political significance. On Monday, President Donald Trump announced that Nvidia would be permitted to ship H200 accelerators to approved customers in China provided the U.S. government receives 25% of the revenue from those transactions. The proposal immediately drew criticism from several Republican lawmakers who argue that any relaxation of controls risks accelerating China’s AI capabilities.

DeepSeek itself contributed to the current tension earlier this year when it released DeepSeek-R1 in January, an open-weight reasoning model that quickly topped both Apple App Store rankings and independent performance leaderboards while reportedly being developed at a small fraction of the compute and capital expenditure required by leading U.S. frontier models. In August, the company further signaled that Chinese entities expect to field domestically produced next-generation AI chips in the near future, reducing longer-term dependence on restricted foreign hardware.

Nvidia remains by far the dominant supplier of the high-performance GPUs required to train and deploy the world’s largest AI systems, a position that has made its compliance with U.S. export controls a recurring focal point in the broader technology competition between the United States and China. The combination of persistent demand from Chinese AI developers, tightening U.S. restrictions, and continued reports of indirect acquisition channels has kept the issue at the center of both market and geopolitical attention.

WallStreetPit does not provide investment advice. All rights reserved.

• Bulenox: Get 45% to 91% OFF ... Use Discount Code: UNO
• Risk Our Money Not Yours | Get 50% to 90% OFF ... Use Discount Code: MMBVBKSM