Jensen Huang, chief executive officer of Nvidia Corp., outside the U.S. Capitol in Washington, D.C., on Wednesday, Dec. 3, 2025.
Bloomberg | Getty Images
Nvidia on Wednesday refuted a report that the Chinese artificial‑intelligence startup DeepSeek has been using smuggled Blackwell chips to develop its next‑generation model.
The United States has barred the export of Nvidia’s Blackwell chips—its most advanced AI silicon—to China as part of a broader effort to maintain a strategic lead in the global AI race.
According to The Information, DeepSeek is allegedly employing chips that entered China without authorization.
“We have not seen any substantiation or received credible tips of ‘phantom data centers’ built to deceive us and our OEM partners, then dismantled, smuggled, and reassembled elsewhere,” a Nvidia spokesperson said. “While such smuggling seems far‑fetched, we investigate every lead.”
Nvidia has been the primary beneficiary of the AI boom, supplying the graphics processing units (GPUs) that power model training and large‑scale inference workloads. Because the hardware is so critical to AI advancement, Nvidia’s commercial ties with China have become a political flashpoint in Washington.
President Trump recently announced that Nvidia could ship its H200 chips to “approved customers” in China and other markets, provided that the United States receives 25 % of the sales revenue. The proposal sparked mixed reactions among lawmakers, with some Republicans expressing concern over technology transfer risks.
DeepSeek captured industry attention in January with the launch of its R1 reasoning model, which quickly rose to the top of app‑store rankings and benchmark leaderboards. Analysts estimate that R1 was developed at a fraction of the cost of comparable U.S. models, underscoring the competitive pressure on Western AI firms.
In August, DeepSeek hinted that China is close to deploying its own “next‑generation” AI chips, suggesting a future where domestic silicon could further reduce reliance on imported technology.
From a market perspective, the lingering uncertainty around chip export controls creates a paradox for investors. On one hand, tighter restrictions protect American IP and preserve a technological moat; on the other, they risk incentivizing Chinese firms to accelerate indigenous chip development, potentially eroding Nvidia’s long‑term addressable market. The company’s stock has already reflected this tension, with volatility tied to policy announcements and geopolitical headlines.
Technologically, the Blackwell architecture introduces architectural improvements—such as higher tensor‑core density and advanced interconnects—that enable more efficient training of massive transformer models. If China succeeds in obtaining or reverse‑engineering these capabilities, it could narrow the performance gap and spur a new wave of competition in AI research, cloud services, and edge computing.
Strategically, Nvidia is balancing short‑term revenue opportunities from approved sales against the broader goal of shaping a global AI ecosystem that remains aligned with U.S. security interests. The outcome will hinge on how effectively regulators can enforce export limits while allowing enough commercial flexibility to keep Nvidia’s innovation pipeline funded.
– CNBC contributed to this report.
