NVIDIA CEO Jensen Huang said at the Milken Institute Global Conference that China should not receive the company’s most advanced AI chips, naming both Blackwell (B200) and the next-generation Rubin (R100). His position was clear: the United States should have “the first, the most, and the best” AI hardware.
The remark came during a May 4 appearance in Los Angeles. Asked whether China should be allowed to access NVIDIA’s top-tier chips, Huang answered with a direct “No.” He said that neither the current Blackwell generation nor the upcoming Rubin line should be made available to China.
NVIDIA says its China AI accelerator share is now zero
Huang also acknowledged that NVIDIA’s market share in China’s AI accelerator segment has dropped to zero. In practical terms, that means no H200 GPUs were shipped into China in the first quarter of 2026. The figure stands out on its own, but the reasons are tied to restrictions on both sides.
On the US side, the Trump administration approved H200 exports to China last December, but attached strict terms: 25% of revenue must be paid to the US government, export volume cannot exceed 50% of US domestic sales, and shipments must pass security testing by a third-party US laboratory. On the Chinese side, import approval for the H200 has still not been granted. NVIDIA said in March that it had restarted production planning for the H200, yet no shipping date has been confirmed.
Blackwell and Rubin are outside any China negotiation
The report notes that H200 is the top-end product of NVIDIA’s previous architecture, while Blackwell and Rubin are ahead of what China can currently access by one and two full generations. Huang’s comments amount to a public signal that these two product lines are not on the table in any talks involving China.
China’s state-backed outlet Global Times responded quickly, criticizing NVIDIA for trying to profit from China while guarding against it, and said that strategy had already been proven unsuccessful.
Domestic alternatives in China are gaining ground
With supply constraints still in place, China’s AI sector is moving faster to reduce reliance on NVIDIA GPUs. According to the source material, DeepSeek V4 has announced a full shift away from NVIDIA GPUs in favor of domestic chips. At the same time, Chinese AI chip makers including Cambricon, Moore Threads, and Muxi posted revenue growth of 453%, 243%, and 121% in 2025, and each achieved full-year profitability for the first time.
Taken together, Huang’s comments and those market changes show that NVIDIA is facing more than export controls in China’s high-end AI chip market. It is also dealing with a faster local substitution cycle.