NVIDIA CEO Jensen Huang clearly stated at the Milken Institute Global Conference that China should not gain access to Blackwell or next-generation Rubin architecture AI chips, emphasizing that the United States should always have the “first, most, and best” AI hardware. At the same time, he admitted that NVIDIA’s market share of AI accelerators in China has fallen to zero.
(Background: NVIDIA and AMD agreed to Trump’s “15% AI chip tax,” allowing H20 and MI308 imports into China, with gross margin flowing directly into Washington)
(Additional context: DeepSeek V4 announces it will abandon NVIDIA! Where does China’s “computing power independence” battle in AI stand now?)

It appears Huang briefly wavered on the “China issue.” On May 4, at the Los Angeles Milken Institute Global Conference, when asked whether China should be allowed to obtain NVIDIA’s most advanced chips, he directly answered, “No.”

He said that neither the current-generation Blackwell (B200) nor the upcoming next-generation Rubin (R100) should be accessible to China, and emphasized that as NVIDIA’s home country, the U.S. should always have the top priority—“the first, the most, and the best.”

China’s market share has fallen to zero; no H200 shipments this year

During the same conference, Huang also acknowledged that NVIDIA’s market share of AI accelerators in China is currently zero, meaning that in the first quarter of 2026, no H200 GPU will actually be shipped to China.

The reason behind this is a double bind. Although the Trump administration approved H200 exports to China last December, it came with strict conditions: 25% of revenue must be remitted to the U.S. government, export quantities must not exceed 50% of U.S. domestic sales, and the products must undergo security testing by U.S. third-party labs.

Meanwhile, China’s government has not yet approved the import license for the H200, creating a stalemate of “the U.S. clears it, but China doesn’t approve.” In March, NVIDIA said it had restarted the H200 production line to rebuild inventory, but the actual shipping dates remain undecided.

Above the red line, Blackwell and Rubin are completely banned from exports to China

H200 is NVIDIA’s highest-end product from the previous architecture, while Blackwell and Rubin each lead by one and two complete hardware generations over what China can access. Huang’s remarks are effectively a formal declaration that these two product lines are outside the scope of any negotiations with China.

Chinese state media, the Global Times, responded immediately, criticizing NVIDIA for “wanting to make money from China while guarding against China,” and noting that this strategy “has already been proven to be a failure.”

In fact, China’s AI industry is accelerating its “de-NVIDIA-ization,” including DeepSeek V4’s announcement that it will fully abandon NVIDIA GPUs and switch to domestic chips. In 2025, domestic AI chipmakers such as Cambricon, Moore Thread, and Muxin saw their revenues surge by 453%, 243%, and 121% respectively, achieving full-year profitability for the first time.

This article does not discuss smuggling chips.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.