Nvidia has told cloud-computing partners they must sell at least 51% of their services to non-Chinese companies, the Financial Times reported Tuesday, citing five people with knowledge of the matter — a new restriction with no prior equivalent at the chipmaker. The instructions were delivered verbally through Nvidia's sales force and regional teams rather than in a written policy document, and no specific cloud partners or customer firms were named in the reporting.
The rule appears aimed at "neocloud" providers, often based in places such as Singapore, Malaysia and Japan, that Chinese firms have used to rent overseas computing capacity and indirectly gain access to high-end American chips for AI model training — a workaround for export controls that has drawn growing scrutiny from Washington. Nvidia has not publicly confirmed or denied the 51% figure, and no company spokesperson was quoted on record in the coverage reviewed, which was also independently relayed by Hong Kong's The Standard. Reuters said it had republished elements of the FT's reporting but could not immediately verify it independently.
Washington has restricted Nvidia's sales of advanced chips to China since 2022, and the company's most recent quarter booked close to zero revenue from its China-specific Hopper-series H20 chip, down from $4.6 billion a year earlier. A limited export license for its H200 chip, approved in February and covering buyers including Alibaba, Tencent, ByteDance and JD.com, has so far generated little booked revenue. The new cloud-partner directive would effectively close an indirect channel that remained open even as direct chip sales to China dried up.
The move lands against a backdrop of aggressive enforcement: Super Micro, which assembles Nvidia-based servers, had executives charged this year in a case involving roughly $2.5 billion of Nvidia AI servers allegedly smuggled into China, illustrating how high the stakes have become for companies caught facilitating workarounds. Shares of Nvidia and Super Micro both dipped modestly on the report Tuesday, a muted reaction that reflects how routine new layers of China-related restriction have become for the chip sector over the past three years. Nvidia has not said when or whether the directive will be formalized in writing, or how it plans to enforce compliance among the dozens of cloud operators it counts as partners.
