MediumV Technology Geopolitics & AI3 November 2026, Tuesday
Nvidia wants cloud partners' new contracts to be majority non-Chinese
According to an FT report based on 5 sources and carried by AFP on 3 November, Nvidia is requiring cloud partners in new contracts to sell at least 51% of their services to non-Chinese companies. The company neither confirmed nor denied the report on 3 November.
According to the report, the 51% rule applies only to new contracts and leaves existing agreements untouched. In a one-sentence statement to AFP, Nvidia said data centres built with American chips are the foundation for American AI developers. In the same 3 November statement, the company added that America cannot win the AI race unless its infrastructure wins around the world.
Chinese Foreign Ministry spokesperson Mao Ning said on 3 November that she was not aware of the report. Beijing's one-sentence response was limited to a call for a fair, transparent and non-discriminatory business environment for everyone, Chinese companies included. The timing falls 2 weeks before China hosts APEC leaders in Shenzhen on 18–19 November.
This time the restriction comes not from a state but from a supplier. The US export regime limits chip sales to China through licensing; the 51% requirement uses a supplier's contract to narrow the route by which Chinese developers rent compute remotely through foreign clouds. It has not been disclosed whether the threshold will be measured by revenue or customer count, or whether there is an audit mechanism; this point could not be verified.
Talay assessment
Bottom line
China's access to AI compute is now being narrowed not only by Washington's licences but by a supplier's contracts. Because the 51% requirement covers only new contracts, its effect will build over months. Beijing's muted response suggests it is postponing escalation until after the Shenzhen summit; a reply through another channel, such as rare earths, is highly likely.
Likely effects
- Asian cloud operatorsNegative1–6 months
Independent operators with a high share of Chinese customers will have to change their customer mix in new contracts or turn to non-Nvidia hardware.
- Chinese AI developersNegative1–6 months
The cost of renting compute remotely and uncertainty over access rise, accelerating the shift to domestic accelerators.
- US–China bargainingUncertainWeeks
Compute access enters the agenda at the 18–19 November Shenzhen summit; Beijing may respond through the licensing channel.
Possibilities, ranked
- 1Rule applied quietly55%
Nvidia applies the requirement in new contracts without publishing a written text, and operators gradually change their customer mix.
Watch: Customer disclosures from GPU cloud operators in Asia in November–December
- 2Softened at the summit25%
After a general understanding at the Shenzhen summit, the requirement is loosely enforced.
Watch: US–China statements at the 18–19 November APEC leaders' summit
- 3Beijing retaliates20%
China responds with an investigation or licensing curbs targeting Nvidia.
Watch: Statements on Nvidia from China's Commerce Ministry and competition regulator
Probabilities are calibrated judgement based on the sources, not measurement, and are revised as new information arrives. Not investment advice.
Market reaction
Indicators affected
- Non-China sales requirement▼ ≥51%
- Number of sources (FT)▲ 5