Nvidia Caps China Share of Cloud Services

Nvidia has asked data center and cloud customers worldwide to limit the share of services they provide to Chinese companies to 49% or less, the Financial Times reported on the 3rd, citing multiple sources. In recent months, Nvidia has verbally told customers to sell at least 51% of their total services to companies outside China. The restriction did not previously exist, and Nvidia’s sales and regional teams reportedly delivered the guidance verbally rather than in writing.

The guidance applies to customers worldwide, including in Asia. Existing lease agreements will remain in place, while the rule applies to new contracts. Nvidia has not disclosed how the 51% share will be calculated or how compliance will be checked. The report said the move follows efforts by the U.S. government to restrict Chinese companies’ access to cloud services abroad as well.

The United States restricts exports of advanced AI chips to China, but Chinese companies have rented Nvidia GPUs at data centers overseas, including in Southeast Asia, to train AI models. In July, Nvidia introduced a whitelist of companies that passed stricter compliance reviews, cutting by more than half the number of Asian customers eligible to buy AI chips. The report said shortages of advanced GPUs in China are also worsening. The FT reported that an Nvidia B300 chip package, which trades for about $600,000 on international markets, sold at auction in China last month for 18.5 million yuan, or about $2.8 million.

Overseas cloud providers may also be affected. Southeast Asian data centers with Chinese companies as major customers may have to reduce their operations if they cannot attract more customers outside China. An industry figure in Singapore said securing enough non-Chinese customers outside the United States is difficult and that strict enforcement of the guidance would significantly affect the business. Nvidia did not explain the criteria in detail, saying that global cloud infrastructure based on U.S. chips is important for strengthening the competitiveness and economic influence of U.S. AI companies.

Nvidia has asked data center and cloud customers worldwide to limit the share of services they provide to Chinese companies to 49% or less, the Financial Times reported on the 3rd, citing multiple sources. In recent months, Nvidia has verbally told customers to sell at least 51% of their total services to companies outside China. The restriction did not previously exist, and Nvidia’s sales and regional teams reportedly delivered the guidance verbally rather than in writing.

The guidance applies to customers worldwide, including in Asia. Existing lease agreements will remain in place, while the rule applies to new contracts. Nvidia has not disclosed how the 51% share will be calculated or how compliance will be checked. The report said the move follows efforts by the U.S. government to restrict Chinese companies’ access to cloud services abroad as well.

The United States restricts exports of advanced AI chips to China, but Chinese companies have rented Nvidia GPUs at data centers overseas, including in Southeast Asia, to train AI models. In July, Nvidia introduced a whitelist of companies that passed stricter compliance reviews, cutting by more than half the number of Asian customers eligible to buy AI chips. The report said shortages of advanced GPUs in China are also worsening. The FT reported that an Nvidia B300 chip package, which trades for about $600,000 on international markets, sold at auction in China last month for 18.5 million yuan, or about $2.8 million.

Overseas cloud providers may also be affected. Southeast Asian data centers with Chinese companies as major customers may have to reduce their operations if they cannot attract more customers outside China. An industry figure in Singapore said securing enough non-Chinese customers outside the United States is difficult and that strict enforcement of the guidance would significantly affect the business. Nvidia did not explain the criteria in detail, saying that global cloud infrastructure based on U.S. chips is important for strengthening the competitiveness and economic influence of U.S. AI companies.