Nvidia cloud customers have been advised to reduce business ties with firms in China, according to a report published by the Financial Times on Tuesday. The guidance was directed at data centre operators that provide computing capacity, citing unnamed sources familiar with the discussions.
Key points
- The Financial Times reported that Nvidia advised data centre operators to lower their exposure to firms in China.
- The advice affects companies leasing server capacity powered by the chipmaker’s hardware.
- US export controls continue to limit direct sales of advanced processors to Chinese buyers.
Advice to data centre providers
Nvidia communicated the guidance to companies running server infrastructure equipped with its graphics processors. These cloud partners buy chips in bulk and rent compute power to external clients, including machine learning startups and enterprise software developers.
According to the Financial Times, the chipmaker asked operators to limit deals with Chinese commercial entities. The report cited people familiar with the request, but did not name individual cloud providers or outline specific compliance deadlines.
Regulatory pressure on remote computing
United States trade rules already prevent direct shipments of advanced artificial intelligence processors to buyers in China. However, offshore compute access remains an active area of interest for regulators monitoring the technology sector.
Foreign server operators can technically allow overseas customers to access clusters over the internet. Nvidia’s reported advice suggests a push to avoid potential regulatory fallout related to training models or processing workloads for Chinese businesses from external server farms.
Financial implications for cloud operators
Any reduction in international client demand could affect data centre companies that borrowed heavily to build hardware capacity. That pressure is already visible as neocloud debt worries mount as AI data centre firms face market pressure across the infrastructure sector.
Specialised hosts rely on high hardware utilisation rates to service financing costs. At the same time, major operators are broadening their chip partnerships, reflecting moves where Oracle expands cloud compute with Google TPUs and rival chips to avoid reliance on a single architecture.
Nvidia has not commented publicly on the Financial Times report. The company has not stated whether the request will lead to formal contractual terms for its infrastructure partners.
