Nvidia has purportedly told its data center customers that they must limit the amount of business that they do with Chinese companies in what is a new policy for the U.S. company. The move would mean that data center operators who sell their services to Chinese companies must now track what proportion of their overall business they make up, limiting them to just 49% of the total and ensuring 51% are non-Chinese outfits. The FT reports the new policy has been communicated to Nvidia's customers by its sales force and regional teams, and does not appear to have been laid out in any official form via documentation.
- While the U.S. flip-flops on chip sanctions, China is building its own chip supply market
- The state of China's decade-long semiconductor push
- Huawei Ascend NPU roadmap examined
- China's CXMT targets 30% DRAM memory market share by 2030 with sixth mega-fab
- China's latest round of rare-earth export controls explained
Notably, Nvidia's alleged new stance on Chinese customers does not only apply to data centers that are located in Asia. Instead, it applies to those based globally, and even then it only affects new and future contracts rather than any existing Nvidia hardware leasing agreements that are in place. In a statement, Nvidia notes that "American chips are the foundation that American model and application makers need to compete worldwide." The statement goes on to say that current U.S. policy "encourages" cloud providers to build on American infrastructure and to accept business from "everywhere they can."
But this latest move to restrict Chinese access to Nvidia's high-performance data center hardware comes amid ongoing tensions between the United States and China with regard to AI capabilities. Nvidia can only sell its chips to whitelisted companies, a situation that has led to a thriving black market for its wares. Earlier this year, it was found that even Nvidia's older servers have become so sought-after that their values have tripled.
Whether limitations on who Nvidia can sell its chips to have had the desired effect is open to debate, with even China's state-owned telecoms giant China Telecom having recently been found to have Nvidia hardware in its data centers. The situation wouldn't be out of place in a blockbuster espionage thriller, with Nvidia CEO Jason Huang reportedly the target of a spy plot by Chinese agents.
The reasons behind the U.S. government's attempts to limit China's access to Nvidia hardware are obvious. The aim is to stymie China's ability to advance its own AI technologies, with Nvidia chips the go-to option for AI servers and data centers around the globe. But this limitation on which customers cloud service providers can accept will put additional strain on companies that are already fighting it out for the same contracts. Reducing the pool of potential customers could potentially see a shrinking in the scale of data center operations moving forward.
