China Reportedly Bars ByteDance From Using Nvidia Chips in New Data Centres

Chinese regulators reportedly barred ByteDance, TikTok’s parent company, from deploying Nvidia chips in new data centres, according to reporting cited by Reuters on November 26, 2025. The report attributed the information to two ByteDance employees. 2

The development was especially significant because ByteDance had reportedly purchased more Nvidia chips than any other Chinese company in 2025. It was seeking additional computing capacity for its more than one billion users while facing the risk that Washington could further restrict supplies. 2

What ByteDance was reportedly told to stop doing

The reported prohibition concerned the use of Nvidia chips in new data-centre projects. It was not presented in the available reporting as a requirement to remove every Nvidia chip already operating across ByteDance’s existing infrastructure.

The reported action also fit a wider set of restrictions and procurement guidance aimed at reducing Chinese technology companies’ dependence on foreign AI hardware. Earlier in 2025, regulators had reportedly asked local companies to stop placing new Nvidia AI-chip orders. 2

Separate November guidance required new data-centre projects receiving state funding to use only domestically produced AI chips. Projects that were less than 30% complete were reportedly told either to remove installed foreign chips or abandon plans to acquire them, while more advanced projects were to be assessed individually. 36

Why the decision was unusually consequential for Nvidia

ByteDance was not a marginal customer. Its reported position as the largest Chinese buyer of Nvidia chips in 2025 made the restriction a direct test of whether even the country’s biggest AI-compute buyers could continue expanding with U.S.-designed accelerators. 2

Nvidia also faced a narrowing product set in China. Washington had restricted exports of the company’s most advanced AI accelerators, while lower-performance China-specific products such as the H20 faced changing licensing conditions and limited availability. That made Nvidia’s China supply less predictable even before Beijing’s reported procurement restrictions took effect. 26

Nvidia summarized the commercial problem directly, saying it was “not able to offer a competitive data center GPU in China.” 2 The statement reflected a market shaped by both governments: U.S. rules limited which products Nvidia could sell, while Chinese policy increasingly discouraged or blocked the purchase and deployment of those permitted alternatives.

A double squeeze on AI infrastructure

The result was pressure on both sides of the supply chain:

• For ByteDance and other Chinese buyers: access to familiar Nvidia hardware became less reliable just as demand for AI computing was rising.
• For Nvidia: the products it could offer were constrained by U.S. rules, and Chinese customers faced regulatory pressure to choose domestic alternatives.
• For domestic chipmakers: Beijing’s procurement rules created a stronger opportunity to supply the country’s expanding AI infrastructure.

Weak Chinese demand for some permitted Nvidia products added to the problem. The available reporting describes a market in which geopolitical restrictions, product-performance limits and domestic competition were all working against Nvidia’s ability to maintain its previous position. 2

What it signaled about Beijing’s AI strategy

The ByteDance report was best understood as part of a longer-term self-reliance campaign rather than an isolated purchasing decision. By directing state-backed data-centre projects toward domestic chips and discouraging new Nvidia orders, Beijing appeared willing to accept near-term trade-offs in performance, compatibility or supply in exchange for reducing exposure to future U.S. restrictions. 23

That strategy also made AI computing a central pressure point in U.S.-China technology relations. The two countries could continue trading in some areas while remaining deeply vulnerable to sudden changes in chip access, licensing and procurement rules. The ByteDance restriction therefore mattered beyond one company: it showed how quickly national technology policy could reshape the infrastructure choices of a major AI platform.