China is drawing a firm line on foreign AI hardware, and NVIDIA is feeling the impact. According to new reports, the country’s cyberspace regulator has directed major domestic tech firms to stop testing the RTX 6000D and to halt new orders. That effectively blocks NVIDIA from selling what could have been millions of units to Chinese customers and signals Beijing’s strongest move yet to reduce reliance on Western AI chips.
This escalation follows earlier restrictions that curbed shipments of NVIDIA’s H20 accelerators amid concerns about potential security backdoors. The latest directive goes further, suggesting a broader pivot: China appears confident its domestic AI chips can meet current demand under export controls. Companies such as Huawei and Cambricon have advanced rapidly, giving local cloud providers and AI developers more homegrown options for training and inference.
NVIDIA’s leadership has acknowledged the shifting landscape. CEO Jensen Huang expressed disappointment at the developments but emphasized patience, saying the company can only serve markets that want its products and recognizing that larger geopolitical issues are at play between China and the United States.
Even with meaningful progress on performance, China still faces significant production bottlenecks. The challenge isn’t just about building competitive semiconductors; it also hinges on securing high-bandwidth memory and scaling manufacturing to data center volumes. Matching or surpassing the H20’s capabilities on paper is only one piece of the puzzle. The real test is whether domestic supply chains can deliver at scale, consistently and cost-effectively.
Key takeaways for the AI market:
– China’s regulators have asked leading tech companies to stop testing and ordering NVIDIA’s RTX 6000D, effectively shutting off a major sales channel.
– Earlier measures targeting the H20 have evolved into a broader effort to prioritize domestic AI accelerators.
– Local chipmakers are closing the gap on performance, but capacity and HBM supply remain critical hurdles.
– NVIDIA’s response signals a long game: acceptance of near-term setbacks while watching how policy and supply chains evolve.
What happens next will influence everything from the pace of AI model training inside China to global chip supply dynamics. If domestic providers can overcome production constraints, China’s AI stack will become more self-reliant. If not, demand and timelines for large-scale AI projects could remain constrained. Either way, the message is clear: the competition for AI computing leadership is no longer just about raw speed—it’s about resilience, sovereignty, and the ability to deliver at scale.
