Chinese regulators have authorised limited shipments of Nvidia H200 chips to mainland firms, allowing ByteDance and Tencent to receive roughly 10,000 units each while other companies await approval. This move permits specific batches to enter the market despite US export controls that ban access to Nvidia’s most powerful processors. The H200 model is at least two generations older than the American equivalent, yet it serves as a critical stopgap for domestic artificial intelligence development. Beijing must obtain sign-off from the National Development and Reform Commission before any purchase occurs, and the total volume remains far below the 100,000 unit ceiling allowed for individual companies in the United States. Logistics remain complicated because Hong Kong lacks the necessary data centre infrastructure and power capacity to absorb these components. Meanwhile, local laboratories including Moonshot, Alibaba, DeepSeek and Z.ai continue to release new models to compete with US providers. Despite these technical advancements, Chinese firms still face significant shortages of inference capacity compared to their American counterparts. Nvidia currently holds approximately half a million H200 units in stock, creating a bottleneck that restricts the speed of domestic innovation.
- Shipments require explicit approval from the National Development and Reform Commission.
- Hong Kong lacks the power grid and data centres needed for deployment.
- Local labs still struggle with insufficient inference capacity despite new model releases.
