First Nvidia H200 shipments reach China, ByteDance and Tencent take deliveries as Beijing loosens its import block

Recent reports indicate that ByteDance and Tencent have each received approximately 10,000 NVIDIA H200 accelerators, marking the first significant shipment of these chips into mainland China since the... Read original →

The first shipment of NVIDIA H200 chips to China signals a tactical easing in U.S.-China tensions over AI compute capabilities. However, the actual deployment remains severely constrained by China’s infrastructure and regulatory bottlenecks. Despite export approvals, the lack of sufficient data center capacity—especially in Hong Kong, a key transit hub—limits large-scale adoption. Each H200 server draws up to 700W, and even a 100,000-unit allowance equates to 125 MW of IT load, far exceeding current regional capacity. Huawei’s Ascend chips are advancing, but they still lag behind NVIDIA in training large AI models. This delay underscores the continued reliance on U.S. hardware. The limited delivery volume reflects a cautious approach by both governments. In the medium term, Chinese firms will remain dependent on imported chips unless domestic data center infrastructure scales rapidly. This marks a broader shift in global semiconductor supply chains, with China forced to navigate between technological sovereignty and market access.