ByteDance and Tencent received initial NVIDIA H200 shipments, but Beijing is restricting mainland deployment to protect domestic chipmakers.
ByteDance and Tencent just took delivery of roughly 10,000 NVIDIA H200 AI chips each. The small shipments mark the first real movement of these powerful processors into China since Washington granted export licenses.
The United States cleared Chinese tech firms to buy up to 100,000 H200 chips apiece. Yet a surprising plot twist comes from Beijing, not Washington. Chinese regulators are actively capping mainland deliveries. They want domestic tech giants to park these processors in Hong Kong instead.
Beijing’s strategy shows clear logic.
Chinese regulators want to protect domestic chipmakers from NVIDIA’s market dominance, but the AI developers there still need its raw compute power to train their cutting-edge models. Sending hardware to Hong Kong creates a clever compromise, even if local data centers face severe power limits.
This dynamic exposes the reality of the global AI race. Government mandates cannot easily erase market demand for top-tier silicon.
Chinese tech giants rely on NVIDIA for heavy AI training, while using domestic chips for basic inference tasks. Beijing balances national self-reliance with immediate technical needs by allowing limited shipments.
