China quietly approves Nvidia H200 imports, ByteDance and Tencent load up
Chinese regulators have reportedly started authorizing limited imports of Nvidia's H200 AI processors into mainland China. The news, reported by Alex on Developpez.com, hands Chinese tech firms the compute they need to close the gap with American rivals.
The H200 is a data center GPU built on Nvidia's Hopper architecture, designed for generative AI and large language model workloads. It sits at least two production cycles behind Nvidia's current flagship. US export controls keep Chinese buyers from the most powerful AI hardware on the market, so the H200 is effectively the ceiling.
Ten thousand units each
The Financial Times, citing a source close to the matter, reports that ByteDance and Tencent have each taken delivery of around 10,000 H200 units in recent weeks. Several other companies could receive authorization soon.
Twenty thousand units across two of China's largest AI players is not a trickle. It is a statement. The gap is real, but the pipeline is open.
From pause to approval
The authorization is a reversal. In early January 2026, Nvidia ramped production to meet what CEO Jensen Huang called "quite high" demand for H200s in China. Beijing responded by slamming the brakes:
- Domestic firms were asked to suspend H200 orders.
- Regulators started defining the conditions for authorization.
- Imports were frozen while China weighed self-sufficiency against its hardware needs.
Now the terms are set, and shipments are moving.
Autarky is a long game. Training runs are not.
The whiplash tells you everything about Beijing's strategy. China wants domestic semiconductor champions, but its AI labs need compute now. The H200 is old by Nvidia standards. It is still far ahead of what most Chinese firms can source at home.
The authorization is limited, conditional, and reversible. But ByteDance and Tencent already have their units. Washington's controls pushed China two generations back. Beijing's answer is to buy the best thing America will sell it.