Nvidia begins shipping H200 chips to China amid $1B-plus compute deal spree

A U.S. official confirmed that Nvidia has begun shipping its powerful H200 AI chips to China, a notable development given the export-control tensions that have shaped the market. The resumption of higher-end shipments signals a policy calibration that could reshape Chinese AI labs' hardware access — even as those same labs (DeepSeek most prominently) pursue domestic inference chips to reduce dependence on Nvidia and Huawei.

On the demand side, compute deals are getting larger. AI startup Reflection inked a $1 billion compute agreement with Nebius for access to Nvidia's latest accelerators, following a comparable SpaceX arrangement. These multibillion-dollar capacity commitments reflect a 'compute-investment' theme running through the week: securing GPU supply is now a strategic differentiator, and neoclouds like Nebius are becoming major channels for Nvidia silicon.

Nvidia also published guidance on evaluating general-purpose robot policies for real-world deployment, complementing its Jetson Thor edge modules and LeRobot collaboration — the company is simultaneously feeding data-center demand and seeding the robotics edge. Its Nemotron 3 Ultra, run on LangChain's agent orchestration, was pitched as leading benchmarks at lower cost than top closed models.

The China shipments are politically sensitive: critics worry about strengthening a strategic rival's AI capability, while proponents argue restricting sales only accelerates China's domestic chip efforts. What to watch: the scale and conditions of H200 exports, whether more $1B-class neocloud deals follow, and how DeepSeek's in-house chip ambitions alter Chinese demand for Nvidia parts over time.