Industry insiders say Chinese robot makers currently rely on Nvidia silicon and software; Nvidia's physical AI business generates ~$10B in annual revenue
Business in ‘physical AI’ is growing, and Chinese companies rely on U.S. chips and software — Nvidia's chips aren't just for training chatbots.
Context & Ripple Effects
Nvidia's China exposure has already been shaped by constraints and workarounds: Chinese AI companies were reported to seek overseas rentals of Nvidia compute, while Alibaba and Baidu had begun using internally designed training chips. The reported reliance of robot makers on Nvidia hardware and software places physical AI alongside those contested compute paths, while the company’s roughly $10 billion annual physical-AI revenue shows the category is material to its business.
Nvidia’s production base is also increasingly concentrated in Asia, according to a May analysis of the company’s supplier costs. That makes a robotics-led expansion of its silicon-and-software stack a demand opportunity tied to the same regional supply network.
First-order effects
- Nvidia’s physical-AI unit gains a sizable revenue base beyond chatbot training, with Chinese robot makers reportedly dependent on both its silicon and software.
- Chinese robot makers using Nvidia’s stack face a combined hardware-and-software dependency rather than a chip-only purchasing decision.
Second-order effects
- Alibaba and Baidu’s move toward self-designed chips for model training highlights a different challenge for local alternatives: matching Nvidia’s robotics software environment as well as supplying compute.
- Asian suppliers supporting Nvidia’s production become more exposed to demand from physical-AI systems as that business expands alongside data-center computing.
Third-order effects
- If robot makers continue to standardize on Nvidia’s integrated stack, competition in physical AI will turn on ecosystem compatibility and deployment tooling, not only on access to accelerators.
- The story fits a widening split in China’s AI supply chain: local chips can reduce dependence in selected training workloads, while embodied AI may preserve demand for full foreign platforms.
The trend: Physical AI is extending the AI hardware contest from training compute into integrated silicon-and-software platforms for robots.