Nvidia is developing a specialized version of Groq chips for the Chinese market

Last year Nvidia acquired the rights to Groq’s startup technology for $17 billion and this month unveiled its own LPU chips that accelerate AI inference. Sources say the company is preparing a Chinese version of these chips in light of U.S. export restrictions.

Reuters reported on the development based on internal information. Meanwhile, Nvidia CEO Jensen Huang announced the resumption of production of its H200 processors, which are slated for sale in China. The company has obtained the necessary U.S. export licenses and Chinese import approvals.

China hosts many competitors producing specialized inference chips, so supplying adapted Groq solutions is critical to maintaining Nvidia’s position in that market. Reuters clarified that the adaptation will not reduce chip performance but simply enable them to work with local systems. Deliveries are set to begin in May; Chinese batches will launch alongside the main ones.

What’s new?
- Deal for $17 billion: Nvidia gained access to Groq developments.
- LPU chips: specialized processors that accelerate AI inference.
- Chinese version: being developed under U.S. export restrictions.
- H200 resumption: Nvidia is again producing its own chips for the Chinese market.
- Licenses and import: obtained from the U.S.; approved in China.
- Competition in China: many local inference‑chip manufacturers.
- Groq adaptation: focused on compatibility, not performance loss.
- Supply launch: planned for May; batches will run concurrently with main ones.

Thus Nvidia is actively expanding its AI chip portfolio and ensuring licensed products are available in China to preserve competitiveness in this important market.