Nvidia has denied a report from The Information alleging plans for small-batch shipments of a Groq-licensed LPU to Chinese customers by year-end. The company stated on August 20, 2026, that it has zero current LPU sales in China and no China-tailored LPU on its roadmap. The reported chip relied on Groq technology to create a compliant inference unit optimized for AI chatbot workloads that navigates U.S. export restrictions. Earlier Reuters reporting in March 2026 described Nvidia working on a China-compatible inference chip. This denial occurs as China once accounted for roughly 20% of Nvidia's data center revenue before export controls tightened, with CEO Jensen Huang noting in May 2026 that the company largely ceded the market to Huawei. The U.S. has approved limited H200 shipments to ByteDance and Tencent, each receiving about 10,000 units, while the Vera Rubin architecture is restricted from China.