Nvidia denies report it will ship Groq-based LPUs to China by year-end

NVIDIA has denied a report claiming it plans to ship Groq-based language processing units (LPUs) to China by the end of 2026, despite earlier claims of pre-orders and product development. The company ... Read original →

NVIDIA's denial of exporting Groq-based LPUs to China underscores the deeper dynamics of U.S.-China semiconductor rivalry. Technologically, the chip uses Samsung's 4nm process and avoids HBM to comply with U.S. export controls, yet this does not guarantee market access in China, where import restrictions remain stringent. Huawei’s Ascend 950DT is emerging as a direct rival, accelerating its Q4 2026 launch. While some U.S. export restrictions have eased, Chinese firms still face tight controls on high-end AI hardware. Companies must reassess supply chain resilience and local alternatives. In the short term, NVIDIA may pivot toward customized solutions, while domestic players leverage this gap to accelerate innovation. Over the next 12–24 months, the trend toward technological decoupling in AI chips will intensify, reshaping the global semiconductor landscape.