Chinese chipmaker Moore Threads Technology, dubbed "China's little Nvidia," launched its next-generation Huagang GPU architecture alongside the AI-focused Huashan chip and gaming-oriented Lushan chip at the MUSA Developer Conference in Beijing on December 20, 2025. CEO Zhang Jianzhong, a former Nvidia executive, claimed Huashan outperforms Nvidia's Hopper series (H100/H200) in computing power, memory bandwidth, and capacity, nearing Blackwell levels, while enabling AI clusters over 100,000 cards with 50% higher density, 10x energy efficiency, and full-precision tensor acceleration.
The rollout follows Moore Threads' blockbuster Shanghai IPO on December 5, with shares surging 480% to 664.1 yuan, fueling ambitions to cut reliance on U.S. chips amid export curbs. Lushan promises 15x better AAA gaming performance via AI Generative Rendering and enhanced ray-tracing, both entering mass production in 2026 on the versatile MUSA platform supporting diverse workloads beyond pure AI. Additional reveals included the KUAE cluster for massive AI models, underscoring software compatibility with mainstream ecosystems.
This escalation in China's homegrown AI race positions Moore Threads against Huawei and Baidu, targeting developers wary of foreign delays while projecting 2025 sales of $170-210 million and breakeven by 2027. Analysts eye its "full-function" edge for broader adoption, testing U.S. dominance in a geopolitically charged semiconductor arena.