On September 3rd, during the 2026 Mid-Year Performance Briefing, Moore Threads-U announced that it has established a complete embodied AI pipeline, covering synthetic data generation, large model training, simulation verification, and edge hardware deployment. The company launched the full-stack embodied AI simulation platform, MTLambda, and open-sourced the GPU-accelerated physics simulation backend, MuJoCoWarpMUSA, based on the MUSA architecture. It has completed Sim2Real real-world verification for a quadruped robotic dog and a bipedal humanoid robot. Additionally, Moore Threads-U established an Industrial Embodied AI Innovation Center in Wuxi, collaborating with ecosystem partners, and founded the 'Embodied AI Computing Power and Simulation Joint Laboratory' with the National Embodied AI Application Pilot Base. Embodied AI requires the coordination of AI computing, physics simulation, graphics rendering, and edge inference, highlighting the differentiated advantages of a full-function GPU.