On September 3, during the 2026 mid-year performance review, Moore Threads unveiled that it has successfully constructed a comprehensive embodied AI chain. This chain encompasses every crucial phase, from synthetic data generation and large-scale model training to simulation verification and edge hardware deployment. Notably, its full-stack embodied AI simulation platform, MT Lambda, has open-sourced the GPU-accelerated physics simulation backend, MuJoCo Warp MUSA, which is built upon the proprietary MUSA architecture. This platform has undergone rigorous Sim2Real real-world validation, demonstrating its efficacy with quadruped robot dogs and bipedal humanoid robots.

In addition to these technological advancements, Moore Threads has taken a significant step forward by establishing an Industrial Embodied AI Innovation Center in Wuxi. This center fosters collaboration with ecosystem partners to drive innovation in the field. Furthermore, Moore Threads has co-founded the "Embodied AI Computing Power and Simulation Joint Laboratory" alongside the National Embodied AI Application Pilot Base, solidifying its commitment to advancing the frontiers of embodied AI research and development.

Embodied AI represents a sophisticated fusion of AI computing, physics simulation, graphics rendering, and edge inference capabilities. The successful implementation of such a system underscores the unique and differentiated advantages offered by a full-function GPU, positioning Moore Threads at the forefront of this transformative technological wave.