Moore Thread-U explained at the 2026 semi-annual performance briefing held on September 3 that the company has opened up a complete link of artificial intelligence from synthetic data generation, large model training, simulation verification to end-side hardware deployment. MT Lambda is a full-stack intelligent simulation platform. The company also open-sourced Mujoco Warp MUSA, a GPU-accelerated physical simulation backend based on the Musa architecture, and completed real machine verification of the four-legged robot dog and the two-legged humanoid robot Sim2Real. The company set up the Industrial Embodied Intelligence Innovation Center in Wuxi to be jointly built with ecological partners; at the same time, it also set up a “Embodied Intelligent Computing Power and Simulation Joint Laboratory” with the National Embodied Intelligent Application Pilot Base. Embodied intelligence requires collaboration between AI computation, physical simulation, graphics rendering, and end-side reasoning, which reflects the differentiated advantages of full-featured GPUs.

Moore Thread-U explained at the 2026 semi-annual performance briefing held on September 3 that the company has opened up a complete link of artificial intelligence from synthetic data generation, large model training, simulation verification to end-side hardware deployment. MT Lambda is a full-stack intelligent simulation platform. The company also open-sourced Mujoco Warp MUSA, a GPU-accelerated physical simulation backend based on the Musa architecture, and completed real machine verification of the four-legged robot dog and the two-legged humanoid robot Sim2Real. The company set up the Industrial Embodied Intelligence Innovation Center in Wuxi to be jointly built with ecological partners; at the same time, it also set up a “Embodied Intelligent Computing Power and Simulation Joint Laboratory” with the National Embodied Intelligent Application Pilot Base. Embodied intelligence requires collaboration between AI computation, physical simulation, graphics rendering, and end-side reasoning, which reflects the differentiated advantages of full-featured GPUs.