Sand AI Secures $200M Funding to Build Embodied Simulation Base with Video Models
In November 2026, video generation foundation model company Sand AI completed a $200 million funding round, bringing its total funding to approximately $300 million within ten months and achieving unicorn status. The rapid fundraising underscores strong investor confidence in video generation as a foundational technology for embodied AI.
Founded by Cao Yue, a core author of the Swin Transformer, Sand AI adheres to an autoregressive technical route. The company has open-sourced a 100-billion-parameter mixture-of-experts video model, slashing inference costs to one-tenth of the industry average. This open-source strategy lowers the barrier for developers and researchers to experiment with large-scale video generation.
Beyond video generation, Sand AI is exploring the use of its video models as a simulation foundation for embodied intelligence. By providing realistic virtual environments, these models could accelerate the training of robots and embodied agents, bridging the gap between virtual content and real-world robotics.