Fei-Fei Li's World Labs buys SceniX to push into robotics
What's the deal? World Labs, the startup founded by AI pioneer Fei-Fei LiDealroom has a profile for this one. Try Dealroom →, has acquired SceniXDealroom has a profile for this one. Try Dealroom →, a robotics company that builds high-fidelity simulation systems for training robots. The deal — announced Tuesday — brings SceniX's cofounders Yunzhu Li and Changxi Zheng into World Labs, along with a team focused on simulation platforms where robots learn skills before running on physical hardware. Terms were not disclosed.
What's the endgame? The acquisition is World Labs' biggest move yet into embodied AI — systems that perceive, reason, and act in physical spaces rather than staying confined to screens. The company wants to combine world models, learning-based simulation, and real-world feedback into a single training loop.
Who joins: SceniX cofounders Yunzhu Li and Changxi Zheng come aboard, bringing expertise in hybrid simulation for general-purpose robotics. Their software creates realistic environments where robots practice manipulation and movement, with what the company says is strong transfer from simulation to real-world deployment.
Why it fits. SceniX's work pairs with World Labs' flagship product, MarbleDealroom has a profile for this one. Try Dealroom →, which generates persistent, spatially consistent 3D environments from text, images, videos, or rough layouts. Connecting those AI-generated worlds to real robotic systems is the stated aim.
Why now? The "sim-to-real" approach has become a top priority in robotics. Training robots entirely in the physical world is slow, costly, and risky, so high-fidelity simulations let developers generate millions of training scenarios cheaply, then transfer learned behaviours onto actual machines.
"The world is not just made of words, and spatial intelligence was never just about perceiving and generating worlds. It's about interacting with them," Li wrote on X, confirming SceniX was joining World Labs.
The signal: Large language models changed how people use computers; the next phase centres on systems that understand physical environments, predict what comes next, and act inside them through robots. World Labs is betting that its work in generative worlds and SceniX's simulation tech together form a path from virtual training grounds to machines that operate in the real world.
Read more: TechStartups
Image credit: Generated with Gemini
