Genesis AI has launched from stealth with a jaw-dropping $105 million seed round co-led by Eclipse Ventures and Khosla Ventures to create a foundational AI model for robots.

The startup was founded in December 2024 by Carnegie Mellon robotics Ph.D. Zhou Xian and ex-Mistral researcher Théophile Gervet.

Genesis aims to automate repetitive tasks such as lab work and housekeeping through a general-purpose robotics model.

Unlike large language models trained on text, robotics models need physical-world data—something Genesis hopes to shortcut using synthetic data.

Its proprietary physics engine, born out of an academic collaboration involving 18 universities, generates this synthetic data to simulate real-world interactions.

The company boasts a team of over 20 researchers with backgrounds in robotics, machine learning, and computer graphics.

Genesis claims its engine develops models faster than competitors relying on NVIDIA’s more common software.

Other startups in the race include Physical Intelligence, which snagged $400 million, and Skild AI, which hit a $4 billion valuation.

Genesis operates out of Silicon Valley and Paris, with plans to release its model to the robotics community by year-end.

Because what better way to train robots to mop floors than with imaginary messes in a virtual world?