Periodic Labs, a groundbreaking AI startup founded by William (Liam) Fedus, formerly of OpenAI, and Ekin Dogus Cubuk, a former Google Brain researcher, has emerged from stealth with an impressive $300 million seed round led by Felicis. The round includes backing from an elite roster of investors such as Andreessen Horowitz, DST, NVentures, Accel, Jeff Bezos, Elad Gil, Eric Schmidt, and Jeff Dean.
The company’s mission is to revolutionize materials science through a fusion of AI, robotics, and simulation. Periodic Labs aims to use machine learning models and robotic experimentation to accelerate the discovery of new materials — from superconductors to advanced compounds — by combining real-world experimentation with AI-driven insights.
The concept was conceived roughly seven months ago when Fedus and Cubuk began exploring how generative AI could transform scientific discovery. They envisioned a system that could integrate robotic arms for powder synthesis, machine learning models for rapid simulations, and large language models (LLMs) for reasoning and optimization. Their goal: an AI-powered lab capable of conducting experiments, analyzing data, and improving results autonomously.
Cubuk’s previous research at Google demonstrated the potential of this approach, where an AI-guided robotic lab successfully produced 41 novel compounds. The founders believe that even failed experiments provide essential data, feeding into an AI-driven feedback loop that continuously enhances the discovery process.
Investor interest surged following Fedus’ departure from OpenAI, triggering what insiders described as a “funding frenzy.” Felicis’ Peter Deng reportedly secured the deal during an informal “pitch walk” in San Francisco. Despite speculation, the startup confirmed it did not receive any financial backing from OpenAI.
Currently, Periodic Labs employs over two dozen specialists across AI, physics, and materials science. The team holds weekly internal lectures to promote cross-disciplinary collaboration, with an initial focus on discovering new superconducting materials while training its robotic systems for more advanced synthesis tasks.
While the approach is ambitious and inherently high-risk, the founders see this as the next major leap in scientific research — where AI doesn’t just assist but actively drives discovery. Periodic Labs joins a growing ecosystem of AI-for-science initiatives, including OpenAI’s own research unit, all aiming to redefine how innovation happens in the age of intelligent experimentation.
