The quiet part of innovation just got loud. Periodic Labs, co-founded by William (Liam) Fedus and Ekin Dogus Cubuk, just came out of stealth with a $300 million seed round led by Andreessen Horowitz, one of the largest in startup history. The investor lineup reads like a tech Mount Rushmore, DST Global, NVentures (NVIDIA’s venture arm), Accel, Felicis, Lightspeed, Bain Capital Ventures, CRV, Radical Ventures, A.Capital, and more. Sprinkle in Jeff Bezos, Eric Schmidt, Jeff Dean, and Elad Gil, and you’ve got the kind of backers who don’t just chase trends, they fund tectonic shifts.
Periodic Labs isn’t chasing hype; they’re engineering a new scientific rhythm. Their mission is to build AI scientists that don’t just analyze the world but interact with it, autonomous systems that can design, run, and learn from physical experiments in real time. Think of it as the first true fusion of intelligence and matter, where data isn’t scraped from the internet but born from experiments in physics and chemistry.
William Fedus brings the OpenAI and Google Brain pedigree, the guy who helped create ChatGPT and co-authored the Switch Transformer that scaled AI into the trillion-parameter era. Ekin Dogus Cubuk brings the deep science play, having led materials and chemistry teams at Google DeepMind and Brain, where he built GNoME, the model that discovered over two million new crystal structures. Together, they’re turning science into a living, learning system.
The team reads like an AI dream roster, alumni from OpenAI, DeepMind, Meta, Microsoft, and Samsung, people who walked away from massive stock packages to chase the next frontier. They’re building out autonomous labs in Menlo Park, where robotics systems run thousands of experiments monthly. The data from those runs doesn’t just fill spreadsheets; it fuels reinforcement learning systems that treat nature itself as the environment. That’s right, Periodic Labs is literally training AI on the laws of physics.
This isn’t another “AI for X” pitch. It’s a moonshot at the core of how knowledge is made. From tackling semiconductor heat dissipation to chasing high-temperature superconductors, Periodic Labs is placing science back in motion. When your models start learning directly from the universe, the feedback loop isn’t theoretical, it’s elemental.
$300 million says the next great discovery won’t come from the cloud, it’ll come from the lab.
