Top OpenAI, Google Brain researchers set off a $300M VC frenzy for their startup Periodic Labs (techcrunch.com)

Periodic Labs, founded by OpenAI veteran Liam Fedus and former Google Brain researcher Ekin Dogus Cubuk, emerged from stealth with a $300M seed round led by Felicis and backed by top-tier VCs and angels (a16z, Accel, DST, NVentures, Jeff Bezos, Elad Gil, Eric Schmidt, Jeff Dean, etc.). The raise followed a viral departure tweet from Fedus (OpenAI did not invest) and a flurry of inbound interest; the company has since hired heavyweight AI and materials-science talent and set up a wet lab to pair experimentation with AI. Their initial target is discovering new superconductors, though the robotics for powder synthesis still require training. Technically, Periodic is betting on the convergence of three maturing elements: reliable robotic arms for powder synthesis, efficient ML simulations capable of modeling complex physical systems, and LLMs with stronger reasoning/post-training capabilities to interpret results and propose next experiments. The approach mirrors a 2023 Google-led proof-of-concept that used language-model-suggested recipes and an automated lab to create 41 novel compounds. Importantly, founders view failed experiments as valuable labeled data for training and post-training, potentially redefining scientific incentives toward exploration. The effort could accelerate materials discovery and generate unique real-world datasets for AI, but it’s a high-risk, long-horizon play—robotic integration, experimental throughput, and the inherent unpredictability of discovery remain major hurdles.