Periodic Labs Building AI Scientist for Materials Discovery
Periodic Labs, a startup founded by Liam Fedus (former OpenAI post-training lead, ChatGPT co-creator) and Dogus Cubuk (former Google DeepMind research scientist), is developing "synthesis superintelligence": an AI system that autonomously conducts experiments in physical labs to discover new materials, starting with high-temperature superconductors. The founders aim to close the loop between hypothesis, experiment, and learning by training AI on the scientific process itself, capturing full experimental traces. They have built a lab that can run thousands of experiments daily, using AI for experiment design, labeling, and analysis. The company raised a new round of funding to scale compute and infrastructure. They plan to monetize through custom systems for partners and potentially outcome-based pricing, retaining IP for major discoveries.
- •Periodic Labs is building 'synthesis superintelligence' to automate materials discovery.
- •Co-founded by Liam Fedus (ex-OpenAI) and Dogus Cubuk (ex-Google DeepMind).
- •First target: high-temperature superconductors.
