Former OpenAI and DeepMind researchers raise whopping $300M seed to automate science (techcrunch.com)

Periodic Labs emerged from stealth with a $300 million seed round led by top-tier investors (a16z, DST, Nvidia, Accel) and luminaries including Jeff Dean, Eric Schmidt and Jeff Bezos. Founded by Ekin Dogus Cubuk (formerly leading materials and chemistry at Google Brain/DeepMind — his GNoME project found >2 million candidate crystals in 2023) and Liam Fedus (ex-OpenAI VP of Research and a creator of the first trillion-parameter model and ChatGPT-era systems), the small team includes researchers from OpenAI and Microsoft materials-AI efforts. The company’s bold pitch: build “AI scientists” and fully autonomous laboratories where robots run physical experiments, collect data, and iteratively learn — effectively automating the scientific discovery loop. Technically, Periodic is pursuing closed-loop experimentation: automated hardware + active-learning models that propose experiments, execute them, and retrain on fresh physical-world measurements. Their initial target is next-generation superconductors with better performance and potentially lower energy requirements, but the platform aims to generalize to broader materials discovery. Critically, Periodic argues internet-scale LLM training has hit limits and that novel physical datasets produced by these labs will be essential to advance models for science. The large seed and heavyweight team signal high conviction (and capital intensity) for robotized lab infrastructure and data curation, positioning Periodic among a growing cohort of startups and consortia racing to industrialize AI-driven chemistry and materials science.