Periodic Labs, a new startup founded by former OpenAI and DeepMind researchers, has emerged from stealth with a $300 million seed funding round—the largest of its kind for a science-focused AI venture. Backed by a powerhouse group of investors including Andreessen Horowitz, DST, Nvidia, Accel, Elad Gil, Jeff Dean, Eric Schmidt, and Jeff Bezos, the company is poised to redefine how scientific discovery is conducted. The company was co-founded by Ekin Dogus Cubuk and Liam Fedus. Cubuk led the materials and chemistry research team at Google Brain and DeepMind, where he played a key role in developing GNoME, an AI system that discovered over 2 million new crystal structures in 2023—some of which could enable breakthroughs in energy storage, electronics, and advanced materials. Fedus, a former Vice President of Research at OpenAI, was instrumental in the development of ChatGPT and led the team that built the first trillion-parameter neural network, one of the largest AI models ever created. Periodic Labs’ mission is ambitious: to build autonomous AI scientists capable of designing experiments, running physical tests, analyzing results, and iterating independently—essentially creating self-improving laboratories. The team’s first target is the discovery of next-generation superconductors that could operate at higher temperatures and lower energy costs than current materials, potentially revolutionizing clean energy and quantum computing. The startup’s approach combines advanced robotics with large-scale AI models trained on real-world experimental data. As robots conduct trials—mixing powders, heating materials, and measuring outcomes—the AI learns from the physical world, generating insights that traditional models trained on internet data cannot access. The company aims to collect and preserve every piece of data produced in this process, creating a vast, high-quality dataset that can fuel future AI advancements. In a blog post, Periodic Labs argues that the era of training AI on internet text has reached its limits. “LLMs have exhausted the internet as a source of new information,” the post states. “At Periodic, we are building AI scientists and the autonomous laboratories for them to operate.” While Periodic Labs stands out for the caliber of its founding team and its substantial funding, it is not alone in this pursuit. Other startups like Tetsuwan Scientific, non-profits such as Future House, and academic initiatives like the University of Toronto’s Acceleration Consortium are also exploring AI-driven materials discovery. However, Periodic’s blend of deep technical expertise, robust financial backing, and a clear path toward physical experimentation sets it apart. The company’s vision goes beyond inventing new materials—it aims to create a new feedback loop where AI scientists generate real-world data, which in turn trains even smarter AI, accelerating scientific progress at an unprecedented pace.