Former OpenAI and DeepMind researchers raise whopping $300M seed to automate science

Periodic Labs has emerged from stealth mode, unveiling a substantial $300 million seed funding round backed by major players in the tech industry. Notable investors include Andreessen Horowitz, DST, Nvidia, Accel, and prominent figures such as Jeff Bezos and Eric Schmidt. Founded by Ekin Dogus Cubuk and Liam Fedus, Periodic Labs is poised to transform scientific research. Cubuk, who previously led materials and chemistry initiatives at Google Brain and DeepMind, developed an AI tool named GNoME. This innovative tool made headlines in 2023 by discovering over 2 million new crystal structures, which researchers believe could fuel the next generation of technological advancements. Liam Fedus, a former VP of Research at OpenAI and a key contributor to the development of ChatGPT, also plays a vital role at Periodic Labs. He was instrumental in creating the first trillion-parameter neural network. The company’s team is packed with seasoned researchers who have contributed to significant AI and materials science projects, including OpenAI’s agent Operator and Microsoft’s MatterGen, an AI focused on materials science. With a mission to fully automate scientific discovery, Periodic Labs aims to create AI scientists that will conduct experiments within robotic laboratories. These labs will gather data, learn from their findings, and refine their methods over time. The primary objective is to innovate new superconductors that not only outperform current materials but do so with reduced energy requirements. However, the startup's ambitions extend beyond superconductors; they are also on the lookout for other groundbreaking materials. A key aspect of their strategy involves compiling comprehensive data generated by AI scientists as they manipulate various materials in pursuit of new discoveries. According to the company, previous advancements in scientific AI relied heavily on models trained using internet data, which they believe has reached its limits. Periodic Labs is set on redefining this by constructing autonomous laboratories equipped with AI scientists. While Periodic Labs represents one of the most formidable teams to enter this sector, the race to automate scientific discovery is gaining momentum. Numerous startups and academic institutions, including Tetsuwan Scientific, Future House, and the University of Toronto’s Acceleration Consortium, are also exploring the potential of AI in chemistry and materials research.