Periodic Labs, a venture from former Google and OpenAI researchers, has emerged from stealth mode after securing a $300 million seed funding round on Tuesday. The company intends to automate scientific discovery using AI-powered robotic laboratories.
The investment round was led by a group of technology investors and companies, including Andreessen Horowitz, DST, Nvidia, Accel, Elad Gil, Jeff Dean, Eric Schmidt, and Jeff Bezos. The company was established by co-founders Ekin Dogus Cubuk and Liam Fedus. Cubuk formerly headed the materials and chemistry team at Google Brain and DeepMind. During his tenure, he developed an AI tool called GNoME, which discovered over 2 million new crystals in 2023. Researchers believe these materials could be used to power future technologies.
Co-founder Liam Fedus is a former Vice President of Research at OpenAI, where he was among the researchers who helped create ChatGPT. He also led the team that built the first trillion-parameter neural network. Periodic Labs has assembled a small team of researchers with experience on other major AI projects. Their past work includes building OpenAI’s agent Operator and contributing to Microsoft’s MatterGen, a large language model specifically designed for materials-science discovery.
The company’s mission is to automate the scientific discovery process by developing “AI scientists.” This strategy involves the construction of physical labs where robots will conduct experiments, collect data from those experiments, and iterate on the process autonomously. The systems are designed to learn and improve as they run successive trials, manipulating various powders and raw materials by mixing, heating, and other methods.
Periodic Labs’ initial focus is the invention of new superconductors that could offer improved performance and potentially require less energy than existing materials. In addition to superconductors, the well-funded startup aims to find other novel materials through its automated process. A parallel goal is to collect and compile all the physical-world data generated by its AI-driven experiments as they search for new discoveries.
