Ex-OpenAI and DeepMind Scientists Secure Record $300 Million Seed Funding for AI-Powered Science Automation

Periodic Labs emerged from stealth mode on Tuesday, announcing a monumental $300 million seed round. The funding is backed by a veritable all-star lineup from the tech industry, including Andreessen Horowitz, DST Global, Nvidia, Accel, Elad Gil, Jeff Dean, Eric Schmidt, and Jeff Bezos.

The company was founded by Ekin Dogus Cubuk and Liam Fedus. Cubuk previously led the materials and chemistry team at Google Brain and DeepMind. There, he spearheaded projects like GNoME, an AI tool that in 2023 identified over 2 million new crystals—materials researchers believe could fuel future technological breakthroughs.

Fedus is a former Vice President of Research at OpenAI and was one of the key researchers behind the creation of ChatGPT. He also led the team that developed the first trillion-parameter neural network.

The core team is similarly composed of researchers with deep experience in major AI and materials science initiatives, ranging from developing OpenAI's agent, Operator, to contributing to Microsoft's MatterGen, an LLM for materials discovery.

Periodic Labs states its ambitious mission is to automate scientific discovery by creating AI scientists. This involves building autonomous laboratories where robotics systems conduct physical experiments, gather data, iterate on findings, and continuously learn and improve.

The lab's primary initial objective is to develop new superconductors that offer superior performance and potentially greater energy efficiency than current materials. However, the well-capitalized startup also aims to discover a wide range of other novel materials.

A parallel goal is to systematically aggregate all the physical-world data generated by its AI scientists as they mix, heat, and manipulate various powders and raw materials in their quest for innovation.

"Historically, advances in scientific AI have relied on models trained on internet data," the company notes in an introductory post, suggesting that large language models have largely "exhausted" the web as a scalable data source. "At Periodic, we are building AI scientists and the autonomous laboratories for them to operate within."

The vision is that these labs will not only invent next-generation materials but also produce a continuous stream of invaluable, fresh experimental data to fuel the ongoing evolution of AI models.

While this assembly of talent is remarkable, Periodic Labs is not alone in pursuing AI-driven scientific discovery. Using AI to automate chemistry breakthroughs has been an active academic research area since at least 2023. It's also the focus of other entities, from smaller startups like Tetsuwan Scientific to nonprofits such as Future House and the University of Toronto's Acceleration Consortium.