AI-Powered Acceleration of Physics Research Through Periodic Labs

SAN FRANCISCO / LONDON – Periodic Labs, a startup founded by Ekin Dogus Cubuk, is pioneering a new approach to materials discovery by integrating large language models (LLMs) with physics research. The company aims to accelerate research in solid-state physics, with a particular focus on discovering new superconductors.

Bridging the Gap Between Theory and Experiment

Ekin Dogus Cubuk, co-founder and co-CEO of Periodic Labs, previously worked at Google Brain and DeepMind. He observed limitations in linking theoretical work with experimental validation within large software companies. Periodic Labs was created to address this challenge, fostering a collaborative environment for physicists and AI experts. The company’s goal is to streamline the process of materials discovery, particularly in the search for improved superconductors.

Leveraging Large Language Models for Materials Science

Periodic Labs utilizes LLMs to analyze extensive datasets and identify patterns that might be missed by human researchers. This capability allows for the generation of new hypotheses and the design of more efficient experiments. The company aims to combine theoretical and experimental physics, accelerating research and development in the field.

A Well-Funded Venture

In September 2025, Periodic Labs secured $300 million in seed funding from investors including Andreessen Horowitz, DST, Nvidia, Accel, Elad Gil, Jeff Dean, Eric Schmidt, and Jeff Bezos. This substantial investment underscores the potential of AI-driven materials science and will enable the company to build and operate physical labs where robots conduct experiments, collect data, and iterate on designs.

Focus on Superconductors and Beyond

While the initial focus is on discovering new superconductors – materials with the potential to revolutionize energy transmission and storage – Periodic Labs’ technology is applicable to a broader range of materials research. The company’s long-term vision is to automate scientific discovery, creating AI scientists capable of accelerating innovation across multiple disciplines.

Ekin Dogus Cubuk’s Background

Ekin Dogus Cubuk holds degrees from Swarthmore College (BA, physics and BS engineering, 2010) and Harvard University (PhD, applied physics, 2016). Prior to founding Periodic Labs, he led the materials and chemistry team at Google Brain and DeepMind, where he developed an AI tool called GNoME that discovered over 2 million new crystals in 2023.

Liam Fedus, co-founder of Periodic Labs, is a former VP of Research at OpenAI and played a key role in the development of ChatGPT and the first trillion-parameter neural network.