Periodic Labs

Periodic Labs is a research and technology venture developing autonomous laboratories and "artificial intelligence scientist" systems to accelerate experimentation and materials discovery in the physical sciences. Grounded in the principle that scientific knowledge is generated not only from textual data on the internet but also from experimental data, the organization jointly designs AI agents capable of conducting experiments and learning from their outcomes alongside the laboratory infrastructure in which they operate. As of 2025, its activities are primarily focused on physical sciences, with emphasis on materials science and condensed matter physics.

Periodic Labs develops autonomous systems that operate the cycle of hypothesis generation, experiment design, execution, and inference from results in a closed-loop manner. This approach aims to move beyond the limited scope of internet-derived textual data by generating high-volume, original, and often unpublished experimental data—including “negative results.” Autonomous laboratories provide measurement data at the gigabyte scale, forming the backbone of data pipelines integrated with reinforcement learning and physics-based modeling. Simulation-verifiable experimental setups with high signal-to-noise ratios enable rapid iterative research into physical and materials systems.

The core of its work is the systematic discovery and optimization of new functional materials. Objectives include identifying compounds that exhibit superconductivity at higher operating temperatures than existing materials, developing semiconductor packaging solutions that overcome thermal management and heat dissipation limitations, and generally automating materials design processes. Within this framework, physics-based simulation tools are linked with experimental discovery cycles; custom AI agents are trained on data sets provided by industrial partners to accelerate the pace of experiment design for engineers and researchers. The organization cites a collaboration with a semiconductor manufacturer facing thermal dissipation challenges as a practical example.

The founding team brings extensive experience in pioneering projects involving large language models and generative AI, deep learning-based atomistic models for materials discovery, scaling autonomous physics laboratories, and contributions to various materials discoveries over the past decade. Team members have contributed to materials intelligence models such as GNoME and MatterGen, as well as architectural innovations including operator/agent frameworks and attention mechanisms. This accumulated expertise establishes an institutional technical foundation for data- and reproducibility-driven scientific process automation.

The organization’s Scientific Advisory Board comprises distinguished academics in chemistry, materials science, and condensed matter physics. Professor Carolyn R. Bertozzi of Stanford University contributes through her work on cell surface glycosylation and the interface of biochemistry and chemistry. Professor Mercouri G. Kanatzidis of Northwestern University brings expertise in inorganic solid-state chemistry, chalcogenides, and thermoelectrics. Professor Zhi-Xun Shen of Stanford University contributes through his work in photon-based spectroscopy and imaging tools in condensed matter physics. Professor Steven A. Kivelson of Stanford University provides theoretical insights into high-temperature superconductivity and quantum liquid crystal phases. Professor Chris Wolverton of Northwestern University contributes through first-principles calculations, materials informatics, and high-throughput computational materials design.

Periodic Labs operates with support from venture capital funds and individual investors. Investors include funds such as a16z, Felicis, DST, NVentures, and Accel, as well as individual supporters including Jeff Bezos, Elad Gil, Eric Schmidt, and Jeff Dean. This funding is used to scale laboratory infrastructure, develop AI agents, and expand the team.

The organization has launched an Academic Grant Program to foster academic collaborations and knowledge exchange within the research ecosystem. The program aims to support bold ideas and pioneering projects. On the industrial side, it seeks to leverage AI to process corporate experimental data and accelerate experiment design cycles through partnerships with materials- and manufacturing-focused companies. Internally, the distinction between “Bits” (software, models, data) and “Atoms” (experimental apparatuses, laboratory automation) defines an organizational framework for the co-development of digital and physical components.

As of 2025, Periodic Labs is focused on developing autonomous laboratory infrastructure and scientific AI agents centered on the physical sciences. The organization is building academic networks and industrial partnerships to scale research outputs, expanding laboratory capacity and accelerating discovery rates in new classes of materials. Career opportunities and open positions indicate growing organizational staffing needs.

The organization’s methods are tracked through metrics such as experimental efficiency, hypothesis-validation cycle time, simulation-validation consistency, and the rate at which material candidates meet targeted performance benchmarks. Outputs include identification of new compound candidates, optimization of thermal and electrical properties, design recommendations for packaging and manufacturing processes, and performance reports derived from industrial pilot applications. This framework aims to align AI-assisted scientific discovery with principles of reproducibility and verifiability.