DeepMind & OpenAI Alums Raise $300M to Unleash 'Robot Scientists' on Materials Discovery

AI's New Frontier: Periodic Labs Aims to Automate Scientific Discovery with Robot Scientists and a $300 Million War Chest

San Francisco, 2025-10-10 - A new venture, Periodic Labs, has surfaced from stealth with a substantial $300 million in seed funding, marking a significant escalation in the quest for automated scientific discovery. Supported by a consortium of prominent investors including a16z, Nvidia, former Google CEO Eric Schmidt, and Amazon founder Jeff Bezos, the startup plans to harness AI-powered robot scientists to conduct self-guided experiments and produce novel scientific data.

The ambition of Periodic Labs is to overcome the perceived constraints of internet-trained AI models by creating purpose-built autonomous laboratories that generate proprietary, high-grade empirical data. This strategy, articulated by the company's founders, squarely tackles a rising worry within the AI community: the possible depletion of publicly-available datasets for advanced model training.

"Until now, scientific AI advances have come from models trained on the internet. But despite its vastness-it's still finite. And in recent years the best frontier AI models have fully exhausted it. Autonomous labs provide huge amounts of high-quality data and give our AI scientists the tools to act," the company's founders state on their website.

Periodic Labs was co-founded by Ekin Dogus Cubuk, formerly of DeepMind, and Liam Fedus, a former OpenAI researcher. Cubuk's prior work at Google Brain and DeepMind included leading the materials and chemistry team and developing GNoME, an AI tool credited with discovering over 2 million new crystals in 2023. Fedus, recognized for his contributions to ChatGPT and for leading the creation of the first trillion-parameter neural network at OpenAI, brings extensive large-scale AI development experience. The company's scientific advisory board comprises professors from Stanford and other leading academic institutions, lending further credibility to its ambitious agenda.

The immediate objective for Periodic Labs is the discovery of novel superconductors that operate at higher temperatures than currently available materials. This pursuit is not merely academic; the company posits that automating materials design will accelerate advancements in electronics, space travel, and nuclear fusion-sectors critically dependent on breakthroughs in material science.

The operational model involves coordinating robots to perform physical experiments, collect empirical data, and iteratively refine approaches, fostering a continuous learning loop for their AI systems. This closed-loop system is designed not only to generate discoveries but also to create vast, unique datasets that can be fed back into AI models, further enhancing their capabilities and potentially sidestepping future data scarcity issues.

The concept of integrating AI for automated chemistry and materials discovery is not entirely nascent. Academic research on this front dates back to at least 2023, with pioneering efforts by institutions such as the University of Toronto's Acceleration Consortium and early-stage startups like Tetsuwan Scientific. However, Periodic Labs distinguishes itself with an unprecedented level of capital infusion at the seed stage, suggesting a strong investor belief in the team's ability to execute this complex vision at scale.

The $300 million seed round included participation from Andreessen Horowitz, DST, Accel, and Elad Gil, alongside the aforementioned tech luminaries. This financial backing positions Periodic Labs to rapidly scale its operations, acquire advanced robotics, and recruit top-tier AI and materials science talent, moving beyond theoretical discussions to practical implementation of AI-driven scientific laboratories.

The implications of Periodic Labs' approach extend beyond specific material discoveries. If successful, the model of autonomous labs generating and consuming their own empirical data could fundamentally alter the pace and methodology of scientific research across various disciplines, potentially accelerating innovation cycles exponentially. The initiative represents a bold bet on the future of AI not just as a computational tool, but as an active participant in the empirical scientific process.