Periodic Labs just shattered seed funding records with a staggering $300 million round backed by tech's biggest names - Andreessen Horowitz, Nvidia, Jeff Bezos, and Eric Schmidt. The startup, founded by former OpenAI VP Liam Fedus and Google DeepMind's materials chief Ekin Dogus Cubuk, wants to automate scientific discovery itself by building AI scientists that run physical experiments autonomously.
Periodic Labs came out of stealth Tuesday with what might be the largest seed round in startup history - $300 million from a who's who of tech titans including Andreessen Horowitz, DST, Nvidia, Accel, and individual investors Jeff Bezos, Eric Schmidt, and Google's Jeff Dean. The round signals massive confidence in a audacious vision: fully automated scientific discovery.
The founding team reads like an AI hall of fame. Liam Fedus served as VP of Research at OpenAI and helped create ChatGPT, plus led development of the first trillion-parameter neural network. His co-founder Ekin Dogus Cubuk ran materials and chemistry at Google Brain and DeepMind, where he built GNoME - an AI tool that discovered over 2 million new crystals in 2023 according to DeepMind's research.
Their team includes researchers who worked on OpenAI's latest agent Operator and Microsoft's MatterGen, an AI specifically designed for materials science discovery. It's the kind of talent concentration that typically takes years to assemble, yet they've done it while the AI boom is still white-hot.
Periodic's approach tackles what the founders see as AI's next bottleneck. "Until now, scientific AI advances have come from models trained on the internet," the company explained in their launch blog post, "and LLMs have 'exhausted' the internet as a source that can be consumed." Their solution? Generate entirely new training data by having AI scientists conduct real physical experiments.
The plan involves building autonomous laboratories where robots mix chemicals, heat materials, analyze results, and iterate based on what they learn. Think of it as ChatGPT, but instead of processing text, it's discovering new superconductors by physically manipulating powders and raw materials in a lab setting.
Superconductors represent their first major target - materials that could conduct electricity without resistance, potentially revolutionizing everything from power grids to quantum computers. Current superconductors either require extremely cold temperatures or rare materials, making them impractical for widespread use. If Periodic's AI scientists can discover room-temperature superconductors using common elements, it could trigger the next industrial revolution.
