Liam Fedus, a former vice president of research at OpenAI, is raising at least 500 million dollars for Periodic Labs, the young company he co-founded with Ekin Dogus Cubuk, who was previously a research scientist at Google DeepMind. The round would value the startup at about 7.5 billion dollars and is led by AMP, the firm started by former Andreessen Horowitz partner Anjney Midha. According to people familiar with the talks, the raise was heavily oversubscribed, and there are already preliminary discussions about a fast follow round at an even higher price.
The idea behind Periodic Labs is a departure from the way most artificial intelligence has been built. Instead of training models on text scraped from the internet, the company runs autonomous robotic laboratories that carry out real physics and chemistry experiments and turn the results into fresh training data. The immediate target is the discovery of superconductors that work at higher temperatures, a long standing scientific prize with obvious value to industry, and the team is already working alongside semiconductor makers. In effect, the startup is trying to teach machines to do science by having them generate the evidence themselves.
The pace of the company's rise is its own story. Periodic Labs launched only in September 2025 with a 300 million dollar seed round at a 1.3 billion dollar valuation, which means the new deal would push its worth up nearly sixfold in roughly eight months. It also earned a place on Forbes' AI 50 Brink list in 2026. To staff the effort, the company has pulled in more than 20 researchers from Meta, OpenAI, and DeepMind, many of them walking away from large equity packages to join, a sign of how much conviction sits behind the project.
What the deal really captures is a shift in where ambition and money are heading in artificial intelligence. The first wave of the boom rewarded systems that could write and converse, but investors are now paying up for the promise of models that can produce new science rather than new sentences. The speed and size of this raise show how fiercely capital is competing for a small pool of elite researchers and for any credible path to discovery. Whether robotic labs actually deliver breakthroughs like a better superconductor remains unproven, yet the market is already pricing the possibility as if the answer is close.
