Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Periodic Labs, a new startup from one of OpenAI’s most respected researchers, Liam Fedas, and his ex-Google brain colleague, Akin Douglas QBook, came out of the blue last month. With a massive $300 million seed round. It was headed by Felicis and included an angel and other top VCs.
The startup began when Fedas had a conversation with Qubuk (whose friends call him “Doz”) about seven months ago. Qubuk was one of Google Brain’s lead machine learning and materials science researchers. What does it feel like after endless Silicon Valley? GenAI will radically change scientific discoveryThey decided the pieces were finally there to make it a reality. Or at least found a startup that tried it.
“Some things have happened in the LLM field, in experimental science and in simulation, that have created such precise timing,” Qubuk told TechCrunch.
For one, he said, robotic arms that can handle powder synthesis — the process of mixing and creating new materials — have recently proven themselves reliable. For another, machine learning simulations have become efficient and accurate enough to model complex physical systems such as those needed to develop new materials.
And, third, LLM now had powerful reasoning capabilities — in part through the work of Fedas and his team at OpenAI. Fedas was one of the small teams that initially built ChatGPT and ran OpenAI Uber important post training teamwhich refine the models after their initial development.
Stitched together, the picture was clear: a simulation could theoretically discover new compounds, a robot could mix ingredients, and an LLM could analyze the results and suggest course corrections. AI-automated materials were ready to build.
In fact, Qubuk was one of the researchers who published a groundbreaking paper in 2023 Documenting a pioneer Google research project. The team built a fully automated, robotic-operated lab and produced 41 novel compounds from recipes suggested by the language model.
TechCrunch event
San Francisco
|
October 27-29, 2025
Equally important, the founders realized that even failed experiments would be valuable to their new startup because data is the lifeblood of AI. AI science offers an entirely new source for real-world training and post-training data. This, the founders believed, could turn the existing scientific motivation system on its head, which seeks success, not research, rewarded through paper publication and grants.
“Getting in touch with reality, experimenting [AI] Loop — we think that’s the next frontier,” Fedas told TechCrunch.
Felicis won the contract; OpenAI does not invest
After a discussion with Qbook, Fedas approached the Powers of OpenAI to share his resignation and his plans. He was happy Tweet OpenAI’s blessing and investment seems to be the world he’s leaving.
However, that investment did not materialize. OpenAI is not a supporter of the periodical, the founders confirmed to TechCrunch. And while FedS declined to say why, they didn’t actually need OpenAI’s money.
Fedus’ tweet set off a VC frenzy for the company. “There was almost a feeling of being the opposite pitch. An investor actually wrote a love letter to Periodic Labs,” laughs Fedas, explaining that neither he nor QBuck “knew what to do with it.” Others sent multi-page documents pitching themselves.
But the first call they took was from Peter Deng, a former OpenAI colleague turned investor for top-tier seed firm Felices. (Deng left OpenAI for Felicis in early 2025.)
“Liam is a big deal at OpenAI, well-loved and a very influential researcher,” Deng told TechCrunch. “When I heard he was gone, I texted him right away.”
Deng met Fedas for coffee in San Francisco’s No Valley neighborhood. Hyped with caffeine and excitement, Fedas invites Deng to end their conversation with a walk in the area’s famous mountain range. Pitch walks may be a Silicon Valley trope, but they really do happen.
The cold day turned hot. Deng, wearing a sweater, was sweating and scrambling to keep up with the fit and friendly Fedas until the founder said something that “literally stopped me in my tracks,” Deng told TechCrunch. He told Deng that “everyone talks about doing science, but to do science you have to actually do science,” Deng recalled.
In other words, they needed to give the AI a fully equipped wet lab to try out its ideas in a real-world, controlled setting.
“The truth about these models is that everything the model knows is within a normal distribution. We take a bunch of data, and it can rearrange what it knows,” Deng said.
Hypotheses must be tested to discover something new.
“And I committed on the spot to write the check, in the middle of the mountains of No Valley,” Deng said.
Fedas also remembers the moment Deng asked how he could get involved and Fedas told him the startup needed cash for laptops and a temporary office. And “He’s good, great, I’ll pay you right now. And it’s just this huge vote of confidence.”
But Deng didn’t actually get his checkbook out on the street. He went back to the office excited for the deal only to be confronted by Felicis’ lawyer, who pointed out that the firm couldn’t immediately sign a contract: the company hadn’t been incorporated yet. It didn’t even have a name, much less a bank account to wire funds. Deng smiled and said, “How quick we were.
Soon they had all that stuff and all the term sheets they could handle. With a $300 million war chest, QBook and Fedus have more than two dozen prestigious AI and scientific talents such as Alexander Passos (creator of o1 and o3); Eric Tauberer (a materials scientist who had already discovered the original superconductor) and Matt Horton, creators of two of Microsoft’s GenAI materials science tools. And The list goes on.
Because the team members all specialize in fields ranging from AI to physics, each week one of them gives a grade-level lecture to the others. “We think that a tight coupling is very important,” Cubic said. He wants everyone to understand all the parts of what they are building.
Periodic Labs has already set up its lab, and is working on experimental data, simulations and testing some predictions. The main initial mission is to find new superconductor materials – potentially a gold mine of discovery. Advanced superconductors could power the next era of powerful, but low-energy technologies.
But the last part – the robot – is still up and running. “They will take some time to train,” Cubic said.
All of this, of course, is a big swing for the fences. Whether powered by AI or not, scientific discovery is generally not fast, easy or predictable. While there’s some indication that this group of experts will find what they’re looking for — or make other discoveries along the way (or simply generate valuable data about their failures), there’s no guarantee.
And we know that the model builders themselves are making their way toward more AI science. OpenAI VP Kevin Weil last month Says he’s launching an OpenAI for science The company’s unit is to “build the next great scientific machine: an AI-powered platform that accelerates scientific discovery.”
The investor who wrote the love letter didn’t win the deal (though Fedas said the letter was “very flattering.”) Other seed investors include Andreessen Horowitz, DST, NVIDIA’s venture capital arm NVentures, Accel, and angels like Jeff Bezos, Eliad Gil, Eric and Jeff Gilt.
Elad Gil will talk about how AI has changed the startup landscape Disruption in San Francisco 29 October.
