Insider Brief
- Periodic Labs emerged from stealth with a $300 million seed round led by Andreessen Horowitz to build “AI scientists” and autonomous labs, funding hiring, lab scale-up, and initial industry products.
- Backers include Felicis, DST Global, Accel, NVentures, and individual investors Jeff Bezos, Eric Schmidt, Jeff Dean, and Elad Gil; founders William Fedus and Ekin Dogus Çubuk cite team contributions to ChatGPT, GNoME, Operator/Agent, the attention mechanism, and MatterGen.
- The company will start in the physical sciences—targeting higher-temperature superconductors and industrial use cases like chip heat dissipation—by coupling large models with automated experimentation to generate proprietary data and compress R&D cycles.
Periodic Labs emerged from stealth with a $300 million seed round led by Andreessen Horowitz (a16z), a bet that “AI scientists” and autonomous labs can accelerate discoveries in materials and other physical sciences. The startup—co-founded by William Fedus and Ekin Dogus Çubuk—said the financing will fund hiring, scale out its laboratory infrastructure, and bring its first products to industry partners.
Along with a16z, funding came from Felicis, DST Global, Accel and Nvidia’s venture arm NVentures, joined the round alongside individual backers including Jeff Bezos, Eric Schmidt, Jeff Dean and Elad Gil, the company noted in an X post. Terms weren’t disclosed, but the company framed the raise as the capital base needed to pair large AI models with automated experimentation—software that forms hypotheses and directs robots to run and learn from real-world tests.
Periodic is starting in the physical sciences, where experiments are relatively fast, results are verifiable, and simulations can narrow the search space, Fedus wrote.
“We’re starting here because experiments have high signal-to-noise and are (relatively) fast, physical simulations effectively model many systems, but more broadly, physics is a verifiable environment,” Fedus wrote in a Post on X on Tuesday. AI has progressed fastest in domains with data and verifiable results – for example, in math and code. Here, nature is the RL environment.”
Fedus pointed out Periodic Labs aims to discover higher-temperature superconductors to enable next-gen transportation and low-loss power grids. More broadly, it wants to automate materials design to accelerate progress across Moore’s Law, space travel, and nuclear fusion.
“We’re also working to deploy our solutions with industry. As an example, we’re helping a semiconductor manufacturer that is facing issues with heat dissipation on their chips,” Fedus wrote. “We’re training custom agents for their engineers and researchers to make sense of their experimental data in order to iterate faster.”
The founding team helped create ChatGPT, DeepMind’s GNoME, OpenAI’s Operator/Agent, the neural attention mechanism, and MatterGen, and has scaled autonomous physics labs while contributing to major materials discoveries, Fedus pointed out, adding they’re uniting to reimagine and scale how science is done by pairing AI systems with automated experimentation.
The thesis is straightforward: the internet’s text is finite and largely mined; the next step is generating proprietary, high-quality experimental data at scale. By coupling AI reasoning with automated lab systems, Periodic aims to create a feedback loop—conjecture, test, learn—that compresses R&D cycles and broadens the search for useful materials and processes.
“Scientific discovery is inherently an out-of-domain task. Experimental iteration is required for significant advances, regardless of the form of intelligence that is modeling the world,” Cubuk wrote on X. “We are building experimental labs that will unlock the next frontier for LLM reasoning.”
