Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesSome links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Periodic Labs announced a $300 million seed round on September 30, 2025, as it emerged from stealth to build AI systems linked to robotic laboratories. The company says it will use models, simulations and physical experiments to pursue advances in materials science, starting with superconductors. The funding is a major bet on that approach—not evidence that Periodic has already made a scientific breakthrough or brought a new material to market.
The $300 million round
Periodic Labs calls the financing a “founding round”; the company’s launch coverage commonly describes it as a seed round. Andreessen Horowitz (a16z) led the investment. Other publicly named backers include Felicis, DST Global, NVentures—the venture arm of NVIDIA—and Accel, along with Jeff Bezos, Elad Gil, Eric Schmidt and Jeff Dean. The public announcements do not disclose how much each investor contributed or identify individual investment vehicles. Periodic’s launch announcement and a16z’s announcement provide the investor list and round framing.
The company was founded by Liam Fedus, a former OpenAI research leader, and Ekin Dogus Çubuk, a former Google Brain and Google DeepMind researcher whose work includes materials science. Their prior experience helps explain why the venture attracted attention, but it does not establish what Periodic itself has achieved. TechCrunch reported that Çubuk was associated with Google’s GNoME research, which identified more than two million candidate crystal structures computationally. Those were candidates, not two million experimentally confirmed, commercially usable materials. TechCrunch’s launch report gives further context on the team and GNoME.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →What Periodic is building
Periodic’s ambition is not simply a chatbot that answers questions about scientific papers. It says it is building a closed loop connecting AI models to simulations and physical experiments, with robotic equipment carrying out some laboratory work. The intended cycle is:
#1 Best Overall
- Review evidence: Use scientific literature and existing experimental records to frame a research question.
- Choose a hypothesis: Generate candidate materials or properties to investigate.
- Rank experiments: Use models and computational tools to select promising tests.
- Run physical tests: Direct laboratory equipment to synthesize or measure candidates.
- Feed back results: Compare measurements with predictions, update the models and choose what to test next.
In principle, this can turn experiments into data the system can use, including results that are not already in published papers. But this is Periodic’s intended workflow, not proof that it has built a generally autonomous scientist. Automation does not remove the hard parts of research: designing meaningful tests, preparing samples, calibrating instruments, controlling contamination and deciding whether an apparent result is reproducible.
Why start with materials and superconductors?
Periodic says it is beginning in the physical sciences, with an early focus on discovering superconducting materials that work at higher temperatures than existing options. It also describes longer-term ambitions in areas such as semiconductors, advanced manufacturing, energy and aerospace. Those wider applications are goals, not evidence of deployed products.
Superconductors can carry electrical current with extremely low resistance under suitable conditions. Materials that operate at higher temperatures—or are otherwise easier and less costly to use—could potentially reduce cooling needs or improve equipment involving power, magnets, computing, transportation or medical and industrial systems. Yet finding a promising candidate is only an early milestone. A material must also be reproducibly made, stable, affordable, scalable and compatible with real operating conditions before it can become a useful product.
Rank #2
Materials science is not an easy shortcut to scientific discovery. Synthesis can fail; measurements can be noisy; simulations may not match laboratory behavior; and a result that works in one setup may not reproduce elsewhere. A model can also optimize a measurable proxy without finding a material that solves a practical problem.
Why raise so much at seed stage?
A company combining frontier AI and laboratory research has substantial costs that a software-only startup may not face. The money could support recruiting specialists in AI, physics, chemistry and robotics; building or equipping labs; acquiring compute; developing automated synthesis and measurement systems; and running experiments over a long research period. These are reasonable implications of Periodic’s stated strategy, not a published allocation of proceeds: the company and a16z have not provided a detailed spending breakdown.
The investment thesis is that models trained on published text alone have limits. Scientific literature can be incomplete, measurements vary, and negative results often do not appear in papers. Physical experiments may provide new, proprietary evidence and let a system test its predictions against the world. That is the rationale Periodic and its investors advance, not a settled conclusion that automated labs will make research faster, cheaper or more reliable in every field. Experiments, instruments, facilities and expert oversight remain expensive, and more data is not necessarily better data.
The scale of the round gives Periodic room to build infrastructure before proving a commercial model. It also raises expectations: investors will eventually need evidence that the system produces repeatable scientific results or useful customer outcomes, not only impressive demonstrations or large candidate lists.
Recommended Free Tools
What is known about commercial work—and what is not
Periodic and a16z have said the company is working with industry, including on a semiconductor manufacturer’s chip heat-dissipation problem. The customer has not been named, and public materials do not disclose contract value, revenue, measured outcomes or whether the work has produced a commercial product. The announcements also do not provide customer counts, pricing, laboratory throughput or a tally of validated discoveries.
Those gaps matter when judging the round. A large financing and prominent backers signal investor confidence and provide resources; they do not independently verify customer traction or scientific performance. For the approach to succeed, Periodic will have to show that its experiments are well calibrated and reproducible, that its models learn scientific relationships rather than lab-specific artifacts, and that promising results can be validated outside its own facility. Ultimately, a discovery must be manufacturable and valuable enough to justify its costs.
Rank #4
Is OpenAI an investor?
No. TechCrunch reported in October 2025 that Periodic’s founders said OpenAI was not a backer. Fedus’s former employer should not be confused with the investors publicly listed for the round. The same report said Periodic had hired more than two dozen researchers and had established a laboratory by October 20, 2025; those are dated snapshots, not current headcount or a measure of the lab’s present capabilities. TechCrunch’s follow-up covered those details.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.A later financing report is not a closed round
On May 7, 2026, Forbes reported that Periodic was in advanced talks to raise at least another $500 million at a $7.5 billion valuation. The report described negotiations, not a completed financing confirmed by Periodic. It also cited a reported launch valuation of about $1.3 billion, a figure not clearly disclosed in the company’s original announcement. Treat those figures as reported terms, not established company facts. Forbes’s report is the source for the update.
How to assess the bet
The case for Periodic is that automated experiments could generate useful data unavailable in public literature, while a feedback loop could help models choose better tests. A well-equipped lab may also be difficult for a conventional software company to replicate. But each link in that chain must work: the experiment must be correctly designed and executed, the measurements trustworthy, the model’s learning generalizable and the result useful beyond the lab.
Potential failure points include sensors that are miscalibrated, sample contamination, models learning laboratory quirks, results that cannot be reproduced elsewhere, or a candidate that proves impossible to manufacture economically. Commercialization can take years after a discovery, with additional validation, scale-up and customer testing. No public announcement establishes that Periodic has solved these problems or announced a breakthrough.
The most meaningful evidence to watch for is therefore not simply the number of AI-generated candidates. It is independently reproducible measurements, validated material properties, results that survive scale-up, and disclosed customer outcomes that show the system improves on conventional research in cost, time or performance.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
