Lila Sciences
Last reviewed
Jun 7, 2026
Sources
18 citations
Review status
Source-backed
Revision
v1 · 1,732 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Last reviewed
Jun 7, 2026
Sources
18 citations
Review status
Source-backed
Revision
v1 · 1,732 words
Add missing citations, update stale details, or suggest a clearer explanation.
Lila Sciences is a Cambridge, Massachusetts technology company that builds artificial intelligence systems coupled with automated laboratories, which it calls "AI Science Factories," in pursuit of what it brands "scientific superintelligence" across chemistry, materials science, and the life sciences. Incubated inside the venture-creation firm Flagship Pioneering and led by chief executive Geoffrey von Maltzahn, the company emerged from stealth on March 10, 2025 with a $200 million seed round, and by October 2025 had raised roughly $550 million in total, one of the largest early-stage funding totals in the history of biotechnology. Lila's stated approach closes a loop in which AI agents propose hypotheses and experiments, robotic laboratories run them, and the physical results feed back to retrain the models. As of late 2025 the company remained early stage, and its headline scientific results were company claims that had not been independently validated or published in peer-reviewed venues.
Lila Sciences was created within the labs of Flagship Pioneering, the Cambridge firm founded by Noubar Afeyan that originated Moderna and Generate:Biomedicines. Flagship and the company describe Lila as having formed in 2023, drawing together internal programs in biology and materials science; some company-background reporting dates the underlying projects to 2022. The venture stayed in stealth until it was publicly unveiled on March 10, 2025.
Leadership combines Flagship operators with established academic scientists. Geoffrey von Maltzahn, a Flagship general partner who co-founded multiple Flagship companies, is co-founder and chief executive. Noubar Afeyan serves as co-founder and chairman. Andrew Beam, a former Harvard Medical School faculty member and a co-founder of Generate:Biomedicines, is chief technology officer, and the Harvard geneticist George Church, known for foundational work in genome sequencing and synthetic biology, joined as chief scientist; both Beam and Church were named with the March 2025 launch. The AI research group has been led in part by Kenneth Stanley, a machine-learning researcher who previously worked at OpenAI. On the materials side, MIT professor Rafael Gomez-Bombarelli is named as a co-founder and senior materials scientist, and John Gregoire, an experimental materials researcher, serves as chief autonomous science officer.
Lila frames its goal as building AI that can execute "every step of the scientific method," from generating a hypothesis to designing an experiment, running it physically, interpreting the data, and iterating. The company calls the eventual product "scientific superintelligence," language deliberately echoing the artificial-general-intelligence ambitions of frontier AI labs but applied to laboratory discovery rather than general reasoning.
This branding should be read as the company's own marketing rather than a measured external assessment. "Superintelligence" here is an aspirational label for a system that the company says exceeds human and prior-AI benchmarks on specific tasks; it does not denote any independently verified general capability. Reporting on the broader field has stressed that no "ChatGPT moment" has yet arrived for autonomous scientific discovery, and that the hardest and most expensive steps in domains such as materials science are physical synthesis and testing, not computation. As John Gregoire put it to MIT Technology Review, simulations are useful for framing what is worth testing, "but there's zero problems we can ever solve in the real world with simulation alone."
Lila's central concept is the "AI Science Factory," which the company describes as a unified facility where AI models, custom software, and laboratory hardware "close the loop between reasoning and real-world verification." In Lila's framing these are "scientific-method machines" that use generative AI and reinforcement learning to propose experiments, then carry them out with robotics and automated instrumentation, capturing physical data that is fed back to improve the models. The company speaks of producing "scientific tokens" by putting more instruments under AI control, an analogy to the way large language models consume text tokens.
As of December 2025 the company operated a single AI Science Factory in development in Cambridge, with stated plans to expand to additional sites including San Francisco and London. Reporting on the Cambridge facility describes AI-directed equipment such as sputtering systems producing thin-film alloy samples for materials work. Lila has signed a 235,500-square-foot lease at Alewife Park in Cambridge, reported as the largest Greater Boston lab lease of the third quarter of 2025. The company has emphasized that human scientists are still required to supervise the machinery, which tempers the "fully autonomous" framing. Headcount grew from roughly 100 employees earlier in 2025 toward several hundred by year end, with one company-background report citing about 249 employees in December 2025.
Lila has raised capital quickly and from an unusually broad investor base spanning venture firms, sovereign wealth, defense-linked funds, and a major chip maker. The seed round of $200 million was announced at launch in March 2025. A Series A first close of $235 million followed in September 2025, and a $115 million extension in October 2025 brought the Series A to $350 million and cumulative funding to roughly $550 million. Some company-background reporting cites a valuation of about $1.2 billion at the September Series A and above $1.3 billion after the October extension; the company itself did not confirm a valuation in its announcements, so these figures should be treated as unconfirmed.
| Round | Date (announced) | Amount | Selected investors |
|---|---|---|---|
| Seed | March 10, 2025 | $200 million | Flagship Pioneering (lead), General Catalyst, March Capital, ARK Venture Fund, Altitude Life Science Ventures, State of Michigan Retirement System, Modi Ventures, a subsidiary of the Abu Dhabi Investment Authority |
| Series A (first close) | September 15, 2025 | $235 million | Co-led by Braidwell and Collective Global |
| Series A (extension) | October 2025 | $115 million | NVIDIA's NVentures, Analog Devices, IQT, Dauntless Ventures, Catalio Capital Management, Pennant Investors, members of Peter Diamandis' Abundance group |
| Total | through October 2025 | about $550 million | Seed plus $350 million Series A |
The participation of NVIDIA's corporate venture arm, NVentures, in the October 2025 extension and of IQT (the venture arm associated with the U.S. intelligence community) drew particular attention, signaling interest in the company's compute-heavy, security-sensitive approach.
Lila has publicized four categories of scientific results, all of which are company claims that, as of late 2025, had not been independently verified or published in peer-reviewed journals. They should be attributed to the company.
Outside experts urge caution. Coverage of AI-driven materials science has noted that AI predictions frequently fail to translate into materials that can actually be synthesized, remain stable at real-world temperatures, and exhibit useful function, and that a separate effort by Google DeepMind to predict millions of new materials was later criticized for thin evidence of compounds meeting the bar of novelty, credibility, and utility at once. Industry analysts have also flagged persistent challenges in biological data quality and in connecting diverse laboratory instruments into a single automated pipeline. None of this disproves Lila's specific claims, but it underscores that the company's results await independent confirmation.
Lila Sciences sits at the center of a wave of well-funded startups applying AI for science and autonomous laboratories to discovery, alongside peers such as Periodic Labs. Its scale of funding, its Flagship pedigree, and the involvement of figures like George Church and an ex-OpenAI researcher made it one of the most visible bets that the generative AI wave can be turned toward chemistry, materials, and medicine rather than only language and images. The "scientific superintelligence" framing is ambitious and unproven, and the company's most striking claims remain unvalidated by independent parties. Whether Lila can convert large capital and a closed-loop lab architecture into reproducible, externally verified discoveries, rather than benchmark results announced in blog posts, is the central open question about the company as it moves from stealth to commercial deployment.