Lila Sciences: The Laboratory Where AI Does the Experimenting

Most AI companies automate cognitive work — writing, coding, analysis. Lila Sciences automates something much harder to digitize: the physical process of scientific experimentation itself. It's arguably the most radical bet among the 2025 unicorn class, betting that AI can run an entire research and development loop — hypothesis, experiment, result, iteration — with minimal human involvement in the routine steps.

What an "AI Factory of Science" Actually Means

Lila Sciences builds physical laboratories where robotic equipment, directed by AI systems, operates around the clock. The loop works like this: the system generates a scientific hypothesis, designs an experiment to test it, physically executes that experiment through robotic hardware, and then feeds the results back into the model to inform the next hypothesis. It's a closed loop that, in theory, never needs to stop for a human researcher to go home for the night — the robots keep running experiments through the same hours a human-staffed lab would sit idle.

This isn't automation of a single repetitive task, the way a pipetting robot in a traditional lab automates one step. It's an attempt to automate the entire cycle of scientific discovery, including the intellectually creative parts — hypothesis generation and experimental design — that have traditionally been considered uniquely human contributions to research.

Why George Church Lends the Company Instant Credibility

Lila Sciences' chief scientist is George Church, a Harvard professor and one of the most cited geneticists in the world. In a field where credibility with the scientific community is a genuine barrier to adoption — pharma and biotech companies won't hand real R&D budget to a black-box system without confidence in its scientific rigor — having a researcher of Church's stature involved provides a level of institutional trust that a purely commercial AI team couldn't generate on its own.

Three Ways to Monetize an Automated Lab

Lila Sciences has structured its business around three distinct revenue tracks: project-based partnerships with pharmaceutical and chemical companies, subscription-based rental of lab capacity through a Lab-as-a-Service model, and licensing of its own developed technologies to partners. That diversification matters for a capital-intensive business model — physical lab infrastructure is expensive to build and maintain, and a single revenue stream would leave the company overly exposed to the sales cycle of any one type of customer.

Fast Capital, Fast Ambition

In September 2025, Lila Sciences raised a $235 million Series A at a $1.2 billion valuation. That round came on top of an already substantial fundraising pace — the company had raised nearly $450 million total in under a year of existence, reflecting investor appetite for a company attempting something genuinely novel rather than an incremental improvement on existing lab automation tools.

The Long-Term Stakes

If Lila Sciences' model works at scale, the implications extend well beyond one company's valuation — it would represent a fundamentally different economics for scientific discovery, where the marginal cost of testing another hypothesis drops toward the cost of robot time and compute, rather than researcher salaries and lab scheduling constraints. That's a big claim, and the jury is still out on how far it generalizes beyond Lila's initial focus areas. But the scale of capital and scientific credibility behind the bet suggests serious people think it's worth testing.

This batch covers three more of the companies profiled in the CrackTheDeck "AI Unicorns 2025" report. The next batch will cover the "picks and shovels" economics of AI training — Mercor, Clay, and Suno.

FAQ

Source: CrackTheDeck Research.