jimmyresearch
Home/AI for Science/Lila Sciences
Company

Lila Sciences

Flagship-incubated "scientific superintelligence" startup running autonomous AI labs (AI Science Factories) across drug discovery, materials, and energy.

1. Core Product / Service

Lila Sciences (Cambridge, MA) is building what it calls a "scientific superintelligence" — an AI system that runs the scientific method end-to-end, from hypothesis generation through experiment design, robotic execution, and interpretation of results [2]. Founded in 2023 and launched from stealth in March 2025 inside Flagship Pioneering (the venture studio behind Moderna), Lila pairs an AI reasoning model — Lila Iris — with AI Science Factories: robotic labs that physically run experiments and feed real-world data back into the model for the next hypothesis cycle [2].

The company frames its output as "scientific tokens" — analogous to how LLMs consume text tokens — produced by putting more instruments under AI control [2]. Its commercialization plan splits into two routes: Catalyst (lab-as-a-service for customers' own scientists) and Creation (end-to-end discovery campaigns that return validated molecules, materials, and IP to partners) [2]. Target domains span drug discovery, chemicals, advanced materials, energy, and defense [2].

2. Target Users & Pain Points

  • Pharma / biotech and industrial partners wanting validated molecules or materials without building their own autonomous-lab stack — served via Creation campaigns.
  • Research teams wanting lab capacity plus an AI co-pilot — served via Catalyst.

Pain solved: the traditional discovery loop is bottlenecked by human wet-lab throughput and fragmented point AI tools. Lila's bet is that a closed loop — AI proposes, robots test, data retrains the model — collapses the hypothesis-to-result cycle time. It acknowledges that human scientists are still required to supervise the machinery, which tempers the "fully autonomous" framing [2].

3. Competitive Landscape

Player Approach Vs. Lila Sciences
recursion-pharmaceuticals Industrialized phenomics + public-company data flywheel Data/imaging-first and public; Lila is lab-loop-first and private
isomorphic-labs DeepMind spinout applying AlphaFold-lineage models to drug discovery Pharma-focused; Lila spans chemistry/materials beyond pharma
generate-biomedicines Flagship sibling using generative protein design Single-domain (proteins); Lila is a general multi-domain autonomous-lab play

Lila's differentiation is the physical closed loop — owning the robotic labs, not just a model — which makes it harder to falsify but also capital-intensive relative to software-only competitors [2].

4. Unique Observations

  • The valuation jump is the story. Lila went from ~$1.3B (Oct 2025 Series A) to an ~$8.5B valuation (2026 Series B) in roughly eight months — with no clinical asset and no disclosed commercial revenue [2][4][5]. Jimmy's local research flags the ~$8.5B as a record for the AI-for-science vertical [1]. Analysts describe it as the highest-valued and "hardest-to-falsify" name in the AI4S wave — a price that rests substantially on the Flagship Pioneering brand premium [2].
  • General-purpose closed loop vs. domain point solutions. Lila's ambition is one machine that does science generically, not a single assay or model. That is a direct bet against the field's historical pattern documented in ai-for-science, where domain-specific task models still dominate; Lila is the cleanest test yet of the "just build the general loop" thesis.
  • Flagship's second-generation platform play. CEO Geoffrey von Maltzahn previously founded generate-biomedicines inside the same studio, and Noubar Afeyan chairs both. Lila is Flagship testing whether the Moderna playbook — incubate a platform company inside the studio, fund it hard, scale — transfers to autonomous science [2][6].

5. Financials / Funding

  • Seed (Mar 2025): $200M, Flagship-led [2].
  • Series A (Sep–Oct 2025): $235M first close (co-led Braidwell + Collective Global) plus a $115M extension (Nvidia NVentures, Analog Devices, IQT) = $350M total, at a valuation above $1.3B [2][3].
  • Total through Oct 2025: ~$550M [2].
  • Series B (2026): Bloomberg reported talks to raise ~$2B at an ~$8.5B pre-money valuation (Jun 2026) [4]; StockAnalysis lists an $8.5B valuation dated Jul 31, 2026 with $2.55B total funding, implying the round closed — though one investor profile noted it was "not yet closed" as of Jul 2026, so close timing is ambiguous across sources [5].

The ~$8.5B figure is the number Jimmy's 2026-09-05 research surfaced as an AI4S funding high-water mark [1].

6. People & Relationships

  • Co-founder / CEO: Geoffrey von Maltzahn — Flagship general partner; founding CEO of generate-biomedicines, Tessera Therapeutics, Quotient Therapeutics, Indigo Ag [2][6].
  • Co-founder / Chairman: Noubar Afeyan — Flagship founder/CEO and Moderna co-founder [2].
  • Other co-founders: Alexandra Sneider, Molly Gibson, Rafael Gómez-Bombarelli, Ben Kompa [6].
  • Key executives: CTO Andrew Beam; SVP of Scientific Intelligence Kenneth Stanley (ex-OpenAI); Chief Scientist George Church [2].
  • Investors: Flagship Pioneering (incubator), Nvidia (NVentures), Braidwell, Collective Global, Analog Devices, IQT [2][3].
  • Facility / scale: 235,500 sq ft Alewife Park (Cambridge) lease; ~250 employees by Dec 2025; planned San Francisco and London expansion [2].

Sources

Related

generate-biomedicinesrecursion-pharmaceuticalsisomorphic-labs
Last compiled: 2026-09-07