Isomorphic Labs
Alphabet's AI-first drug-discovery subsidiary — the benchmark "AI + science experts" model, applying DeepMind's AlphaFold lineage to wet-lab drug design.
1. Core Product / Service
Isomorphic Labs is an AI-driven drug-discovery company spun out of google-deepmind and controlled by Alphabet. It applies deep learning — anchored on AlphaFold-class structure prediction (alphafold3) — to design drug candidates end-to-end: target identification, hit discovery, lead optimization, and preclinical validation. Its founding thesis is "digital biology": treat biology as an information-processing system that ML can model at scale, rather than using AI merely as a screening aid (Internal research notes).
2. Target Users & Pain Points
- Pharma partners via multi-year discovery collaborations that aim to compress early R&D timelines and raise clinical success rates — preclinical failure is the industry's dominant cost (see ai-for-science).
- Pain solved: traditional drug discovery is a low-throughput, high-failure trial-and-error process; Isomorphic positions AI to move the failure point earlier and cheaper.
3. Competitive Landscape
| Player | Positioning | Vs. Isomorphic |
|---|---|---|
| google-deepmind / alphafold3 | Open structure-prediction research | Isomorphic is the commercial wet-lab arm applying that research |
| insilico-medicine | Generative-AI-first biotech (Pharma.AI) | Independent biotech running its own clinical programs |
| xtalpi | Crystal-structure prediction + AI drug discovery (China) | China-listed, physics-first CSP origin |
| Recursion / Schrödinger | Computational + wet-lab drug discovery | Pre-AI-era incumbents adding ML |
4. Unique Observations
- The "AI + science experts" playbook that Chinese AI4S teams (incl. kimi) have benchmarked against — general-model scale paired with domain experts, versus a pure-science startup building a small bespoke model (Internal research notes).
- Structural moat: inside Alphabet and inheriting google-deepmind's research, Isomorphic can fund long-horizon R&D with patient capital — a structural advantage over independent AI4S biotechs.
- Shared model lineage: AlphaFold 3 was developed jointly by google-deepmind and Isomorphic Labs — the clearest example of the "research lab → commercial subsidiary" transfer in AI4S.
5. Financials / Funding
- External rounds: Series A $600M (2025-03, led by Thrive) and Series B $2.1B (2026-05-12, led by Thrive; Alphabet, GV, MGX, Temasek, CapitalG and the UK Sovereign AI Fund participating) — about $2.7B raised in total; valuation undisclosed. Alphabet remains the controlling shareholder.
- Partnership economics (reported): multi-year discovery collaborations with Eli Lilly and Novartis announced January 2024, each reported up to ~$3B including milestones (upfronts in the tens of millions USD). Headline figures from public reporting — verify before relying (Internal research notes).
- Status: pre-revenue / pre-commercial as of 2026.
6. People & Relationships
- Founder / CEO: Demis Hassabis (also co-founder/CEO of google-deepmind).
- Parent: Alphabet.
- Partners: Eli Lilly, Novartis.
- Related: alphafold3, google-deepmind, ai-for-science.
Isomorphic Labs
DeepMind 药物发现分拆公司,闭源 IsoDDE
做什么
Google DeepMind 的药物发现分拆公司,与 DeepMind 共同署名 alphafold3 论文 [3];2026-02 发布闭源药物设计引擎 isodde [1]。
数据策略
IsoDDE 的训练数据未披露;AF3 论文方法部分无 Isomorphic 内部专有数据的披露。
融资与客户
B 轮 $21 亿,2026-05-12,Thrive 领投,称临床前已开始、人体试验计划 2026 年底;客户含 Novartis、Lilly、J&J [4]。媒体报道估值 $150–200 亿 ⚠️ 待复核。
争议
管线口径:ISM8969 属 Insilico,不属 Isomorphic,不可写入其管线。IsoDDE 在蛋白-配体泛化基准上"准确率超 AF3 两倍"为公司口径,无独立复现 [1][2]。
- https://isomorphiclabs.com/articles/the-isomorphic-labs-drug-design-engine-unlocks-a-new-frontier (2026-10-15)
- https://www.scientificamerican.com/article/an-alphafold-4-scientists-marvel-at-deepmind-drug-spin-offs-exclusive-new-ai/ (2026-10-15)
- https://pmc.ncbi.nlm.nih.gov/articles/PMC11168924/ (2026-10-15)
- https://www.isomorphiclabs.com/press/isomorphic-labs-funding (2026-10-15)