AI-Bio (TechBio / AI Drug Discovery / Synthetic Biology)
AI-driven drug discovery, generative biology, protein foundation models, synthetic biology, autonomous labs, and the techbio thesis
AI-Bio (TechBio / AI Drug Discovery / Synthetic Biology)
AI-Bio is the convergence of machine learning with biological discovery — protein design, generative chemistry, autonomous wet labs, and synthetic biology. The field is in its most critical phase: moving from platform promises to clinical outcomes. As of mid-2026, over 173 AI-originated drug programs are in clinical development (up from ~24 in late 2023), but no AI-designed molecule has yet achieved full FDA approval. That milestone is projected for 2026–2027, with Insilico Medicine's Rentosertib (TNIK inhibitor for IPF) as the most advanced candidate after Phase IIa results published in Nature Medicine in June 2025.
Three structural bets define the sector. First, protein and molecule foundation models (ESM3, AlphaFold3, BioNeMo, RFdiffusion) compress sequence-structure-function reasoning into a single model, enabling de novo protein design at scale. Second, lab-in-the-loop closes the design-test-learn cycle by integrating AI design with robotic wet labs — moving from weeks to hours per iteration. Third, the data moat is the durable advantage: Recursion's 2.5 trillion cellular images, Tempus's multimodal oncology dataset, and Basecamp Research's biodiversity data are harder to replicate than any algorithm.
The field splits cleanly into two archetypes. Platform companies (Recursion, Schrodinger, AbCellera, Ginkgo) derive value from the engine itself — licensing access, generating partnerships, and slowly building their own asset pipeline. Asset companies (Insilico, Iambic, Genesis) bet that a specific AI-discovered molecule will clear Phase 2/3. The market is still trying to price which archetype wins, while pharma is hedging by doing both (Isomorphic Labs, DeepMind/Google's drug spinout, owns the pipeline and is negotiating like an asset play).
The cautionary sub-ledger is critical context: BenevolentAI's lead failed efficacy in 2023, Exscientia was acquired in distress by Recursion in 2024, Zymergen was liquidated, Amyris filed Chapter 11, and several post-SPAC synbio names (Ginkgo -95%, Senti, Lanzatech) trade near zero. The hype cycle from 2020–2022 is correcting in public markets; private markets remain elevated via Isomorphic ($2.1B Series B, 2026), Xaira ($1B launch), and Lila Sciences ($550M autonomous lab).
Frontier — What's Moving Now
- First AI-designed drug to Phase 3 — Insilico's Rentosertib approaching Phase 3 for IPF; would be history's first AI-to-Phase-3 molecule. Phase 3 design expected 2026.
- Isomorphic Labs first trials — DeepMind spinout targeting first internal IND by end of 2026; $2.1B Series B (Jan 2026); the sector's most watched catalyst.
- Lab-in-the-loop hitting commercial scale — AWS Amazon Bio Discovery (Apr 2026), Insilico's LabClaw, Lila Sciences AI Science Factories; the automated loop is now a product category.
- Generate Biomedicines IPO — $400M raise at $16/share (Feb 2026, GENB); first major AI-biologics public offering of the cycle.
- Data-moat consolidation — Tempus AI's 126% NRR (2025) shows data monetization working; Recursion CEO transition (Najat Khan, Jan 2026) signals scale-up phase.
- ESM3 / protein foundation model convergence — EvolutionaryScale's ESM3 (98B params, 2.78B proteins) enables generative reasoning over sequence+structure+function simultaneously; a16z/Nvidia-backed.
Concept Map
Concepts
| Concept | Status | Last Updated |
|---|---|---|
| Protein & Molecule Foundation Models | Frontier — Active | 2026-06-15 |
| Generative Biology | Frontier — Active | 2026-06-15 |
| Lab-in-the-Loop | Frontier — Rapid Progress | 2026-06-15 |
| In-Silico Screening | Established — Active | 2026-06-15 |
| The Data Moat Question | Active Debate | 2026-06-15 |
| Platform vs Asset Archetype | Active Debate | 2026-06-15 |
| The TechBio Thesis | Core Concept | 2026-06-15 |
| Synthetic Biology | Active | 2026-06-15 |
Entities
| Entity | Type | Status | Key Fact |
|---|---|---|---|
| Recursion Pharmaceuticals | Company (public, RXRX) | Operational | >2.5T cellular images; new CEO Jan 2026 |
| Isomorphic Labs | Company (private) | High conviction | $2.1B Series B Jan 2026; DeepMind spinout; IND target end-2026 |
| Insilico Medicine | Company (public, 2546.HK) | Most advanced pipeline | Rentosertib Phase IIa (Nature Medicine Jun 2025) |
| EvolutionaryScale | Company (private) | Foundation model leader | ESM3 98B param model; Nvidia/a16z-backed |
| Generate Biomedicines | Company (public, GENB) | Newly public | $400M IPO Feb 2026; Novartis deal |
| Xaira Therapeutics | Company (private) | Well-funded stealth | $1B launch; David Baker / RFdiffusion lineage |
| Tempus AI | Company (public, TEM) | Data monetization | Largest US multimodal oncology dataset; 126% NRR |
| Lila Sciences | Company (private) | Autonomous lab | $550M raised; AI Science Factories; Nvidia-backed |
| Ginkgo Bioworks | Company (public, DNA) | Distressed synbio | Cell-programming platform; post-SPAC -95%; Bayer partnership May 2025 |
| Chai Discovery | Company (private) | Structural prediction | Chai-1/Chai-2; Pfizer license |
| Schrodinger | Company (public, SDGR) | Computational chem | Physics-based simulation + own drug pipeline |
| NVIDIA BioNeMo | Platform | Infrastructure | Cloud bio foundation models; compute backbone for the sector |
Timeline
See timeline.md for chronological developments.
Research Frontier
See frontier.md for active research directions, 2026 catalysts, and knowledge gaps.
Sources
Initial wiki built from web research (Jun 2026). Primary sources ingested + compiled 2026-06-24.
| Source | Type | Date | Informs | Status |
|---|---|---|---|---|
| UniGenX: Unified Generative Foundation Model | paper (arXiv:2503.06687) | 2025-08-26 | protein-foundation-models, generative-biology, in-silico-screening | compiled |
| Survey of ESM Protein Language Models | paper (DOI 10.1002/qub2.70013) | 2025-09-21 | protein-foundation-models, in-silico-screening, evolutionaryscale | compiled |
| Biohub Protein-Biology World Model | technical-report (GEN) | 2026-05-27 | generative-biology, protein-foundation-models, data-moat, evolutionaryscale | compiled |
| Isomorphic Labs — IsoDDE + Human Trials | technical-report (summary-derived) | 2026-04-25 | in-silico-screening, platform-vs-asset, isomorphic-labs, recursion | compiled |