Recursion Pharmaceuticals (RXRX)

Type: Platform company (phenomics + AI drug discovery) — public, NASDAQ: RXRX

Recursion built the largest proprietary phenomics dataset in drug discovery: over 2.5 trillion cellular images across hundreds of disease-relevant contexts and compound perturbations, generated by a highly automated robotic biology infrastructure. Its core model maps chemical perturbations to cellular phenotypes — essentially asking "what does this molecule do to cells across thousands of disease contexts simultaneously?" — and uses that map to identify novel biology and drug candidates.

In November 2024, Recursion completed the acquisition of Exscientia (a competing UK-based AI drug discovery platform) in a $688M merger, creating a combined company with end-to-end AI capabilities (phenomics mapping + small-molecule design). The integration was financially strained — Exscientia's pre-acquisition valuation was significantly higher.

In January 2026, Najat Khan became CEO, replacing founder Chris Gibson. Khan had been CSO; the transition signals a shift from founder-led startup to operational scale-up.

Pipeline (as of mid-2026)

• REC-1245 (RBM39 degrader) — Phase 1, solid tumors / lymphoma; dose-escalation data updates expected H1 2026
• REC-4209 (for IPF) — IND-enabling studies ongoing
• 10+ clinical and preclinical programs in combined pipeline; $20B+ in possible partner milestones

Financials (key metrics)

• Revenue: ~$85M projected 2026 (milestones + partner revenue)
• Data asset: >2.5T cellular images (10-100x any academic equivalent)
• Partners: NVIDIA, Sanofi, Bayer, Roche/Genentech among others
• Total milestones possible from partners: >$20B

Key Relationships

• NVIDIA: Recursion runs its models on BioNeMo; NVIDIA is a key infrastructure partner
• Sanofi / Roche / Bayer: Research collaboration partners
• Exscientia (now wholly owned): Adds small-molecule AI design capability

Cautionary note

RXRX has underperformed significantly from its 2021 peak. The platform thesis requires pharma partners to convert milestone payments, which is slow. The Exscientia merger added capabilities but also complexity and cost.

Key Claims

• Recursion's Boltz-2 (with MIT) approaches FEP accuracy ~1,000× faster — Per the IsoDDE-context report, Boltz-2 reaches free-energy-perturbation-grade binding accuracy at roughly 1,000× the speed — a concrete data point on the ML-surrogate-for-physics trajectory in in-silico screening. Evidence: moderate (Isomorphic IsoDDE context)
• >$1B spent over a decade without a commercial drug — The same report cites Recursion as the cautionary anchor for the unproven platform→asset bridge: heavy spend, real tooling, no approval yet. Evidence: moderate (Isomorphic IsoDDE context)

Sources

• Recursion IR — public filings
• FierceBiotech: Exscientia merger pipeline rework
• Recursion open datasets
• Isomorphic IsoDDE + human trials (2026) — Boltz-2 + spend context

Changelog

• 2026-06-24 — Compiled new sources (isomorphic-isodde-human-trials)