Recursion Pharmaceuticals, Inc. (NASDAQ: RXRX) — Business Overview

Executive Summary

Recursion Pharmaceuticals is a Salt Lake City-based, clinical-stage "TechBio" company that fuses high-throughput, automated wet-lab biology with machine learning to industrialize drug discovery. Its Recursion Operating System ("Recursion OS") runs more than two million experiments per week to generate proprietary phenomic, chemical, transcriptomic, proteomic, genomic and real-world clinical datasets, which it mines with purpose-built foundation models (Phenom, MolPhenix, MolGPS) and the NVIDIA-built BioHive-2 supercomputer to nominate and advance drug candidates. Following its 2024 all-stock combination with Exscientia — the largest merger to date in the TechBio sector — Recursion added a generative chemistry engine (Centaur Chemist) and an AI-enabled clinical operations system (ClinTech), giving it an integrated discovery-to-clinic stack. The company monetizes its platform two ways: (1) a self-funded internal oncology-focused pipeline of seven clinical and preclinical small-molecule candidates, led by REC-4881 (FAP), REC-617 (solid tumors) and REC-1245 (solid tumors/lymphoma); and (2) multi-year platform collaborations with Roche/Genentech, Sanofi, Bayer and Merck KGaA that have generated over $500 million in cumulative upfront and milestone payments and carry up to $20 billion of contingent future milestones before royalties. For fiscal year 2025, Recursion reported $74.7 million of collaboration revenue against a $644.8 million net loss, ending the year with $753.9 million in cash, cash equivalents and restricted cash and a stated cash runway extending into early 2028. The investment case rests on whether the platform's data and AI advantages translate into materially better clinical success rates than conventional R&D before the company's cash runway — and investor patience — runs out; recent Phase 2 data for REC-4881 and continued large-pharma renewals offer early, but not yet conclusive, validation.

---

1. Core Business Model

Recursion operates a closed-loop "biology-to-bedside" flywheel: automated experimentation generates proprietary data, machine learning turns that data into candidate hypotheses, and both an internal pipeline and external pharma partners convert those hypotheses into drug programs. Data generated across every program — whether internal or partnered — continuously retrains the underlying foundation models, so each additional experiment or partnership is intended to make the next candidate search more efficient (a data-network-effect dynamic rather than a classical economies-of-scale one).

 ┌────────────────────┐     ┌───────────────────┐     ┌────────────────────────┐
 │  AUTOMATED WET LAB  │     │   RECURSION OS      │     │   CANDIDATE DISCOVERY   │
 │  2M+ experiments/wk │ --> │  Phenom / MolPhenix  │ --> │  Hit -> Lead -> IND     │
 │  ~65 PB phenomics,   │     │  MolGPS foundation   │     │  candidate (Centaur     │
 │  chem, -omics data   │     │  models + BioHive-2  │     │  Chemist generative     │
 │                      │     │  (NVIDIA) compute     │     │  chemistry)             │
 └─────────┬────────────┘     └─────────┬───────────┘     └────────────┬────────────┘
           │                              │                              │
           │                              ▼                              │
           │                  ┌───────────────────────┐                  │
           └────────────────> │  Patient / RWD overlay │ <────────────────┘
                              │  (Tempus, Helix) +      │
                              │  ClinTech trial design  │
                              └────────────┬────────────┘
                                           │
                     ┌─────────────────────┴─────────────────────┐
                     ▼                                             ▼
         ┌─────────────────────────┐                 ┌─────────────────────────────┐
         │   INTERNAL PIPELINE      │                 │   PHARMA PARTNERSHIPS         │
         │   REC-4881, REC-617,     │                 │   Roche/Genentech, Sanofi,    │
         │   REC-1245, REC-3565,    │                 │   Bayer, Merck KGaA           │
         │   REC-4539, REC-7735,    │                 │   (platform access fees +     │
         │   REC-102                │                 │   milestones + royalties)     │
         └─────────────────────────┘                 └─────────────────────────────┘

The model is capital-intensive and pre-commercial: all current revenue is collaboration/milestone-based, not product sales, and the internal pipeline has no approved drugs. Value creation depends on (a) whether phenomics-driven target and candidate selection shortens discovery timelines and improves clinical success rates versus traditional R&D, and (b) whether pharma partners keep renewing and expanding platform access as proof points accumulate.

2. Business Segments / Platform vs. Pipeline Breakdown

Recursion does not report discrete financial segments; management and the 10-K instead describe the business along two commercial tracks that share the same underlying technology stack:

• Platform / Collaboration track — Licenses access to Recursion OS, phenomics maps, and target-identification/generative-chemistry workflows to large pharmaceutical partners (Roche/Genentech, Sanofi, Bayer, Merck KGaA) in exchange for upfront fees, R&D funding, milestone payments and downstream royalties. This track has generated the company's entire reported revenue to date ($74.7 million in FY2025; $58.8 million in FY2024) and over $500 million cumulatively since inception, with up to $20 billion of unearned milestones disclosed as potentially available before royalties.
• Internal Pipeline track — Uses the same platform to advance Recursion's own wholly-owned or in-licensed oncology and rare-disease candidates through IND-enabling studies and Phase 1/2 clinical trials, with the intent of either advancing them to approval directly or out-licensing/partnering them once de-risked (management has stated an intent to selectively monetize pipeline assets to help fund operations). In May 2025 the company narrowed this track, discontinuing or seeking partners for several legacy non-oncology programs (REC-994/CCM, REC-3964/C. difficile, REC-2282/NF2 meningioma) to concentrate capital on oncology.

A supporting technology layer — chemistry (Centaur Chemist, inherited from the 2024 Exscientia combination) and ClinTech (AI-enabled clinical trial design and execution) — underlies both tracks rather than standing as a separate reporting segment.

3. Product Portfolio (Pipeline Snapshot, per FY2025 Form 10-K)

Representative early efficacy data disclosed in the 10-K: daraxonrasib... (not applicable — see RVMD). For Recursion: REC-4881 Phase 2 data showed polyp-burden reduction in FAP; zoldonrasib/REC-9805 data does not apply here. Alongside the owned pipeline, Recursion maintains multi-target discovery collaborations with Roche/Genentech (neuroscience and GI-oncology phenomaps; $213 million in cumulative upfront/milestone payments), Sanofi (joint small-molecule immunology/oncology portfolio; $134 million cumulative, with a fifth milestone achieved February 2026), Bayer (precision oncology data packages and LOWE workflow software), and Merck KGaA (target identification in oncology/immunology). A technology (non-drug) collaboration with NVIDIA underpins the BioHive-2 supercomputer used to train Recursion's models, and data-access arrangements with Tempus and Helix supply real-world patient data for causal AI work.

4. Competitive Landscape

Recursion competes across two overlapping fronts: (1) other AI-native drug discovery platforms seeking pharma partnerships and advancing owned pipelines, and (2) the installed base of traditional pharma/biotech R&D, which increasingly licenses or builds its own AI tools rather than buying them externally. Key named platform peers include Insitro (privately held; ML-first platform paired with in-house wet-lab biology, partnered with Bristol Myers Squibb and Eli Lilly), Isomorphic Labs (DeepMind/Alphabet spinout; structure-prediction-centric, partnered with Eli Lilly and Novartis, no clinical-stage pipeline of its own disclosed), and Schrödinger (publicly traded; physics-based computational chemistry platform combined with a software-licensing business and an owned oncology pipeline, including the MALT1 inhibitor SGR-1505 that directly competes with Recursion's REC-3565).

                        HIGH: Owned clinical pipeline + platform revenue
                                           │
                                   RECURSION (RXRX)
                                   Phenomics + chemistry +
                                   clinic; multi-pharma deals
                                           │
                     SCHRÖDINGER ●         │        ● EXSCIENTIA (merged into RXRX, 2024)
                  (software + pipeline)    │
                                           │
   LOW ─────────────────────────────────── ┼ ─────────────────────────────────── HIGH
   Platform/software monetization          │          Data/wet-lab breadth
                                           │
                  ISOMORPHIC LABS ●        │        ● INSITRO
               (structure prediction,      │     (ML + proprietary wet lab,
                mega pharma deals,         │      pharma co-development deals)
                no clinical pipeline)      │
                                           │
                            TRADITIONAL PHARMA/BIOTECH R&D
                       (highest wet-lab + clinical breadth,
                        lowest AI-native data advantage;
                        increasingly licenses external AI tools)
                                           │
                        LOW: No independent clinical pipeline

Recursion's relative positioning strength is the breadth of its proprietary experimental data (phenomics at massive scale) combined with an owned, advancing oncology pipeline and a growing roster of top-tier pharma partners — a combination few pure-software competitors (Isomorphic Labs) or earlier-stage private peers (Insitro) currently match. Its relative weakness is that, unlike Schrödinger, it has no software-licensing revenue stream and, unlike large pharma, no commercial products or sales infrastructure, leaving it entirely dependent on partnership economics and eventual pipeline approvals to reach sustainable profitability.

5. Strategic Strengths & Risks

Strengths

• Massive proprietary phenomics dataset (~65 petabytes) and throughput (2M+ experiments/week) create a data moat that is difficult and capital-intensive for competitors to replicate from scratch.
• Validated, renewing relationships with four major pharmaceutical companies (Roche/Genentech, Sanofi, Bayer, Merck KGaA) generating real cash milestones (>$500 million cumulative) well ahead of any product approval, which both funds operations and serves as third-party validation of the platform.
• Post-Exscientia combination gives an integrated "biology + chemistry + clinic" stack (Recursion OS, Centaur Chemist, ClinTech) that is broader than most single-point-solution AI drug discovery peers.
• Strong balance sheet for a clinical-stage biotech: $753.9 million in cash/restricted cash at year-end 2025 and management-guided runway into early 2028, reducing near-term financing risk relative to typical small-cap biotech peers.
• Multiple early but encouraging oncology/rare-disease clinical readouts (notably REC-4881's Phase 2 FAP polyp-reduction data) provide incremental proof that the platform's candidates can show clinical activity.

Risks

• Still pre-revenue from products: all revenue is collaboration/milestone-based and small relative to spend; net losses are widening ($644.8 million in FY2025 vs. $463.7 million in FY2024) as R&D investment accelerates.
• Will need to raise substantial additional capital beyond the current runway to fund the business through pipeline maturation and any eventual commercialization, creating dilution risk for existing shareholders.
• The company has previously disclosed a material weakness in internal control over financial reporting (identified in the FY2024 Form 10-K), which, until fully remediated and sustained, is a governance/financial-reporting risk.
• Clinical and regulatory risk is substantial and typical of early-stage biotech: most pipeline assets are Phase 1/2 or earlier, and several non-oncology legacy programs were already discontinued or deprioritized in 2025, underscoring high pipeline attrition even with an AI-driven discovery process.
• Intensifying competition from both well-capitalized AI-native peers (Isomorphic Labs' mega-partnerships with Lilly/Novartis, Insitro's wet-lab-plus-ML model) and traditional pharma building internal AI capabilities, which could compress the exclusivity/pricing of future platform collaborations.
• Integration risk from the 2024 Exscientia combination, the largest in TechBio to date, including a February 2025 agreement to take a 49% stake in a carved-out Austrian entity (Alpha Biotechnology GmbH) and potential obligations (e.g., a Gates Foundation share-repurchase provision tied to global-access commitments) that add structural complexity.

6. Financial Overview

All figures as reported in Recursion's Form 10-K filings and related earnings press releases (unaudited press-release figures for the most recent period; amounts in USD millions unless noted).

No going-concern qualification appears in either the FY2024 or FY2025 Form 10-K; the company explicitly frames its need for "substantial additional funding" as a forward risk factor rather than an immediate doubt about continuing as a going concern, consistent with its multi-year cash runway guidance.

7. Summary Conclusion

Recursion Pharmaceuticals represents a well-capitalized, differentiated bet on AI-native drug discovery rather than a distressed or speculative shell: it combines a genuinely large proprietary experimental dataset, a broadened post-Exscientia technology stack, renewing top-tier pharma partnerships worth hundreds of millions of dollars, and a cash position sufficient to fund operations into early 2028 without immediate financing pressure. The core open question for investors is translation risk — whether phenomics- and AI-driven candidate selection meaningfully improves clinical success rates and timelines versus conventional R&D, and whether that improvement shows up convincingly enough in Phase 1/2 data (as REC-4881's FAP results have begun to suggest) to justify the platform's widening annual losses before the runway narrows. With no approved products, rising R&D spend outpacing revenue growth, and an increasingly crowded field of AI-native and traditional competitors, Recursion sits squarely in "promising but unproven at scale" territory: a credible platform business with real near-term liquidity, but one that still needs several more years of clinical and partnership proof points before its AI-driven thesis can be considered de-risked.