Recursion Review
VisitAI-powered drug discovery at biological scale
Recursion is a clinical-stage biotechnology company using AI and automation to accelerate drug discovery.
AI Panel Score
7.4/10
6 AI reviews
Reviewed
AI-powered drug discovery at biological scale
Recursion is a clinical-stage biotechnology company using AI and automation to accelerate drug discovery.
AI Panel Score
6 AI reviews
Reviewed
Recursion is a clinical-stage biotechnology company headquartered in Salt Lake City, Utah, that integrates artificial intelligence, machine learning, and large-scale biological experimentation to industrialize the drug discovery process. The company operates one of the largest biological and chemical datasets in the industry, generated through automated wet-lab experiments that capture cellular responses to genetic and chemical interventions at massive scale.
At the core of Recursion's approach is the Recursion OS, a vertically integrated platform that spans data generation, storage, and AI-driven analysis. The system uses high-content imaging to observe how cells change under various conditions, producing multimodal datasets that feed proprietary machine learning models. These models are designed to find patterns and relationships in biology that would be difficult or impossible to detect through traditional research methods.
Recursion serves both its own internal drug pipeline and external partners, including large pharmaceutical companies. Through partnerships and collaborations, the Recursion OS is made available to organizations seeking to accelerate their own discovery programs. Notable partners have included Bayer and Roche, reflecting the platform's positioning as an enterprise-grade infrastructure for drug discovery.
The company competes in the emerging field of AI-driven drug discovery alongside companies such as Insilico Medicine, Exscientia, and Schrödinger. Recursion differentiates itself through the scale of its experimental data generation capabilities and its fully integrated technology stack, rather than relying solely on computational modeling applied to existing datasets.
As a clinical-stage company, Recursion also advances its own therapeutic programs into human trials, giving it a dual identity as both a technology platform provider and a drug developer. Pricing and access to the Recursion OS for external partners is handled through direct commercial agreements rather than publicly listed tiers.
Includes absorption, distribution, metabolism, and excretion (ADME) data within the platform to inform and optimize drug candidate selection and design.
Machine learning models trained on proprietary datasets rapidly identify new biological targets for potential first-in-class and best-in-class drug candidates.
AI models design and optimize novel molecules to fuel Recursion's pipeline, reducing reliance on traditional trial-and-error medicinal chemistry approaches.
Leverages de-identified patient data within the discovery platform to improve the relevance and translational potential of AI-driven drug candidates.
Captures large-scale cellular phenotype data from high-throughput imaging to map how genetic and chemical perturbations affect living cells.
Aggregates over 50 petabytes of fit-for-purpose data spanning phenomics, transcriptomics, proteomics, ADME, and de-identified patient data to fuel drug discovery models.
Incorporates transcriptomics and proteomics datasets into the platform to provide multi-modal biological context for AI model training and drug discovery.
Automated wet lab uses robotics and computer vision to capture millions of cell experiments per week, generating imaging data from genetic and chemical perturbations.
Offers pharmaceutical, technology, and data partners access to Recursion OS capabilities to accelerate collaborative discovery of new medicines.
Biopharma's most powerful supercomputer, built in partnership with NVIDIA, used to process the massive biological and chemical datasets generated by Recursion's platform.
Tracks therapeutic candidates across pipeline stages—Preclinical, IND-Enabling, Phase 1/2, and Pivotal/Phase 3—across oncology and rare disease indications.
An integrated drug discovery and development platform that combines data, models, and compute in a continuously improving feedback loop from hit identification to IND-enabling studies.
Recursion does not license its drug discovery platform as a self-serve SaaS product. Commercial engagement happens through pharma partnership deals, joint venture programs, and academic collaborations. Deal structures range from milestone-based licensing to multi-year strategic partnerships with major pharmaceutical companies.
50 petabytes of proprietary bio data is a real moat — if you can access it.
“Recursion isn't a SaaS buy. It's a strategic partnership decision with a clinical-stage company that has Bayer and Roche already at the table.”
NASDAQ-listed, Bayer and Roche as named partners, BioHive-2 built with NVIDIA. That's not a startup pitch — that's a company with institutional validation and a 36-month runway story that holds up. The 50 petabytes of phenomics, transcriptomics, and ADME data is genuinely differentiated versus Exscientia or Insilico Medicine, who lean harder on pure compute modeling without the same experimental data flywheel.
The tradeoff is access. There's no pricing page, no free trial, no API docs. You're not buying a seat — you're negotiating a milestone-based deal. That means 6-12 months before you see anything useful, and your legal team will earn their salary.
If your pipeline has a genuine discovery bottleneck and you can staff a real partnership, this is worth the conversation. If you need something that ships value in a quarter, look elsewhere.
50 petabytes of proprietary biological data puts this ahead of Exscientia and Insilico Medicine on raw discovery infrastructure.
Roche and Bayer are already in — no board is going to squint at this vendor relationship.
Custom milestone-based deal structures and no self-serve access mean you won't see business outcomes inside 12 months.
Recursion OS directly advances first-in-class drug identification — this isn't cost savings, it's pipeline acceleration.
NASDAQ-listed, NVIDIA partnership for BioHive-2, and named pharma partners like Bayer and Roche signal institutional staying power.
Large pharma or well-funded biotech with an active discovery bottleneck and a team that can manage a multi-year strategic partnership.
You need demonstrable ROI inside 12 months or lack the legal and science staff to manage a milestone-based licensing deal.
50 petabytes of proprietary biology and BioHive-2 compute is a genuine moat.
“Recursion isn't a tool you adopt — it's an infrastructure bet you partner into. For pharma data science teams, the multimodal dataset depth (phenomics, transcriptomics, proteomics, ADME, patient data) is the real asset, not the ML models sitting on top.”
The 50-petabyte proprietary dataset is the architectural anchor here. Most competitors — Exscientia, Insilico Medicine — compete on generative chemistry and computational modeling applied to existing public datasets. Recursion's differentiation is the wet-lab-to-model feedback loop: millions of cell experiments weekly feeding a continuously improving system. That's not a feature, that's a data flywheel, and it compounds.
The BioHive-2/NVIDIA partnership signals serious compute infrastructure, not cloud-rented capacity. For a Head of Data Science evaluating a partnership, that matters — model latency on 50PB isn't a solved problem with commodity GPUs. The tradeoff is real access friction: no API docs, no self-serve tier, no public pricing. You're negotiating milestone-based licensing, not spinning up a trial environment.
If we enter a Recursion OS partnership, in 3 years we've either co-generated proprietary data assets we own (good) or we've built discovery workflows dependent on their closed-loop platform (risk). Clarity on data ownership terms in the commercial agreement is the due diligence priority before anything else.
Bayer and Roche partnerships signal enterprise validation; proprietary experimental data generation separates Recursion from purely computational competitors like Schrödinger.
Multimodal data (phenomics, transcriptomics, ADME, patient data) maps directly to how serious drug discovery data science teams structure their feature engineering and model inputs.
No public API, no docs, no changelog — integration happens through negotiated partnership structures, not engineering, which limits how deeply internal data science teams can embed their own tooling.
The compounding data flywheel is a strategic asset, but closed-platform dependency and opaque commercial terms create 3-year lock-in risk if data ownership isn't contractually protected.
Vertical integration from automated wet lab through BioHive-2 compute to AI models is genuinely best-in-class architecture for biological-scale ML.
Pharma or biotech data science orgs with enterprise BD resources who need biological-scale training data they can't generate internally.
Your team needs direct API access, iterative prototyping capability, or transparent pricing before committing to a multi-year commercial structure.
50 petabytes of data, zero public pricing — procurement starts blind
“Recursion OS is enterprise-only, contract-driven, no tiers visible. Every number gets negotiated behind closed doors.”
No pricing page. No trial. No self-serve. Commercial access runs through direct partnership deals — milestone-based licensing, joint ventures, custom structures. Zero sticker to anchor against. Procurement can't even build a rough model without a sales cycle.
The scale is real: 50 petabytes of proprietary data, BioHive-2 supercomputer built with NVIDIA, high-throughput imaging generating millions of cell experiments per week. Bayer and Roche have signed on. That's signal, not noise. But none of that tells a finance team what year-3 looks like.
Compare to Schrödinger — also enterprise, also opaque, but computational-only, no wet-lab integration. Recursion's dual identity as platform provider and drug developer is the tradeoff: partners get genuine capability access, but the vendor's internal pipeline competes for the same assets. Deal terms almost certainly reflect that complexity.
Custom deal structures mean lengthy procurement cycles — no standard invoicing model, no self-serve onboarding.
Milestone-based licensing and multi-year strategic partnerships suggest long terms with limited exit optionality.
No published tiers, no starting price — 100% contact-sales, per their own pricing page absence.
Addressing the 90% drug discovery failure rate is a compelling headline, but measuring platform-specific ROI vs. internal R&D spend is structurally murky.
No invoices, no overages, no public deal structure — TCO modeling is impossible without NDAs and term sheets.
Large pharma or well-funded biotech that can absorb a complex partnership negotiation and has 18+ months of runway to absorb deal timelines.
Any organization that needs predictable SaaS-style budgeting or standard procurement timelines.
50 petabytes of biological data, zero self-serve access — partner or walk away
“Recursion OS is serious infrastructure: BioHive-2 supercomputer, multimodal data across phenomics, transcriptomics, and proteomics, high-throughput imaging generating millions of cell experiments weekly. But it's not a platform you onboard — it's a platform you negotiate.”
The scale here is real. Over 50 petabytes of fit-for-purpose biological data, a dedicated NVIDIA-partnered supercomputer, and automated wet labs running genetic and chemical perturbations at industrial throughput. For an ML engineer working on bioassay modeling or target identification, that data moat is the whole story. Schrödinger and Exscientia both lean heavily on computational modeling against existing datasets. Recursion generates its own ground truth. That's a genuine architectural difference.
Day three looks like this: you're inside a pharma partnership, not a SaaS trial. No API docs in the evidence, no changelog, no self-serve tier. Workflow integration happens at the deal-structure level — milestone licensing, joint ventures, custom agreements. That's not a friction point, it's a category constraint. You don't file a Jira ticket to get data access; you wait on legal.
For an ML engineer at a major pharma partner, the ADME modeling and patient data integration features suggest the models are built for translational relevance, not just benchmark performance. That's encouraging. But with no public docs and no free trial, evaluating model quality before committing to a multi-year deal is opaque by design.
No self-serve access, no trial — day three is whatever your partnership agreement allows, not what you can explore independently.
Blog exists but no docs portal, no API reference, and no changelog in the evidence — reads like a BD-facing site, not an ML engineer's resource.
No changelog, no public docs, and contact-only access mean every workflow question routes through account management, not documentation.
BioHive-2 compute, multimodal data across phenomics, transcriptomics, proteomics, and ADME, plus proprietary model training loops suggest real depth for serious ML workloads.
Recursion OS spans hit identification to IND-enabling studies, suggesting end-to-end fit for pharma ML pipelines, but integration depth is opaque without public API docs.
ML engineering teams embedded inside a major pharma or biotech organization with budget and legal capacity to negotiate a multi-year Recursion OS partnership.
You need to evaluate model quality, run exploratory queries, or prototype a pipeline before signing a commercial agreement.
50 petabytes of biological data and a supercomputer — not a SaaS tool, a serious bet
“Recursion isn't software you try out — it's a platform you partner into. For big pharma, that's a feature, not a bug.”
This isn't Notion. There's no free trial, no pricing page, no spinner to complain about. Recursion OS is a deal you negotiate, a partnership you sign, and a relationship you commit to. The BioHive-2 supercomputer built with NVIDIA, over 50 petabytes of proprietary data, Bayer and Roche already at the table — this is enterprise infrastructure that happens to run AI, not the other way around.
For pharma buyers, the pitch is real. High-throughput cell imaging generates millions of experiments per week. ADME modeling, transcriptomics, phenomics — it's all in one stack, which is genuinely rare against competitors like Exscientia or Schrödinger who lean heavier on computational modeling of existing data.
The tradeoff is access. No docs, no API, no changelog publicly visible. A solo researcher or a small biotech can't evaluate this without a business development call. If your org can't get a seat at that table, the platform doesn't exist for you.
No public UI, no changelog, no micro-copy to evaluate — the daily experience is entirely locked behind partner agreements.
The integrated Recursion OS stack spanning hit identification to IND-enabling studies is powerful but steep for any new partner team to absorb.
Web-only platform with no stated mobile experience — for a tool used in enterprise deal contexts, this dimension barely applies.
Onboarding is a pharma partnership negotiation, not a product flow — no free trial, no docs, no self-serve path.
BioHive-2 supercomputer and Bayer/Roche partnerships suggest infrastructure-grade reliability, even if it can't be observed directly.
Large pharmaceutical companies or well-resourced biotechs that can negotiate enterprise partnership deals.
You need to evaluate, pilot, or budget a tool without going through a sales and legal process first.
50 petabytes of proprietary data is real — but this isn't SaaS, it's a pharma deal
“Recursion OS is genuinely differentiated infrastructure for drug discovery. But calling it a 'platform' blurs the line between enterprise partnership and licensable software.”
Three tells upfront. One: no pricing page, no docs, no API, no changelog. Two: 'pioneering' is in the H1 — the kind of superlative that ages poorly. Three: the free tier is literally 'Partnership Engagements.' That's not a product tier. That's a BD pipeline.
What's real: BioHive-2 built with NVIDIA, 50 petabytes of proprietary phenomics data, Bayer and Roche as named partners. High-throughput imaging at millions of cell experiments per week isn't a slide claim — that's physical infrastructure you can't fake. Exscientia and Insilico don't have this at scale. That's a genuine moat, maybe.
The tradeoff: you're not buying software. You're entering a multi-year commercial negotiation. Exit portability is near-zero — your compounds and datasets live inside their stack. If Recursion's clinical pipeline stumbles, partner programs could freeze. Clinical-stage means cash burn. Watch the pipeline.
50 petabytes of proprietary multimodal biological data plus BioHive-2 supercomputing is a physical scale advantage Exscientia and Insilico Medicine can't easily replicate computationally.
No API, no data export docs, and custom deal structures mean partner data and workflows are deeply embedded — migration would require renegotiation, not just a CSV export.
Clinical-stage means ongoing cash burn; no public funding round listed in evidence, and pipeline execution risk is real — but named enterprise pharma partners suggest institutional confidence.
'Pioneering AI-driven solutions' is vague enough to mean nothing; no pricing, no docs, and a changelog-free site suggest the product is harder to evaluate than the messaging implies.
Bayer and Roche partnerships are named and public; clinical pipeline spanning Phase 1/2 programs in oncology and rare disease shows the platform moves beyond demo-ware.
Large pharma or biotech with budget for multi-year strategic partnerships and a specific need for phenomics-scale data generation.
You need transparent pricing, API access, or any ability to exit cleanly within 18 months.
Common questions answered by our AI research team
Recursion OS is Recursion's platform offered to external partners, giving them access to Recursion's AI-driven drug discovery capabilities and proprietary datasets.
Recursion trains AI on images of cells to understand biological space and cellular disruptions driving disease, using those models to identify potential drug candidates and reduce drug discovery failure rates.
The platform generates massive proprietary datasets from high-throughput imaging of cells, capturing how genetic and chemical perturbations affect living cells.
Yes, external partners can access Recursion's capabilities through the Recursion OS platform.
Recursion addresses the 90% failure rate of traditional drug discovery by using AI and automation to decode biology and identify viable drug candidates more efficiently.
Recursion is a Salt Lake City-based biotech company using AI and automation to industrialize drug discovery through its proprietary Recursion OS platform.
