Recursion

Recursion is a clinical-stage biotechnology company that combines artificial intelligence, machine learning, biological research, chemistry, automation, and computational science to discover and develop new medicines. Founded in 2013 and headquartered in Salt Lake City, Utah, the company operates at the intersection of biotechnology and technology, a field often described as TechBio. Its mission is to decode biology and use that understanding to develop treatments for diseases with significant unmet medical needs. Recursion works across pharmaceutical research, drug discovery, preclinical development, and clinical development, with a particular focus on oncology, rare diseases, neuroscience, and immunology.

The company was founded by Blake Borgeson, Chris Gibson, and Dean Li. Rather than operating as a conventional pharmaceutical company that depends primarily on individual laboratory experiments and traditional research methods, Recursion has developed an integrated technology platform that combines physical experimentation with computational analysis. Its approach is based on the idea that biological systems can be studied at scale by generating large quantities of experimental data and using AI models to identify relationships that may not be obvious through conventional research alone. This allows its scientific teams to investigate disease biology, identify potential therapeutic targets, design candidate molecules, and evaluate their suitability for further development.

At the centre of the company’s technology is the Recursion Operating System, commonly called Recursion OS. This proprietary platform connects biological research, chemical design, laboratory automation, machine learning, and clinical development within one continuous system. It is designed to support the drug discovery process from the identification of biological targets through molecular design and the preparation of clinical development programmes. Information generated at each stage can be incorporated into the platform’s datasets and models, helping researchers refine their predictions and guide subsequent experiments.

Recursion OS combines computational and physical research rather than treating them as separate activities. The computational side uses machine learning models and large datasets to generate hypotheses about biological processes and potential medicines. The laboratory side tests those hypotheses through controlled experiments. Results from physical experiments are converted into digital data that can be analysed by the company’s models. This repeated process creates what Recursion describes as a lab-in-the-loop system, in which experimental results inform computational predictions and computational predictions help determine which experiments should be conducted next.

One of the most important parts of this approach is biological data generation. Recursion uses automated laboratory systems to study how cells respond to different genetic and chemical changes. These experiments can involve genetic perturbations, small molecules, cellular imaging, and other biological measurements. The resulting information helps researchers understand how changes in genes, proteins, and chemical compounds affect cellular behaviour. By collecting data across many experimental conditions, the company aims to build detailed representations of biological systems that can support the identification of disease mechanisms and potential treatment opportunities.

Cellular imaging and phenomics are important elements of Recursion’s research strategy. Phenomics involves studying the observable characteristics of biological systems, including changes in cell appearance and behaviour. Recursion uses high-content microscopy and machine learning to examine cellular images and identify patterns associated with different biological conditions. This can help researchers investigate how cells respond to genetic changes, disease-related processes, and experimental compounds. The resulting measurements provide another way to understand biology beyond relying exclusively on previously established research hypotheses.

The company’s automated laboratories support the large-scale production of experimental data. Recursion reports that its laboratory infrastructure can conduct up to two million experiments per week and contributes to a proprietary dataset exceeding 50 petabytes. These figures describe the company’s reported research capacity and data resources rather than the performance of a publicly available software subscription. The scale of this infrastructure is important because modern drug discovery can involve examining large numbers of biological conditions and potential chemical compounds. Automating suitable experiments allows researchers to collect information more consistently and use computational systems to analyse results across many research programmes.

Recursion also uses advanced computing infrastructure to train and operate its machine learning models. Its BioHive-2 supercomputer was developed with NVIDIA technology and supports computationally intensive biological and chemical research. The system provides substantial GPU computing capacity for processing large datasets, training foundation models, and conducting complex scientific analyses. This infrastructure is part of the company’s broader strategy to combine high-volume laboratory experimentation with large-scale computational modelling.

Another major part of Recursion OS is biological target identification. Before developing a medicine, researchers need to understand which biological mechanism or target may be relevant to a particular disease. Recursion uses its biological datasets and computational models to investigate relationships between genes, proteins, cellular responses, and disease-associated processes. Its Maps of Biology approach is intended to help researchers navigate these relationships and identify potential therapeutic opportunities. By comparing experimental results and biological information, the company seeks to identify targets that may not be obvious through traditional research approaches.

The platform also supports precision chemical design. Once a promising biological target has been identified, researchers need to discover or design molecules that can interact with it in a useful way. Recursion applies computational chemistry, generative AI, molecular modelling, and compound evaluation to help identify potential drug candidates. Its capabilities include analysing molecular properties, predicting potential interactions, and exploring chemical structures that may be suitable for further testing. The objective is to connect biological discoveries with molecules that can be synthesised and evaluated in laboratory experiments.

The company strengthened its digital chemistry capabilities through the acquisition of Cyclica and Valence in 2023, bringing additional expertise in computational drug discovery and generative chemistry. These capabilities support the development and evaluation of potential therapeutic molecules. Recursion has also worked with Enamine to explore chemical compounds and enriched compound libraries. Its MatchMaker technology has been used to identify potentially relevant compounds within Enamine’s REAL Space, a large collection of synthetically accessible chemical structures. This work demonstrates how computational predictions can be connected to practical chemistry and compound availability.

Recursion’s drug discovery process also incorporates a Design-Make-Test-Learn approach. Researchers use computational tools to design potential molecules, select compounds for synthesis, test them in laboratory systems, and use the results to improve subsequent designs. This process is intended to reduce the gap between theoretical molecular predictions and compounds that can actually be produced and evaluated. The approach is particularly important because a molecule that appears promising in a computer model may not necessarily have the required chemical properties, biological activity, or safety characteristics when tested experimentally.

Clinical development is another part of Recursion’s integrated research strategy. Discovering a potential drug candidate is only one stage in developing a medicine. Researchers must also determine which patients may benefit, identify relevant biomarkers, evaluate safety and efficacy, and design suitable clinical trials. Recursion applies data analysis and computational methods to support these activities. Its ClinTech capabilities are intended to connect biological and molecular discoveries with clinical development, including patient selection and clinical trial planning. These capabilities help the company investigate whether a candidate treatment may be appropriate for a particular patient population.

Recursion is also developing a broader predictive representation of biological and chemical systems, which it describes through its World Model concept. The goal is to create computational representations that can help researchers predict how biological systems and chemical compounds may behave under different conditions. Such models could help scientists generate hypotheses, investigate molecular relationships, and prioritise experiments. The concept is part of the company’s longer-term effort to improve the accuracy and usefulness of AI models in drug discovery. It should be understood as an evolving research capability rather than a guarantee that complex biological outcomes can already be predicted perfectly.

The company makes some of its research datasets and models available through RxRx.ai, its open-science platform. This resource provides access to selected biological datasets, machine learning tools, and scientific research created using data generated through Recursion OS. The public resources are intended to support academic research, machine learning development, and collaboration across the scientific community. They include cellular imaging datasets and other biological research resources that can be used to investigate machine learning approaches for biological analysis.

One of the major public datasets associated with Recursion is RxRx3, a large collection of cellular screening data. It contains information generated from experiments involving different biological perturbations and has been used in research on machine learning methods for drug discovery. The public platform also features MapApp, which allows users to explore relationships between biological and chemical information. MapApp is designed to help researchers investigate connections, generate hypotheses, and examine possible therapeutic relationships using available datasets. Access to these public resources should not be confused with access to the full proprietary Recursion OS platform. Recursion states that its publicly available datasets represent less than one per cent of its complete proprietary dataset.

Recursion has also contributed foundation models to the broader scientific and machine learning ecosystem. One example is OpenPhenom-S/16, a foundation model designed for microscopy data. It was made available through Google Cloud’s Vertex AI Model Garden and Hugging Face for non-commercial use. The company has also collaborated with MIT on Boltz-2, an AI model for molecular structure and binding-affinity prediction. These research releases provide opportunities for scientists and developers to investigate selected technologies without obtaining access to the company’s entire internal research platform. However, the availability and licensing conditions of individual models and datasets should be checked separately.

Strategic partnerships are an important part of Recursion’s business model. The company works with pharmaceutical organisations, biotechnology companies, technology providers, and research institutions to explore disease biology and develop potential medicines. Its therapeutic collaborations have included Bayer, Roche and Genentech, Sanofi, Merck KGaA, and the Gates Foundation. These relationships allow Recursion to combine its computational and experimental capabilities with the disease expertise, research resources, and development experience of established organisations.

Its collaboration with Roche and Genentech focuses on neuroscience and oncology research. The companies use Recursion’s platform and biological mapping capabilities alongside partner-generated research data to identify novel biological relationships and potential therapeutic programmes. The collaboration has included the development of phenomaps and the investigation of neuroscience targets. Its Bayer partnership focuses on precision oncology programmes and the identification of potential therapeutic targets in challenging cancer indications. The Sanofi collaboration has focused on discovering precision-engineered small-molecule medicines in oncology and immunology. Recursion has also worked with Merck KGaA on AI-enabled drug discovery programmes in oncology and immunology.

Technology partnerships are equally important. NVIDIA supports Recursion’s computing and AI development through GPU infrastructure and related software technologies. Google Cloud provides cloud computing and generative AI capabilities, including the use of Gemini models and BigQuery to support Recursion’s research environment. Partnerships with Tempus and Helix provide access to de-identified patient and genomic information that can support research into disease mechanisms, biomarkers, and patient stratification. These relationships help connect laboratory-generated biological information with real-world clinical and genomic data.

Recursion is not simply a software company selling an AI application to individual users. It is a biotechnology business that uses its proprietary platform to develop its own investigational medicines and collaborate with other pharmaceutical organisations. Its commercial activities include research collaborations, technology partnerships, and the advancement or licensing of potential therapeutic assets. The platform is therefore primarily an internal and partnership-based research system rather than a conventional self-service product with a public monthly subscription.

The company’s internal medicine pipeline includes investigational programmes in oncology and rare diseases. REC-4881 is a MEK1/2 inhibitor being developed for familial adenomatous polyposis, a rare condition associated with an increased risk of gastrointestinal tumours. REC-1245 is an RBM39 degrader being investigated for biomarker-enriched solid tumours and lymphoma. REC-4539 is an LSD1 inhibitor being studied in solid tumours and haematological cancers, while REC-7735 is a selective PI3Kα H1047R inhibitor being investigated for certain solid tumours. REC-102 is an ENPP1 inhibitor being developed for hypophosphatasia and remains at the preclinical stage. These are investigational programmes, and their development status and potential benefits remain subject to clinical evaluation and regulatory review.

The company’s research has applications across several scientific and healthcare areas. Pharmaceutical companies can use its discovery capabilities through collaborations to investigate new biological targets and potential medicines. Academic researchers can use its public datasets and research models to explore biological imaging, machine learning, and computational drug discovery. Biotechnology organisations may find its platform relevant when investigating disease mechanisms, designing compounds, or developing new therapeutic programmes. The public RxRx resources also provide a way for machine learning researchers to work with biological data without needing to operate an industrial-scale laboratory.

Recursion’s technology and business model have important limitations. Its platform does not guarantee that a predicted drug candidate will succeed in clinical trials or receive regulatory approval. Drug discovery remains a complex process with substantial scientific, financial, and clinical risks. AI predictions must be supported by experimental evidence, and promising laboratory results do not necessarily translate into effective treatments for patients. The company also relies on partnerships, research data, laboratory infrastructure, and substantial computing resources to support its work.

Overall, Recursion represents an integrated approach to AI-enabled drug discovery and development. Its combination of automated laboratory experimentation, biological data generation, computational chemistry, machine learning, high-performance computing, and clinical research gives it a broader role than a standalone scientific AI application. Its proprietary Recursion OS, public research resources, pharmaceutical partnerships, and investigational medicine pipeline make it relevant to biotechnology, pharmaceutical research, biomedical science, and academic machine learning. It is best understood as a TechBio company using AI and automation to improve the process of discovering and developing potential medicines.

Pricing

Recursion does not publish a standard subscription price for Recursion OS or a public per-user pricing plan. Its main platform is used internally and through commercial research and development partnerships with pharmaceutical and biotechnology organisations. Prospective partners are encouraged to contact the company to discuss collaboration opportunities.

The commercial arrangements may involve research funding, upfront payments, development and regulatory milestones, commercialisation milestones, licensing arrangements, and royalties. The financial terms depend on the partnership and the programmes involved. These collaboration payments are not equivalent to a fixed software subscription price.

Recursion also provides selected public datasets, research tools, and models through RxRx.ai. Some resources are available for research use, while individual models and datasets may have specific licensing restrictions. The availability of these public resources does not mean that the complete proprietary Recursion OS platform is free or open source.

Review

Recursion is an interesting example of how AI, laboratory automation, and large-scale biological data can be combined to support pharmaceutical research. Its main strength is the integration of computational predictions with physical experiments. Rather than relying exclusively on AI-generated hypotheses, the company uses automated laboratories to test biological and chemical predictions and feeds the results back into its models. This creates a connected research process that can help scientists investigate complex biological relationships and prioritise potential drug candidates.

The Recursion OS platform is one of its strongest features. It brings biological research, molecular design, experimental testing, and clinical development into a connected environment. This broad approach is valuable because drug discovery involves several stages that can become disconnected when handled through separate systems and teams. By connecting these activities, Recursion aims to improve how information moves from early research into potential therapeutic programmes.

Its laboratory automation and proprietary datasets are also significant advantages. Large-scale cellular experiments can generate information that would be difficult to collect through manual processes alone. Recursion’s use of high-content imaging, machine learning, and computational biology provides a way to analyse these results and investigate patterns across different experimental conditions. Its BioHive-2 computing infrastructure further supports the processing and modelling required for this research.

The company’s public research initiative is another positive feature. RxRx.ai provides selected datasets, models, and tools that can support academic research and machine learning development. This gives researchers opportunities to investigate biological data and contribute to scientific methods without requiring access to Recursion’s entire proprietary platform. Public resources such as OpenPhenom-S/16 also demonstrate the company’s contribution to the wider scientific AI community.

Recursion’s pharmaceutical partnerships provide another important strength. Working with organisations such as Bayer, Roche and Genentech, Sanofi, and Merck KGaA allows the company to combine its technology with established pharmaceutical research and development capabilities. These collaborations can create opportunities to investigate disease areas, identify new targets, and advance potential medicines. Its internal pipeline also demonstrates that the company applies its platform to its own investigational programmes rather than relying exclusively on external partnerships.

However, Recursion is not a simple AI application that an individual researcher can subscribe to and use immediately. Its core platform is proprietary and primarily supports the company’s internal research and commercial collaborations. Organisations interested in using its capabilities may need to establish a partnership and agree on project-specific commercial terms. The absence of public software pricing also makes it difficult for prospective customers to compare costs without contacting the company.

Another limitation is the uncertainty associated with drug development. AI can help identify patterns, predict molecular behaviour, and prioritise research, but it cannot eliminate the scientific risks involved in developing medicines. Potential drug candidates still need to demonstrate appropriate safety and efficacy through preclinical studies and clinical trials. The company’s investigational pipeline should therefore be viewed as research in progress rather than a collection of approved treatments.

The platform also depends on extensive data, laboratory infrastructure, specialised scientific expertise, and high-performance computing. This makes its approach resource-intensive and may limit the ability of smaller organisations to reproduce the same research environment independently. Furthermore, public datasets represent only a small portion of the company’s complete proprietary data resources, so external researchers cannot assume that every internal capability is available through RxRx.ai.

Overall, Recursion is a strong example of an AI-enabled biotechnology company with a broad research and development platform. Its integration of biological experimentation, computational chemistry, machine learning, automation, and clinical development makes it relevant to organisations working on complex drug discovery challenges. Its public research resources add value to the scientific community, while its partnerships and internal pipeline demonstrate the practical direction of its technology. It is most suitable for pharmaceutical companies, biotechnology organisations, and biomedical researchers interested in advanced drug discovery rather than individuals seeking a general-purpose AI tool.