Recursion Pharmaceuticals (RXRX) – AI-Driven Drug Discovery At An Inflection Point

Recursion was founded in 2013 by Chris Gibson, Ph.D., during his MD/PhD training at the University of Utah, together with mentors and co-founders at the intersection of biology, data science and engineering. From the start, the idea was to industrialise cell-based experimentation: use automated microscopes, robotics and cloud compute to generate a huge image-based dataset of how cells respond to drugs, genetic edits and other perturbations, then train AI models to read that dataset at scale.

Over roughly a decade, Recursion has:

• Built one of the largest proprietary phenotypic screening datasets in pharma, with millions of microscopy images capturing cellular responses.
• Invested heavily in computing infrastructure, including high-end NVIDIA hardware, to run deep-learning models for representation learning and target discovery.
• Expanded into a multi-site operation in the US, Canada and Europe, hiring a mixed team of biologists, chemists, data scientists and engineers.
• Converted parts of this platform into concrete clinical programs and external collaborations with large pharma players.

The stated mission is to “decode biology to radically improve lives”. For investors, the practical question is whether this infrastructure can consistently translate into drugs that get approved and generate meaningful revenue. REC-4881 in FAP is the first serious datapoint in that direction.

Strategic pharma partnerships

Recursion’s business model is not limited to its own internal assets. The Recursion OS is positioned as a discovery engine that can feed both proprietary and partnered pipelines. Two anchor collaborations stand out:

• Roche / Genentech collaboration (2021–): up to 40 programs in neuroscience and oncology. Roche pays Recursion for mapping disease-relevant biology and nominating candidates, with billions in potential milestones and royalties if drugs reach the market.
• Sanofi collaboration (2022–): up to 15 programs in oncology and immunology, anchored by a 100 million USD upfront payment and up to 5.2 billion USD in milestones across the portfolio.

These deals both validate the technology and provide non-dilutive capital. They also give Recursion access to big-pharma development and commercial infrastructure for some of the resulting drugs.

Internal clinical pipeline (selected)

Not every early program has worked. In 2025, Recursion discontinued several first-generation assets, including REC-994 (for cerebral cavernous malformation), as part of a pipeline reprioritisation. That pruning is normal in biotech, but also a reminder that the AI layer does not remove biological risk: it tries to improve the odds, not eliminate failure.

Disease background

Familial adenomatous polyposis (FAP) is a rare, inherited syndrome driven by pathogenic variants in the APC gene. Patients develop hundreds to thousands of adenomatous polyps in the gastrointestinal tract and face a near-certain risk of colorectal cancer if left untreated. Current management revolves around surveillance, endoscopic polypectomies and major surgeries (often colectomy), with no approved drug therapy specifically indicated for FAP.

Key TUPELO Phase 1b/2 results

• Design: multicentre, dose-escalation and expansion study of REC-4881 (4 mg once-daily) in adult FAP patients, evaluating polyp burden reductions and safety.
• Efficacy at Week 12: 75 percent of evaluable patients showed reductions in total polyp burden, with a 43 percent median reduction in total polyp count from baseline.
• Durability at Week 25 (12 weeks after stopping therapy): 82 percent of evaluable patients maintained a reduction in polyp burden, with a 53 percent median reduction relative to baseline.
• Upper GI benefit: approximately 40 percent of patients experienced a one-stage or greater improvement in Spigelman stage, suggesting meaningful disease modification in duodenal polyposis.
• Safety: adverse events were consistent with MEK-inhibitor class effects. Most were Grade 1–2; around 16 percent of patients experienced Grade 3 events. No Grade 4 or higher treatment-related events were reported in the presented dataset.

Compared with natural-history cohorts where the vast majority of FAP patients show progressive polyp growth over a year, these reductions and their durability after treatment cessation are clinically significant. For Recursion, the more important point is that REC-4881 is not a random MEK inhibitor tossed into FAP; it is the output of an AI-driven phenotypic insight linking MEK1/2 modulation to APC-loss biology. That closes the loop from platform prediction to human proof.

Regulatory path and next steps

Following these results, Recursion plans to engage the FDA in the first half of 2026 to discuss a potential registration pathway for REC-4881 in FAP. Management has indicated:

• An intention to expand the eligible population by lowering the age threshold from 55 years to 18 years in upcoming cohorts.
• An exploration of alternative dosing schedules and combination approaches, aiming to preserve efficacy while further optimising tolerability.
• Openness to designing a pivotal trial that could support approval in the highest-risk FAP subgroups if the agency agrees on endpoints and design.

J.P. Morgan models more than one billion USD of potential peak annual sales in the US alone for REC-4881 if developed successfully in FAP and related indications. That makes this program one of the most important value drivers in the Recursion story.

Recursion is still pre-revenue from a product perspective, but it does generate collaboration revenue and occasional milestones:

• Q3 2025 revenue: roughly 5 million USD, well below prior analyst expectations around 17 million USD, underlining the volatility of milestone-based income.
• Q3 2025 loss per share: approximately 0.36 USD, slightly better than consensus but still reflecting a large operating loss.
• Operating expenses: R&D hovering around 120 million USD for the quarter, G&A over 40 million USD, illustrating the cost of running both an AI platform and multiple clinical studies.
• Cash and equivalents: about 785 million USD after recent financings and milestone receipts, bolstered by an at-the-market equity program in 2025.
• Runway: management repeatedly guides for a cash runway extending through 2027, assuming current plans and expected collaboration inflows.

The financial picture is typical of an ambitious platform biotech: strong balance sheet today, meaningful but manageable dilution behind, and a clear dependency on future clinical and business execution to avoid further heavy dilution later in the decade.

Retail and social sentiment (Reddit, Stocktwits, X)

• On retail forums (Reddit communities focused on growth and speculative names), RXRX is framed as a classic “AI plus biotech” bet. Enthusiastic posts describe Recursion as a potential “AWS of biotech” or a future platform that could license out its discovery engine across the industry. The tone is often strongly bullish but explicitly speculative.
• There is also a more cautious camp: threads highlight the high price-to-sales multiple, lack of approved products and the fact that this is still essentially “a bet” on management and the platform, not on current numbers.
• On Stocktwits, sentiment oscillated over 2025: periods of bearishness around earnings and dilution, followed by spikes in optimism when AI partnerships or the FAP data hit the tape. Daily sentiment recently shifted more positive after the REC-4881 TUPELO data, but the message stream is still split between optimists and sceptics.
• On X, most discussion centres on the FAP data and the broader theme of AI-enabled drug discovery. Quantitative sentiment tools classify recent news flow as clearly positive, but again, this reflects news headlines, not a guarantee of stock performance.

Note: all these are comments from non-professional traders and investors. They can be useful to gauge attention and crowd mood, but they are not research and should never be treated as a reliable basis for financial decisions.

Analyst and institutional sentiment

• Street coverage remains relatively limited. Consensus 12-month target is in the mid-single digits, with ratings clustered around Hold, a handful of Buy ratings and some cautious equal-weight views.
• The fresh J.P. Morgan upgrade to Overweight with an 11 USD target is one of the clearest institutional votes of confidence so far and could attract incremental fundamental interest if additional positive data arrives.
• Leading AI-thematic investor ARK Invest holds a significant stake (around seven percent of shares outstanding across its ETFs), signalling long-term conviction in the TechBio narrative despite volatility.
• Short interest remains elevated, reflecting a sizeable group of investors betting against Recursion’s ability to convert AI buzz into durable value. This keeps the stock in a high-beta, high-volatility regime.

1. 1H 2026: FDA interaction on REC-4881 FAP registration pathway (study design, endpoints, potential expedited paths).
2. 2026: Additional TUPELO data updates, including younger cohorts (down to 18 years) and potentially longer follow-up on durability and safety.
3. 2026–2027: Early-stage data from REC-617 and other oncology programs (REC-1245, REC-3565), which will test whether the platform can deliver multiple clinically meaningful assets.
4. Ongoing: Milestone announcements under Roche and Sanofi collaborations as new AI-derived programs advance or hit data delivery checkpoints.
5. Structural: Leadership transition on January 1, 2026, with Najat Khan stepping in as CEO and Chris Gibson moving to Chair of the Board, and how this translates into execution discipline and capital allocation.