Recursion (RXRX) Stock: Can AI Drug Discovery Prove Itself?

1. Introduction: Why Recursion Pharmaceuticals Stock Is at a Turning Point

Recursion Pharmaceuticals stock closed at $4.36 on 9 October 2026. In July 2021 the same shares closed above $40. Over the past twelve months alone the stock has traded between $2.77, a low set on 19 May 2026, and $7.18. For a company that set out to “industrialise” drug discovery with machine learning, the share price has told a harsh story: investors have stopped paying for the idea and are waiting for proof.

That proof is now close. Recursion’s lead programme, REC-4881 for a rare inherited colon cancer syndrome, has updated data scheduled for 2 November 2026, and the company expects an update on its registrational path with the FDA in the second half of 2026. Several other programmes report early human data across 2026 and 2027. At the same time, the business is burning roughly $100 million of cash a quarter against revenue of under $8 million.

The setup is unusually clear. If Recursion’s platform produces drugs that work in patients, today’s valuation of around $2.3 billion could look small for a company with more than $500 million already collected from pharmaceutical partners. If the clinical data disappoint, the platform story weakens, the cash runway shortens in practice, and dilution becomes the most likely path.

This article provides a complete, investor-grade deep dive into Recursion’s technology, business model, financials, risks, and long-term investment outlook, with the bear case given the same weight as the bull case.

2. Company Overview: A TechBio Pioneer Rebuilt for the Clinic

Recursion was founded in 2013 and is headquartered in Salt Lake City, Utah. It describes itself as a “TechBio” company: instead of starting with a single disease hypothesis, it runs enormous numbers of automated experiments on human cells and uses machine learning to find relationships between genes, compounds, and disease that a human scientist would struggle to see.

The company has changed shape twice in two years. In November 2024 it completed its combination with Exscientia, a UK-based AI drug design company, which became a wholly owned subsidiary and stopped trading on Nasdaq. Then on 1 January 2026, Najat Khan, who joined from Johnson & Johnson in July 2024, became chief executive, succeeding co-founder Chris Gibson. Recursion has since added Hoifung Poon as Chief AI Officer.

The practical effect is a company trying to move from “we have the biggest biological dataset” to “we have drugs that work.” Exscientia is no longer a separate competitor; its chemistry design tools now sit inside Recursion.

What it sells today: platform access and research work for large pharmaceutical partners, paid through upfront fees and milestones
What it is building: a wholly owned pipeline of small molecule drugs, mostly in oncology and rare disease
Key partners: Roche and Genentech, Sanofi, Bayer, and Merck KGaA, plus a data partnership with Tempus AI
Scale: more than 2 million automated experiments per week, according to its latest annual report
Listing: Nasdaq, ticker RXRX, about 535 million Class A shares outstanding as of 31 July 2026

3. Technology and Core Innovation: The Recursion Operating System

Traditional drug discovery usually starts with a target, a protein scientists believe drives a disease, and then searches for a molecule that changes it. The weakness of that approach is that the target choice is often wrong, and the error is only discovered years later in human trials. Recursion’s central bet is that you can choose better targets, and fail faster on bad ones, if you map biology at industrial scale before committing to a drug.

How phenomics works

The core technique is called phenomics. Robotic labs grow human cells, switch off individual genes using CRISPR or expose the cells to chemical compounds, and then photograph them under a microscope. A change in a gene or a drug leaves a visual “fingerprint” on the cell’s shape and structure. Machine learning models compare millions of these fingerprints. If knocking out gene A produces the same fingerprint as knocking out gene B, the two may sit in the same biological pathway; if a compound reverses the fingerprint of a disease gene, it may be a drug candidate.

Recursion packages these comparisons as “maps.” Two neuroscience maps built for Roche and Genentech give a sense of the scale: the microglia map used more than 100 billion lab-grown immune cells from the brain and contains about 46 million images, while the neuron map drew on a pool of 1 trillion lab-grown neurons.

The compute layer

Models of this size need serious hardware. Recursion trains most of them on BioHive-2, its own supercomputer built in collaboration with NVIDIA from DGX H100 systems. When it was completed in 2024 it ranked 35th on the TOP500 list of the world’s fastest supercomputers. Recursion also builds foundation models, AI models trained on large general datasets, including a transcriptomics model called TxFM that reads gene expression data.

Differentiator: proprietary wet-lab data generated in-house, rather than public data any competitor can download
Feedback loop: model predictions are tested back in the lab, and the results become new training data
Chemistry layer: Exscientia’s precision design tools now help turn targets into optimised molecules
Data partnerships: licensed de-identified patient data from Tempus AI connects lab results to real patient biology
Known limitation: a cell in a dish is not a patient. The platform can improve target choice and speed, but it cannot remove the risk that a drug fails in humans

Roadmap: from maps to medicines

The milestone that matters most is not a bigger map. It is a platform-derived drug succeeding in a controlled human trial, and 2026 to 2027 is the first period in which several such tests arrive close together.

Automated robotic laboratory with rows of cell culture plates and a wall of microscope images feeding a supercomputer — Recursion’s labs can run more than 2 million experiments a week, feeding models trained on its BioHive-2 supercomputer.

4. Business Model and Revenue Engine

Recursion runs two engines at once. The first is partnerships: large drugmakers pay for access to its platform and maps, then pay milestones as programmes advance, and royalties if a drug reaches the market. The second is its own pipeline, where Recursion keeps the full economics but also pays the full cost.

Partnership revenue is lumpy by nature. It is recognised as work is performed and milestones are hit, so a quarter with no milestone can look sharply weaker than one with a map acceptance. That is exactly what happened in Q2 2026, when revenue fell to $7.7 million from $19.2 million a year earlier because some Roche and Genentech project phases had finished in the prior period.

The partnership book

The numbers below come from Recursion’s own Q2 2026 results release and its SEC filings.

Partner	Received to date	Potential per programme	Status
Roche and Genentech	$216 million	Over $300 million, up to 40 programmes	Active; first Validated Target Option exercised in 2026
Sanofi	$134 million	Up to $343 million, up to 15 programmes	Active; fifth milestone ($4 million) in February 2026
Merck KGaA	$20.1 million upfront	Up to $73.0 million per project	Active since September 2023
Bayer	Not separately disclosed	Up to $1.5 billion in total	Active; precision oncology, amended November 2023

Source: Recursion Q2 2026 results, Form 10-Q and 2025 Form 10-K. Potential milestones are not guaranteed.

Across all partnerships, Recursion says it has collected more than $500 million in upfront and milestone payments and has potential for more than $20 billion in further milestones before royalties. Investors should treat the second figure as a theoretical ceiling. It assumes every optional programme is started and succeeds, which never happens in drug development.

The most meaningful recent signal is Genentech exercising its first Validated Target Option in 2026, moving a neuroscience target identified with Recursion into early discovery. No fee was disclosed. It matters less for cash than for what it shows: a partner with its own large research organisation chose to advance a target from Recursion’s maps.

Upfront and milestone income: the main revenue line, and the most variable
Royalties: potentially valuable, but only after a partnered drug is approved, which is years away
Data licensing: Tempus extended its data partnership through November 2029 on 21 September 2026, replacing discretionary fees of up to $84 million over two years with $42 million of committed payments over three years, and took a non-exclusive licence to TxFM
Wholly owned pipeline: no revenue yet, but full value capture if a drug succeeds
Moat: proprietary data and automation that take years and large capital to replicate

The NVIDIA correction

The material behind this article described NVIDIA as a current investor. That is no longer accurate. NVIDIA bought about 7.7 million Recursion shares at $6.48 in July 2023, a $50 million investment. Its 13F filing published in February 2026 showed it had sold the entire position during the fourth quarter of 2025. The technology relationship behind BioHive-2 remains, but NVIDIA is not a shareholder, and investors should not treat it as a source of financial support.

5. Financial Analysis and Growth Outlook

Metric	Value
Share Price (9 Oct 2026 close)	$4.36
Market Cap	About $2.3 billion
52-Week Range	$2.77 to $7.18
Q2 2026 Revenue	$7.7 million (Q2 2025: $19.2 million)
Q2 2026 R&D Expense	$89.6 million (Q2 2025: $128.6 million)
Q2 2026 Net Loss (GAAP)	$131.0 million, or $0.25 per share
Q2 2026 Operating Cash Outflow	$105.9 million
Cash, Equivalents and Restricted Cash	$556.8 million (30 June 2026)
Debt and Finance Leases	$14.2 million
Cash Runway (company guidance)	Into early 2028
Analyst Ratings	3 Buy, 4 Hold (7 analysts)

Data as of 9 October 2026. Market cap uses 535.3 million Class A shares outstanding at 31 July 2026. Always verify current figures before publishing or investing.

Revenue is the least important line for Recursion today. The company is a clinical-stage biotech, and its value rests on whether its pipeline and partnerships produce approved drugs. The more useful lens is cost discipline and cash. Here the trend is genuinely improving: R&D expense fell 30% year over year in Q2 2026, G&A fell to $41.5 million from $46.7 million, and the net loss narrowed to $131.0 million from $171.9 million.

Management also cut full-year 2026 cash operating expense guidance to below $375 million, from below $390 million. First-half cash operating expense was $191.0 million. Cash fell from $753.9 million at the end of 2025 to $556.8 million at the end of June, and Recursion says that, with no additional financing, its cash lasts into early 2028.

Testing the runway claim

The brief behind this article treated the 2028 runway as a guarantee. It is a company projection. A simple check: if Recursion spends close to its full-year guidance, the second half of 2026 costs roughly $185 million, leaving somewhere around $370 million entering 2027 before any partner payments. That is consistent with a runway into early 2028 only if spending keeps falling or milestones keep arriving. Neither is certain, which is why the runway date deserves to be watched each quarter rather than taken as settled.

Recursion also has a $300 million at-the-market equity programme with TD Cowen, set up in February 2026. It sold no shares under it in the first half of the year. The facility does not mean dilution is coming, but it means the mechanism to issue stock quickly is already in place.

Q2 2026 result: revenue down 60% year over year on lumpy partner income; losses and R&D spending narrower
Analyst view: sources tracking seven analysts show three Buy and four Hold ratings, with average price targets of roughly $7 to $8 and a range of $4 to $10
Near-term catalysts (0 to 12 months): updated REC-4881 data on 2 November 2026; REC-4881 FDA registrational path update in 2H 2026; more REC-1245 dose-escalation data in 2H 2026; start of the REC-7735 Phase 1/2 ZINNIA trial in 2H 2026; possible Sanofi milestones over the next 6 to 12 months
2027 catalysts: early REC-617 and REC-3565 safety data in 1H 2027; more REC-4881 data in 1H 2027; REC-4539 data in 2H 2027
Key financial risk: any delay or setback that pushes spending back up before the next large milestone

Financial dashboard showing a gold cash balance bar shrinking over time beside a red burn rate line on a black background — Recursion held $556.8 million of cash at mid-2026 and guides to a runway into early 2028.

6. The Pipeline: Where the Platform Meets Patients

For a cornerstone view, the pipeline needs more than a list. Each programme tests a different part of the platform’s promise.

REC-4881: the lead programme

REC-4881 is a MEK1/2 inhibitor being developed for familial adenomatous polyposis (FAP), an inherited condition in which hundreds of polyps grow in the colon and the lifetime risk of colorectal cancer is very high without surgery. There is no approved drug treatment, which is why the FDA has granted REC-4881 both Orphan Drug and Fast Track designations.

In the Phase 1b/2 TUPELO trial, Recursion reported a median 43% reduction in polyp burden at week 13, deepening to 53% at week 25 after a treatment break, with 40% of patients improving on the Spigelman stage, a standard severity score. Grade 3 side effects occurred in 15.8% of patients, with no Grade 4 or higher treatment-emergent events. Discussions with the FDA began in the first half of 2026, and updated data are scheduled for the CGA-IGC meeting on 2 November 2026.

Why it matters: a small rare-disease market, but a clear unmet need and a potential first approval for the company
What to watch: durability of the polyp reduction, tolerability over longer dosing, and whether the FDA agrees a registrational design

The oncology follow-ons

REC-1245 (RBM39 degrader, solid tumours): early Phase 1 data in 16 patients showed no dose-limiting toxicities and mostly Grade 1 to 2 treatment-related side effects; more dose-escalation data due in 2H 2026
REC-7735 (PI3Kα H1047R inhibitor): IND cleared, designed for more than 100-fold selectivity for the mutant form over the normal enzyme; Phase 1/2 ZINNIA study expected to start in 2H 2026
REC-617 (CDK7 inhibitor): early Phase 1 combination safety and pharmacokinetic data expected in 1H 2027
REC-3565 (MALT1 inhibitor): early Phase 1 monotherapy safety data expected in 1H 2027
REC-4539 (LSD1 inhibitor): early Phase 1 monotherapy safety data expected in 2H 2027

The base rate investors must respect

AI does not change the statistics of human trials overnight. A widely cited MIT analysis of industry-sponsored drugs estimated that only about 13.8% of candidates entering Phase 1 go on to FDA approval, and the figure for oncology was just 3.4%. Most of Recursion’s pipeline is in oncology. Early safety data, which is what most of the 2027 readouts are, tell investors whether a drug is tolerable, not whether it works.

Clinical trial data screens with green and red result markers in a dark control room above a timeline of upcoming milestones — Updated REC-4881 data on 2 November 2026 is the first of several pipeline tests.

7. Industry Overview and Competitive Landscape

Developing a new drug takes many years and costs billions of dollars, and most candidates fail. AI drug discovery promises to help by choosing better targets and designing better molecules faster. The open question is which business models capture that value.

The market’s answer so far has favoured companies with revenue today. Over the past twelve months to 9 October 2026, Schrödinger (SDGR) rose about 43%, while Recursion fell about 18%. The divergence story is not as clean as it first appears, though. Tempus AI (TEM) is up about 20% year to date but down about 26% over twelve months, so “every AI healthcare stock rallied except RXRX” is not accurate.

Company	Model	9 Oct 2026 Close	12-Month Change	2026 YTD
Recursion (RXRX)	Platform plus own pipeline	$4.36	About -18%	About +7%
Schrödinger (SDGR)	Physics-based software plus drug pipeline	$27.28	About +43%	About +53%
Tempus AI (TEM)	Diagnostics and clinical data	$70.93	About -26%	About +20%

Price changes calculated from daily closing prices. Data as of 9 October 2026.

Schrödinger: sells computational chemistry software with recurring revenue, which gives it a commercial base Recursion does not have
Tempus AI: a diagnostics and data company; it is also a Recursion partner and data supplier, not only a competitor
Large pharma in-house AI: the quiet competitor; every major drugmaker is building its own capability, which can reduce demand for outside platforms
Recursion’s position: the deepest proprietary wet-lab data set among listed peers, but the least revenue, and the most dependent on clinical outcomes

8. Institutional Ownership: What ARK’s Stake Does and Does Not Mean

Cathie Wood’s ARK Invest remains one of the most visible Recursion holders. According to ARK’s own daily holdings file dated 9 October 2026, the ARK Genomic Revolution ETF (ARKG) holds 14,727,265 RXRX shares, a 2.84% weight in the fund. That is a larger share count but a lower weight than the figures circulating in some recent coverage, which reflects both ARK’s buying and the falling share price.

A large ETF position is evidence of conviction, not a valuation floor. ARK funds trade frequently, driven by fund flows and the manager’s thesis rather than Recursion’s clinical results. Insider selling also continues at a modest pace: director Blake Borgeson sold 50,000 shares at a weighted average of $4.84 on 6 October 2026 under a pre-arranged trading plan adopted in August 2025.

9. Risk Analysis: Where the Thesis Can Break

At Recursion the risks are not hidden. They are the main story.

Clinical Risk

REC-4881 is the most advanced programme; a weak 2 November update or an unfavourable FDA path would hit both the asset and the platform narrative
Most other programmes are early-stage oncology, where historical success rates are lowest

Financial Risk

Operating cash outflow of $105.9 million in Q2 2026 against $7.7 million of revenue
A $300 million at-the-market programme is in place; issuing shares near current prices would dilute existing holders

Partnership Risk

Partner programmes are optional; partners can decline to advance targets, and milestones are paid only on success
Revenue can fall sharply in quarters without milestones, as Q2 2026 showed

Technology Risk

Better target selection in cells may not translate into better results in people; this is the core claim the platform has yet to prove clinically

Competitive and Sentiment Risk

Large drugmakers building their own AI capabilities could reduce demand for platform deals
NVIDIA’s exit from the shareholder register removed a widely cited source of credibility, and the stock remains heavily traded on sentiment

10. Bull vs Bear Case

🐂 Bull Case

More than $500 million already collected from partners shows the platform has commercial value before any approval
Genentech exercised its first Validated Target Option in 2026, a concrete sign a top-tier partner values targets from the maps
REC-4881 has encouraging Phase 2 data, Orphan Drug and Fast Track designations, and active FDA discussions in a disease with no approved drug
Costs are falling: R&D down 30% year over year and 2026 cash opex guidance cut to below $375 million
The Tempus deal converts discretionary fees into $42 million of committed payments and extends the data partnership to 2029

🐻 Bear Case

Cash burn of about $100 million a quarter against revenue under $8 million leaves little margin for setbacks
The early 2028 runway is a projection that depends on continued cost cuts and partner payments
Most of the pipeline is early-stage oncology, the therapeutic area with the lowest historical approval rate
NVIDIA sold its entire stake in Q4 2025, and the material advantage of the platform has not yet been proven in a late-stage trial
An at-the-market facility of $300 million gives the company a ready route to dilution if the stock stays low

11. Valuation: Multiple Perspectives

Recursion has no earnings and little revenue, so standard multiples such as P/E are meaningless. Three other lenses are more useful.

Enterprise value versus cash: with a market cap of about $2.3 billion and about $557 million of cash at mid-year against $14.2 million of debt, the market is assigning roughly $1.8 billion to the platform, the pipeline and the partnerships
Versus partner payments: that figure is roughly three and a half times the more than $500 million partners have already paid, which suggests the market expects meaningful but not exceptional future success
Versus cash burn: at around $100 million of operating cash outflow a quarter, the current cash balance covers roughly five to six quarters at the Q2 pace, which is why cost cuts and milestones matter so much to the runway guidance

None of these lenses says the stock is cheap or expensive on its own. They say that Recursion’s valuation is mostly a bet on the clinical results due over the next 18 months, which is exactly where the uncertainty is greatest.

A note on the chart

The brief behind this article proposed a technical trading plan with buy zones and breakout levels. OneMoreStock does not publish trading instructions. For context only: the stock’s 250-day moving average, based on daily closing prices, sits at about $3.85, and shares have traded between roughly $3.16 and $4.78 over the past month. Single-session moves of 7% to 16% happened several times in the past month, which is a reminder of how much of the price is driven by sentiment rather than fundamentals.

12. Investment Outlook: A Long-Duration, High-Risk Platform Bet

Recursion is a long-duration bet on whether industrial-scale biology can produce better medicines. It suits only investors with a horizon of several years who can watch the shares fall sharply on a single trial update without being forced to sell.

The thesis changes in the bull direction if REC-4881 shows durable benefit and a clear FDA path, if partners keep exercising options, and if the runway extends rather than shrinks. It changes in the bear direction if the lead programme stalls, if partner revenue stays weak, or if the company turns to its at-the-market facility at low prices.

Bull scenario: REC-4881 moves toward a registrational study, partner milestones accelerate, and the market begins to value the platform on clinical evidence
Base scenario: mixed readouts, steady cost discipline, and a stock that remains volatile and catalyst-driven while the 2027 data arrive
Bear scenario: a disappointing REC-4881 update or FDA outcome, followed by dilutive financing and a weaker platform narrative
Catalysts to monitor: 2 November 2026 REC-4881 data; FDA path update in 2H 2026; Q3 2026 results and runway update; Sanofi milestones; 2027 Phase 1 safety readouts

13. Final Thoughts

Recursion matters because it is one of the clearest public tests of whether AI can change the economics of drug discovery. Its biggest opportunity is that a platform which has already earned more than $500 million from partners produces its first drug that clearly works in patients. Its biggest risk is the opposite: that the data advantage does not show up in the clinic, while cash keeps falling.

The long-term outlook depends on evidence that will arrive over the next 18 months, not on narrative. Investors who follow the company should judge it on trial results, partner decisions and the runway each quarter. For ongoing coverage of AI and healthcare stocks, subscribe to OneMoreStock on YouTube at www.youtube.com/@onemorestock.

FAQ

Is Recursion Pharmaceuticals stock a good buy right now?

That depends on your risk tolerance and time horizon. Recursion is a clinical-stage biotech with about $557 million of cash, a runway guided into early 2028, and key clinical updates due from November 2026. It suits only investors who can accept large losses if trials disappoint.

Is Recursion profitable?

No. Recursion reported a GAAP net loss of $131.0 million in Q2 2026 on revenue of $7.7 million. Its income comes mainly from partner payments, which vary sharply from quarter to quarter.

Does NVIDIA still own Recursion shares?

No. NVIDIA invested $50 million in Recursion in July 2023, buying about 7.7 million shares, but its 13F filing published in February 2026 showed it had sold the whole stake in the fourth quarter of 2025. Recursion’s BioHive-2 supercomputer was built in collaboration with NVIDIA.

Does Cathie Wood’s ARK own Recursion?

Yes. ARK’s daily holdings file for 9 October 2026 shows the ARK Genomic Revolution ETF (ARKG) holding 14,727,265 Recursion shares, a 2.84% weight. ARK positions change frequently, so check the current file before relying on it.

What is Recursion’s next major catalyst?

Updated data for REC-4881, its drug for familial adenomatous polyposis, is scheduled for the CGA-IGC meeting on 2 November 2026. The company also expects an update on the drug’s registrational path with the FDA in the second half of 2026.

How does Recursion compare with Schrödinger and Tempus AI?

Schrödinger sells computational chemistry software with recurring revenue, and Tempus AI is a diagnostics and data business. Recursion has less revenue than either and depends more on its own clinical results. Tempus is also a Recursion partner, having extended a data licence agreement through November 2029.