AI Drug Discovery Stocks: Investing in the Biotech Revolution of 2026

The State of AI Drug Discovery in 2026: Beyond the Hype

Drug discovery is one of the hardest problems in science. The average cost to bring a new drug to market exceeds $2.6 billion (Tufts CSDD). The average timeline from target identification to FDA approval is 12–15 years. The clinical trial success rate — from Phase 1 entry to approval — is approximately 7.9% across all therapeutic areas. These economics have driven big pharma R&D productivity into secular decline for three decades: despite R&D spending increasing from $50 billion in 2000 to over $250 billion in 2025, the number of new molecular entities approved annually has barely increased.

AI promises to break this pattern. Not by making clinical trials succeed more often — that remains unproven — but by dramatically reducing the time and cost of finding promising drug candidates, identifying the right patient populations, and designing more efficient clinical trials.

And unlike 2021, when AI drug discovery was mostly PowerPoint presentations and aspirational timelines, 2026 offers real clinical data points. Insilico Medicine's ISM001-055 (idiopathic pulmonary fibrosis) became the first fully AI-discovered drug to enter Phase 2. Relay's RLY-2608 produced a 63% response rate in a heavily pretreated breast cancer population. Recursion has multiple programs in clinical development and acquired Exscientia in a $688 million deal that consolidated the two largest AI drug discovery platforms.

The AI drug discovery sector is no longer pre-clinical. It is pre-proof. And that distinction matters enormously for how investors should think about position sizing, catalysts, and risk management.

Recursion Pharmaceuticals: The Platform Play

Recursion (RXRX) is building what we believe is the most ambitious AI drug discovery platform in existence. The company's core insight is that traditional drug discovery starts with a known target and searches for molecules that interact with it — a needle-in-a-haystack approach. Recursion inverts this. It uses high-throughput automated labs to systematically perturb human cells with genetic knockdowns, small molecules, and other interventions, then captures images of the resulting phenotypic changes using automated microscopy.

The result is a massive biological dataset — over 50 petabytes, the largest of its kind in the world — that maps how different perturbations affect cell biology. Machine learning models then mine this data to discover relationships between genes, diseases, and potential drug candidates. Instead of starting with a hypothesis and testing it, Recursion generates millions of data points and lets the AI find the patterns.

The Pipeline

REC-994 is Recursion's most advanced wholly-owned program, a small molecule in Phase 2 for cerebral cavernous malformations (CCM), a rare disease affecting approximately 300,000 Americans. The drug was identified through Recursion's phenomics platform by screening existing compounds against cellular models of CCM. Phase 2 results are expected in the first half of 2027, with interim data readouts likely in late 2026.

REC-3964, a Phase 1 candidate for familial adenomatous polyposis (FAP), targets a rare hereditary condition that leads to colorectal cancer. This program demonstrates Recursion's strategy of pursuing rare diseases where smaller trials, lower competition, and orphan drug pricing can generate attractive economics even with modest patient populations.

The Exscientia acquisition, completed in mid-2025 for $688 million, was transformative. Exscientia brought AI-driven chemistry capabilities (complementing Recursion's biology-focused platform), multiple clinical-stage assets including programs in oncology and immunology, and validated pharma partnerships with Sanofi, Bristol-Myers Squibb, and Bayer. The combined entity has over 10 programs in clinical or IND-enabling stages — the largest clinical pipeline of any pure-play AI drug discovery company.

The Financials

Recursion reported $55 million in revenue for 2025, primarily from the Roche-Genentech partnership ($150 million upfront signed in 2023, with up to $12 billion in potential milestones) and the Bayer relationship (focused on oncology). Operating expenses were approximately $520 million, resulting in a net loss of roughly $465 million. Cash and equivalents stood at $740 million at year-end 2025, providing runway through 2028 at current burn rates.

The market cap sits at approximately $2 billion. To justify that valuation, Recursion needs to demonstrate one of three things: clinical proof-of-concept for an AI-discovered drug (Phase 2 success), a major new pharma partnership that validates the platform's economics, or consistent progression of multiple pipeline programs toward value-inflecting data readouts. We believe the Roche partnership and expanding pipeline provide enough near-term catalysts to justify a 1–3% portfolio position.

Relay Therapeutics: The Precision Medicine Pioneer

Relay Therapeutics (RLAY) takes a fundamentally different approach to AI-driven drug discovery. Where Recursion screens millions of cells phenotypically, Relay uses molecular dynamics simulations — essentially physics-based modeling — to understand how proteins move at atomic resolution. This matters because proteins are not static crystal structures (which is how the pharmaceutical industry traditionally studies them). They are dynamic machines that adopt different conformations, and drugs that target specific conformational states can achieve selectivity that is impossible with traditional approaches.

The flagship demonstration is RLY-2608, a mutant-selective PI3Kα inhibitor for HR+/HER2- breast cancer with PIK3CA mutations. PI3Kα has been one of the most important and most difficult oncology drug targets for two decades. Previous PI3K inhibitors (like Novartis's alpelisib) work but cause severe side effects because they inhibit both the mutant and wild-type versions of the protein. Relay used its molecular dynamics platform to design a molecule that selectively targets only the mutant form — leaving the normal protein alone.

The early clinical results are remarkable. In Phase 1/2, RLY-2608 combined with fulvestrant showed a 63% overall response rate in patients with PIK3CA-mutant breast cancer who had failed prior therapies. The safety profile was significantly cleaner than alpelisib, with lower rates of hyperglycemia (the class's dose-limiting toxicity). If these results hold in registrational trials, RLY-2608 could become a best-in-class PI3K inhibitor for a market worth $3–5 billion annually.

Valuation and Risks

Relay's market cap is approximately $3.5 billion with $900 million in cash, giving an enterprise value of roughly $2.6 billion. For a company with a potentially best-in-class drug in a $3–5 billion market, this is not unreasonable if RLY-2608 succeeds in registrational trials. The risk-adjusted NPV of RLY-2608 alone, using standard probability-of-success assumptions (30% for Phase 2 to approval in oncology), supports approximately $2 billion in value.

The downside scenario is straightforward: if RLY-2608 fails in pivotal trials, the stock likely drops 50–60%. The company has additional pipeline programs (RLY-5836 for CDK2 and an undisclosed SHP2 program), but the entire investment case rests on RLY-2608 delivering. This is the essence of clinical-stage biotech risk, and no amount of AI sophistication changes the binary nature of registrational trial outcomes.

Absci Corporation: Generative AI for Biologics

Absci (ABSI) approaches AI drug discovery from an entirely different angle: generative AI for biologic drug design. While Recursion and Relay focus on small molecules, Absci uses large language models — trained on protein sequences rather than text — to design novel antibodies and other biologic drugs from scratch.

The company's Integrated Drug Creation platform combines three capabilities. First, a generative AI engine that designs de novo antibodies with specified binding properties, specificity, and developability characteristics. Second, a wet-lab validation platform that rapidly tests AI-designed molecules using high-throughput assays. Third, a cell line development technology (the SoluPro system) that produces biologic drugs in E. coli rather than the expensive mammalian cell cultures used by most of the industry, potentially reducing manufacturing costs by 70–80%.

In November 2024, Absci published results in Nature Biotechnology demonstrating that its AI could design functional antibodies that bound to clinically validated targets (HER2, VEGF-A) with zero evolutionary starting points — a world first. The model generated antibodies that were not only functional but showed binding affinities comparable to clinically approved drugs, validated through wet-lab experiments.

The Business Model Challenge

Absci's challenge is commercial. The company generates revenue primarily through partnerships — it signed deals with AstraZeneca, EQRx (now acquired), and several undisclosed pharma companies. But annual revenue remains below $20 million and the company burned approximately $110 million in 2025. At a market cap near $700 million, Absci is valued primarily on the optionality of its generative AI platform.

We view Absci as the most speculative of the three companies profiled here. The technology is genuinely differentiated — generative antibody design is a frontier capability that Big Pharma does not yet have internally — but the path from Nature publications to commercial drug revenues is long and uncertain. This is a 0.5–1% position for investors who want exposure to the generative biology thesis without concentrating risk.

Big Pharma's AI Adoption: Competition or Validation?

A common bear argument against AI drug discovery stocks is that Big Pharma will simply build these capabilities internally, eliminating the need for platform companies like Recursion and Absci. The evidence so far suggests the opposite: Big Pharma is partnering with, acquiring, and investing in AI drug discovery companies rather than replicating their platforms from scratch.

Roche signed a $150 million upfront deal with Recursion in 2023, with up to $12 billion in milestones. Sanofi committed $1.2 billion in total deal value to Exscientia (now part of Recursion). AstraZeneca partnered with Absci. Novartis invested over $100 million in its internal AI capabilities but also signed external partnerships. Pfizer hired hundreds of AI scientists and spent billions building internal platforms, yet continued to acquire externally developed programs.

We believe this pattern — build internally but partner externally — will persist for at least the next 5–7 years. The reason is data moats. Recursion's 50 petabytes of phenomics data took a decade and hundreds of millions of dollars to generate. Relay's molecular dynamics simulation platform required years of computational physics development. These are not assets that a pharma company can replicate by hiring an AI team and buying cloud compute.

The bigger risk is not internal pharma AI programs. It is whether AI-discovered drugs actually show higher clinical trial success rates than traditionally discovered drugs. If they do, the platform companies are worth multiples of their current valuations. If they do not, the thesis collapses from “AI revolutionizes drug discovery” to “AI makes preclinical work slightly cheaper.” We are tracking this question closely across every AI biotech earnings transcript and clinical readout, and we detail similar analytical approaches in our coverage of AI-powered biotech pipeline analysis.

How to Build an AI Drug Discovery Portfolio

We believe the correct way to invest in AI drug discovery is a barbell approach that combines platform biotechs with AI-adopting pharma:

• Core (50–60% of AI biotech allocation): Recursion Pharmaceuticals (RXRX) as the broadest platform with the largest pipeline and data moat. Position size: 1–3% of total portfolio.
• High-conviction clinical bet (20–30%): Relay Therapeutics (RLAY) for RLY-2608 upside. Position size: 0.5–2% of total portfolio, with willingness to reduce ahead of binary data events.
• Optionality (10–20%): Absci (ABSI) for generative biology exposure. Position size: 0.5–1% of total portfolio maximum.
• De-risked AI pharma exposure: Roche (RHHBY) and Eli Lilly (LLY) as large-cap pharma companies with the most aggressive AI adoption strategies. These provide AI drug discovery exposure without the binary risk of pre-revenue biotechs.

Total AI drug discovery exposure should not exceed 5–8% of a diversified equity portfolio. This is a sector where the upside is potentially 5–10x on individual names, but the probability of permanent capital loss on any single position is 30–50% over a 5-year horizon. Diversification across platforms and clinical stage is essential.

For investors who prefer a more systematic approach to identifying biotech catalysts and monitoring pipeline progress, our guides on AI-powered quantitative screening and SEC filing analysis provide complementary analytical frameworks.

The 2030 Outlook: What Success (and Failure) Looks Like

By 2030, the AI drug discovery thesis will be resolved one way or another. The sector needs at least one AI-discovered drug to achieve Phase 3 success and FDA approval to prove that the technology produces drugs that are not just faster to discover but also effective in humans. If that happens — and we assign roughly 50–60% probability to this outcome — the sector re-rates dramatically, and companies like Recursion with the broadest platforms and largest data moats will be worth 5–10x their current valuations.

The failure scenario is not that AI proves useless in drug discovery. It is that AI-discovered drugs show the same ~8% clinical success rate as traditionally discovered drugs, meaning the technology only saves time and money in preclinical stages without improving the probability of clinical success. In that world, AI drug discovery platforms are worth their NPV of preclinical cost savings — perhaps $500 million to $1 billion for the largest players — rather than the multi-billion-dollar valuations the market currently assigns.

We believe the intermediate scenario is most likely: AI-discovered drugs will show modestly higher clinical success rates (perhaps 12–15% versus 8% historically) driven by better target selection and more optimized molecules, while dramatically reducing preclinical timelines and costs. This outcome justifies current valuations for the best platforms (Recursion, Relay) but not for speculative names with limited data.

The next 12–18 months will be critical. Phase 2 readouts from Recursion (REC-994), Relay (RLY-2608 expansion data), and Insilico (ISM001-055) will provide the first real evidence on whether AI drug discovery delivers clinically meaningful improvements. Position accordingly.

Frequently Asked Questions