Isomorphic Labs, Recursion, and Insilico Medicine are the three most-watched companies betting that AI can fix pharma's brutal economics — a $2.6 billion, 15-year process where nine out of ten clinical candidates fail — and by 2026 more than 200 AI-discovered drugs have entered clinical trials, but the number that matters most is still zero: no AI-discovered drug has received FDA approval. That's not a verdict against the technology. It's the honest starting point for understanding what each of these three companies has actually proven so far, versus what they're still promising.
The Honest Framing: Real Discovery Wins, an Unproven Clinic
The most useful way to read AI drug discovery in 2026 is as a field that's cleanly split in two. On one side, AI has genuinely delivered: molecules designed computationally are reaching human trials faster and cheaper than traditional discovery timelines, in some documented cases compressing a process that normally takes 6 to 8 years down to under 30 months. On the other side sits the part AI hasn't changed: getting through Phase 2 and Phase 3 trials still depends on the same underlying biology, the same regulatory scrutiny, and the same odds of failure that have always governed drug development. AI speeds up finding a promising molecule. It doesn't yet shorten proving that molecule actually works and is safe in people.
Isomorphic Labs: The Best-Funded, Least Clinically Proven
Isomorphic Labs is Alphabet's dedicated drug discovery subsidiary, built on DeepMind's AlphaFold protein-structure prediction technology, and it leads the category on one specific dimension: validation through partnership capital, with roughly $3 billion in combined deals across Eli Lilly and Novartis. The company has also pushed the underlying science forward directly, releasing an AI system that doubles AlphaFold 3's prediction accuracy. Isomorphic's stated strategy is to advance internal oncology and immunology programs through early-stage development, then license them out once they clear initial human trials. As of 2025 and into 2026, the company's own executives describe lead candidates as being at the cusp of first-in-human testing — meaning Isomorphic remains the most capital-backed of the three companies covered here, and also the one furthest from having any actual clinical trial results to point to yet.
Insilico Medicine: The Speed and Clinical Proof Leader
Insilico Medicine holds the most clinically advanced pipeline of any AI drug discovery company, and its flagship result is the strongest clinical proof point the field has produced so far. Its lead candidate, INS018_055 (also known as rentosertib), an AI-discovered and AI-designed treatment for idiopathic pulmonary fibrosis, moved from initial target identification to Phase 2 trials in under 30 months — versus the traditional 6-to-8-year timeline for comparable programs, and it was widely described as the first fully AI-designed drug to enter human testing back in 2022. The Phase IIa results, published in the peer-reviewed journal Nature Medicine, showed a genuine positive signal: patients on the 60mg daily dose saw lung function improve by a mean of 98.4 mL in forced vital capacity, compared to a 20.3 mL decline in the placebo group. That's a real, published clinical result — the field's first end-to-end demonstration that an AI-originated molecule can produce a measurable therapeutic effect in actual patients, not just a computational prediction.
Recursion: Scale Plus Discipline
Recursion operates at a different scale than either competitor, running what it calls Recursion OS, a platform generating over 100 million candidate molecules annually by combining cellular imaging, multi-omics data, and AI-driven chemistry. After merging with Exscientia in 2024, Recursion became the largest publicly listed AI drug discovery company by pipeline breadth. What's notable about Recursion's 2025-2026 trajectory isn't unchecked expansion — it's discipline: the company trimmed weaker programs from its pipeline in 2025 rather than continuing to fund everything simultaneously, and guided investors toward roughly 10 clinical readouts across an 18-month window, including a Phase II readout for its ALDER program expected around the first quarter of 2026. Industry analysts have specifically flagged this pipeline discipline, cutting programs rather than claiming universal success, as a healthier signal than a company insisting every bet is working.
So Which Company Is Actually Winning?
It depends entirely on which metric you weight. On capital and pharma-partnership validation, Isomorphic Labs leads clearly, backed by Alphabet and roughly $3 billion in deals with two major pharmaceutical companies. On actual clinical proof that an AI-designed molecule works in patients, Insilico Medicine leads decisively, with a published, peer-reviewed positive Phase IIa result already in hand. On sheer platform scale and pipeline breadth, Recursion is the largest operation by molecule volume and program count, tempered by its own 2025 pipeline cuts. None of the three has an FDA-approved drug yet, and industry analysts consistently point to 2028-2030 as the most credible window for the field's first AI-derived approval, contingent on how the current wave of Phase 2 and Phase 3 readouts actually lands.
Head-to-Head Comparison
| Company | Core technology | Clinical status (2026) | Standout metric |
|---|---|---|---|
| Isomorphic Labs | AlphaFold-based protein structure prediction (Alphabet/DeepMind) | Lead candidates approaching first-in-human trials | ~$3 billion in Eli Lilly and Novartis partnership deals |
| Insilico Medicine | Pharma.AI generative chemistry and target discovery engine | Positive Phase IIa results published in Nature Medicine | Target-to-Phase-2 in under 30 months (vs 6-8 years typical) |
| Recursion | Recursion OS: cellular imaging + multi-omics + AI chemistry | ~10 clinical readouts guided across an 18-month window | 100+ million candidate molecules generated annually |
Frequently Asked Questions
Has any AI-discovered drug actually been approved by the FDA?
No. As of 2026, more than 200 AI-discovered drugs are in clinical trials across the industry, but zero have received FDA approval. Industry analysts consider 2028-2030 the most credible window for the field's first AI-derived approval.
Which company has the strongest clinical evidence so far?
Insilico Medicine. Its lead candidate, rentosertib, produced positive Phase IIa results published in Nature Medicine, showing measurable lung function improvement in idiopathic pulmonary fibrosis patients compared to placebo, the field's first published end-to-end clinical validation.
Why is Isomorphic Labs considered well-funded but clinically unproven?
Isomorphic Labs has secured roughly $3 billion in partnership deals with Eli Lilly and Novartis and is backed by Alphabet, but as of 2026 its lead oncology and immunology candidates are still approaching first-in-human trials rather than having published clinical results.
Does AI actually speed up drug development?
Yes, specifically in the discovery phase. Insilico Medicine moved a candidate from target identification to Phase 2 trials in under 30 months, versus a traditional 6-to-8-year timeline. However, AI has not been shown to shorten Phase 2 or Phase 3 clinical trials themselves, which still depend on the same biology and regulatory review as traditionally discovered drugs.