AI-Designed Drugs Are Now in Human Trials — The First Results Will Make or Break a $50 Billion Industry

What Happened

AI-designed drug candidates are reaching the clinical stages where the technology’s value will be proven or discredited. As of early 2026, at least 173 AI-discovered programs are in clinical development, spanning oncology, fibrosis, rare diseases, and neurology. Oncology dominates with 126 active clinical studies.

The most-watched candidate is Insilico Medicine’s rentosertib, formerly known as ISM001-055. In November 2024, Insilico announced positive Phase IIa topline results for the drug in idiopathic pulmonary fibrosis (IPF). The double-blind, placebo-controlled trial enrolled 71 patients across 21 sites in China, and patients on the highest dose (60 mg once daily) showed a 3.05% mean improvement in percent predicted forced vital capacity, compared to a 1.84% mean decline in the placebo group.

Rentosertib holds a specific distinction: both its disease target (TNIK) and the molecular structure were identified and designed using generative AI, with no human hypothesis guiding either step. Insilico CEO Alex Zhavoronkov has said the company is pursuing regulatory discussions for a pivotal Phase III trial, and in early 2026 Harvard Business School published a case study on the drug’s development.

Why It Matters

Traditional drug discovery takes 10 to 15 years and costs an average of $2.6 billion per approved compound. AI platforms promise to compress that timeline to as little as 18 months for lead identification and reduce costs by 30% to 50%. But these claims have largely been theoretical. The 2026 clinical readouts are the first real-world stress tests.

The AI drug discovery market is projected to reach between $5.1 billion and $24.5 billion in 2026, depending on the estimate, with growth rates ranging from 11% to 27% CAGR. Major pharmaceutical companies are placing large bets: Eli Lilly and Nvidia announced a $1 billion joint investment in AI-driven drug development in January 2026.

If Phase III trials confirm the earlier-phase results, it would validate the entire techbio thesis. If they fail, investor confidence in a sector that has already burned through billions in venture capital will take a serious hit.

Technical Details

The leading AI-discovered candidates use different computational approaches:

Insilico Medicine used its proprietary Chemistry42 platform, a generative AI system that designs novel molecular structures optimized for target binding, drug-likeness, and synthetic accessibility. The platform identified TNIK as a novel target for IPF using its PandaOmics disease-modeling engine, then generated the rentosertib molecule from scratch. The Phase IIa results, published in Nature Medicine, showed dose-dependent efficacy with a favorable safety profile.

Relay Therapeutics uses motion-based drug design — computational simulations of protein dynamics to identify allosteric binding sites that traditional structure-based methods miss. Its lead candidate, zovegalisib (RLY-2608), is a first-in-class mutant-selective allosteric PI3Kα inhibitor for HR+/HER2- breast cancer. Updated data showed 11-month median progression-free survival in second-line patients, with a 39% objective response rate. Relay has launched a Phase III trial (ReDiscover-2) pitting zovegalisib against AstraZeneca’s capivasertib.

Recursion Pharmaceuticals takes a different approach entirely, using high-throughput cellular imaging combined with deep learning to map biological relationships at scale. Its CDK7 inhibitor REC-617 is in a Phase I ELUCIDATE trial across multiple solid tumor types, with 29 patients treated across six dose levels showing encouraging tolerability. Full data is expected at a medical conference in 2026.

Who’s Affected

Patients with idiopathic pulmonary fibrosis — a progressive lung disease with limited treatment options — stand to benefit most immediately if rentosertib’s Phase III trial succeeds. The current standard of care (nintedanib and pirfenidone) slows disease progression but does not reverse it.

In oncology, breast cancer patients with PI3Kα mutations who have progressed on CDK4/6 inhibitors are the target population for Relay’s zovegalisib. The existing PI3K inhibitor class has been hampered by toxicity, particularly hyperinsulinemia, which zovegalisib’s allosteric mechanism is designed to avoid.

Pharmaceutical companies and biotech investors face a binary outcome. Positive Phase III readouts would likely trigger a wave of M&A activity, licensing deals, and further investment. The FDA’s forthcoming final guidance on AI in drug development — expected in Q2 2026 — will also clarify regulatory expectations for the next generation of AI-discovered candidates.

What’s Next

The next 12 to 18 months will be decisive. Insilico Medicine is expected to begin its pivotal Phase III trial for rentosertib in IPF, likely enrolling patients across multiple geographies. Relay Therapeutics’ Phase III ReDiscover-2 trial for zovegalisib is recruiting, with interim data anticipated in late 2026 or early 2027.

Recursion has guided investors to expect seven clinical readouts within approximately 18 months, and has expanded its AI capabilities to optimize clinical trial design itself — not just molecule discovery.

The FDA and EMA’s joint ten principles for AI in medicine development provide the first international regulatory alignment on the topic. The final U.S. guidance will introduce a tiered risk framework: low-risk AI use cases (hypothesis generation) will require minimal documentation, while high-risk applications (influencing regulatory decisions) will demand full transparency and prospective validation.

Whether AI-designed drugs succeed or fail in these trials, 2026 will produce the first definitive clinical evidence on whether artificial intelligence can fundamentally change how medicines are discovered.