AI Drug Discovery Finally Has a Scoreboard. It Doesn't Reward the Algorithm.

In September, at Peking Union Medical College Hospital in Beijing, a patient with idiopathic pulmonary fibrosis swallowed a small-molecule pill called rentosertib — and the drug industry quietly crossed a line it has been promising to cross for a decade. Rentosertib, developed by Insilico Medicine, is what the company and industry trackers describe as the first drug discovered and designed with generative AI to enter a Phase 3 trial: both the biological target and the molecule itself came out of Insilico's AI platform. The GENESIS-IPF-3 study will enroll 320 patients across 47 centers in China over 52 weeks.

That is, by any standard, the milestone the entire AI-drug-discovery field has been waiting for. So here is the uncomfortable part: Insilico's Hong Kong-listed shares trade at less than half their 52-week high. The year the field finally produced its landmark was also the year the market repriced almost everyone in it.

That contradiction is the story. AI drug discovery in 2026 is no longer a pitch-deck category — it has public tickers, clinical readouts, IPO prices, and now a scoreboard. And the scoreboard is delivering a verdict most of the sector's marketing never prepared investors for.

The verdict era arrived fast

Three IPOs in eleven months turned a private hype cycle into a public market.

Insilico Medicine listed in Hong Kong in December 2025, raising roughly $293 million in what was the city's largest biotech IPO of the year. Generate:Biomedicines, the Flagship Pioneering-backed "generative biology" company, raised $400 million on Nasdaq in February 2026 at $16 per share — described at the time as the largest biotech IPO of the young year. And in October, Nvidia-backed Iambic Therapeutics priced 9.4 million shares at $16, raised $150 million, and popped 16% on its Nasdaq debut.

The money arriving from outside biotech is bigger still. Alphabet's Isomorphic Labs — the Google DeepMind spinout built on AlphaFold — has research deals with Eli Lilly and Novartis worth a reported $3 billion in potential milestones, and this year closed one of the largest private biotech funding rounds on record, reported at $2.1 billion. In October, OpenAI acquired Orchestra Bio, a startup building AI agents that help pharma companies manage drug development, folding it into Rosalind, its life-sciences tools unit. The two best-funded AI labs on earth have both now planted flags in pharma.

A decade of "AI will fix drug discovery" rhetoric is finally being tested the only way the industry tests anything: in randomized, controlled human trials, with public share prices keeping score in real time.

The market changed the question

For most of the past five years, the question investors asked these companies was: how good is your model? Bigger foundation models, better protein-structure prediction, more automated labs — all of it was fundable.

In 2026, the question changed to the one that has governed biotech forever: what did your molecule do in humans?

The cautionary tale is Neumora Therapeutics, which went public pitching machine-learning-driven "precision neuroscience." In June, its lead antidepressant navacaprant failed its second and third Phase 3 trials — missing both primary and key secondary endpoints — completing an 0-for-3 sweep. The company discontinued the drug and cut roughly 35% of its staff. Computational sophistication, it turns out, does not repeal clinical risk: the industry's roughly 90% clinical failure rate was always the thing AI was supposed to fix, and so far the sample says it hasn't fixed it yet.

The market has drawn its conclusion with brutal consistency. Both of 2026's flagship AI-biotech IPOs now trade below their $16 offer prices. Recursion Pharmaceuticals, the sector's original platform story and a $502 million IPO in 2021, trades at a fraction of the valuation it listed at. Even Insilico, holder of the field's crown-jewel milestone, has been cut in half from its post-IPO peak.

And yet, inside the same sector, a handful of names are up 200% or more over the past twelve months. The dispersion is the message: the market is no longer paying for AI platforms. It is paying for something much more specific — and the companies that have it share a pattern that is easy to check and widely misunderstood.

Who is actually getting paid, what separates them from the companies getting repriced, and which readouts between now and 2028 decide the next leg?

A free membership gets you access to every free article at AlphaBriefing.com: Join AlphaBriefing

The rest of this briefing is free — it just requires a free AlphaBriefing account: the full winners-and-losers scoreboard with live prices, the single trait the 200% gainers share, and the catalyst calendar from Absci's year-end readouts to the GENESIS-IPF-3 endpoint.

Create your free account → — 30 seconds, no card.