Insilico Is Winning. On Paper.

Insilico Medicine is just on fire. I didn't plan on covering them again this week. I really did try to write about something else, and then... here we are.

Last week it was ISM8969, their NLRP3 candidate, getting its first dose in a person. I figured that bought us a break. But just a couple of days ago, on June 22, they signed SK Biopharmaceuticals to a deal worth up to $2.5 billion.

And SK is just the latest. Eli Lilly's come back three times: a software license, then a research deal, then a $115 million upfront global one in March. Menarini took one molecule, liked it, and came back for a second. Qilu went from just running Insilico's software to a near $120 million partnership. These are companies with their own scientists, and they wouldn't keep coming back if the science didn't hold up. The stock agrees, up about 35% since the December IPO.

Of all their drug candidates, Rentosertib is the one we should keep a close eye on. The target, TNIK, is something Insilico says its models dug up from scratch, and the molecule was theirs to design. Then it went into people: a placebo-controlled Phase 2a in lung fibrosis where the 60 mg dose pushed lung function up 98.4 mL while placebo slid 20.3 mL. An AI drug doing something you can measure in patients is still rare. They've got one.

So. Is it too early to celebrate?

Yeah, I think it is (not trying to be negative here…). The SK headline says $2.5 billion, but the money with a real date on it is up to $18 million. So the part you can bank on is, at most, 0.72% of the part in the headline, and the release doesn't name a target, a molecule, an indication past "neuroimmune CNS," or one preclinical number.

What I really wanted to highlight here are the numbers beyond the headline, which are the cash they've collected. They'll tell you $4.6 billion in cumulative collaboration value. Announced money and collected money are two different accounts, and guess which one keeps making the headlines.

There's one more thing to note here, and it's less about the money than about the word “AI” itself. Where's the line on what counts as AI drug discovery? Admittedly, I'm still working that out, but the high bar is this: the AI finds the target and then designs the drug for it.

By that bar, most of these deals fall short. The targets were already known: USP1 for Exelixis, KAT6 for Menarini, NLRP3 for Hygtia. Insilico designed sharp molecules to hit them, but it didn't find them, and finding the target is the part that's supposed to be the breakthrough. Rentosertib is the one drug that clears the bar, and even it is early: a 71-patient, twelve-week Phase 2a, safety as the endpoint, a handful of dropouts over liver injury, and the authors asking for bigger, longer trials.

No pure AI-discovered drug has cleared Phase 3 yet. Not one.

None of that makes the excitement wrong. It just makes it early. Strip out the billions nobody's paid yet, and the proven part of Insilico is small: one mid-stage drug, and not much money that's changed hands. It's a long way from a proven drug company. Their big June 30 update won't move that line unless it brings a real drug or real money.

So for now, Insilico is the best show in AI drug discovery. Watch it. Cheer for it. But just keep your hand off the champagne, for now.

Eli Lilly joined a $100 million stock sale for Absci the same day Absci's AI-designed antibody passed its first human safety test. ABS-201, Absci's drug for hair loss, was safe and well tolerated in the first part of its Phase 1 HEADLINE trial (NCT07317544, 32 healthy volunteers) and stays in the body long enough to need only two or three shots over six months; that part tested safety, not whether it regrows hair, and the part that tests that has only just started. The same day, Absci sold about $100 million of stock (13,495,277 shares at $7.41), with Eli Lilly among the buyers, though Lilly's share was not disclosed. [Link]

Bayer signed an AI drug-discovery deal with Iambic Therapeutics and revealed none of what it is paying. Announced June 22, the partnership puts Iambic's AI to work finding small molecules for targets that have been hard to drug in Bayer's early pipeline. Iambic gets an upfront payment plus later milestones and royalties, but no dollar figures were given, and its claim of reaching the clinic in "about one-third" the usual time is its own, not independently shown. [Link]

Sanofi put a longtime AI-biotech executive in charge of all its drug research as its new CEO tries to fix a stalled pipeline. On June 22 Sanofi named Paulo Fontoura, who had been chief medical officer at AI startup Xaira Therapeutics and spent 15 years at Roche before that, as its head of R&D starting September 1, under recently appointed CEO Belén Garijo. He replaces Houman Ashrafian, who is leaving after about three years, following a run of research setbacks that also cost the previous CEO his job. [Link]