AstraZeneca-insitro AI Pact Pays $40M Upfront for MASH Targets

The deal

AstraZeneca has agreed to pay insitro, the machine-learning drug discovery company founded by Daphne Koller, $40 million upfront to find new drug targets in two hard-to-treat cardiometabolic areas: metabolic dysfunction-associated steatohepatitis (MASH) and cardiomyopathies. insitro can also earn up to about $1.46 billion in milestones, plus single-digit royalties on global sales of any resulting products.11 The worldwide exclusive agreement was announced on October 15. insitro will run its Virtual Human platform to find causal disease biology in both conditions, then take targets nominated by AstraZeneca from discovery to development candidate. AstraZeneca will handle clinical development, regulatory work and commercialization.1115

Coverage of the basic terms is consistent. Trade outlets rounded the headline value to roughly $1.5 billion, and all of them repeated the same split between upfront cash and milestones.1416 The details are what make the deal interesting, and some of them don't line up across the two companies' announcements.

Read the fine print, not the headline number

The most important figure is the smallest one. By PharmExec's calculation, the upfront payment is less than 3% of the total headline value. Neither company disclosed how many targets are involved, the program timelines or the payment schedule.12 RuntimeWire made a similar point: most of the value depends on development and sales milestones that may never be reached, so the deal is really a test of whether insitro's target discovery yields programs AstraZeneca chooses to advance.3

The two press releases also describe the milestones differently. insitro's release mentions discovery, development and commercial milestones. AstraZeneca's version mentions only development and commercial ones, a gap PharmExec noted.1211 The AstraZeneca-distributed release, republished in Europe, says the same thing, describing "development and commercial milestone payments."13 This may just be a difference in wording. But if discovery-stage payments exist, they are the money insitro could collect soonest, and the big pharma partner chose not to highlight them. insitro's release calls the royalties "tiered" single-digit royalties; AstraZeneca's text says "potential single digit royalties."1113

The reading here is that this is a sizable vote of confidence wrapped in a typical big-pharma risk structure. AstraZeneca gets exclusive worldwide rights to the programs it picks for a small fraction of the headline price, and most of the risk stays on insitro's discovery engine.13 That isn't a criticism of the deal. It is how platform partnerships usually work, and insitro's earlier deals followed a similar pattern.

Why MASH and cardiomyopathies

Both companies say patient heterogeneity is the reason for picking these diseases. insitro's release argues that in each condition, one diagnostic label covers patients whose underlying biology differs, which has made treatment hard. It presents that as exactly the situation where machine learning on human data is most useful.11 AstraZeneca R&D chief Sharon Barr put it in terms of target selection, calling the choice of target one of the most critical decisions in drug development.12

The cardiomyopathy side is vague. PharmExec noted that neither release says which forms of cardiomyopathy are included.12 The MASH side has a clearer commercial logic. Pharmaceutical Technology says the market is expected to reach $20.3 billion by 2032. It also describes the deal as AstraZeneca doubling down on MASH: the company is testing a small-molecule FAP inhibitor, AZD2389, in Phase II after dropping AZD2693, an antisense candidate that produced disappointing Phase II results.16 The same outlet ties the deal to AstraZeneca's cardiovascular, renal and metabolism franchise, where its SGLT2 drug Farxiga lost US exclusivity earlier this year and now faces generic competition.16 AstraZeneca itself calls that franchise a key growth driver and emphasizes the links between the heart, kidneys, liver and pancreas.13

Read together, the coverage suggests AstraZeneca is buying more shots on goal for new targets in a disease area where its own clinical record is mixed and an older blockbuster is fading.

insitro's funding story and the clinic test

For insitro, the deal is another pharma partnership in a business model that relies on them. The company says it has raised more than $750 million from investors including a16z, ARCH, BlackRock, CPP, Foresite, GV, T. Rowe Price and Temasek. It already works with Bristol Myers Squibb, Lilly and Gilead.11 The funding figures vary slightly by source and accounting method. An InfluenceAsia analysis puts the total at about $800 million when collaboration revenue is included.2 An Observer profile counts roughly $140 million of pharma collaboration revenue within the $750 million-plus total.1 InfluenceAsia cited a figure of about $150 million in collaboration revenue.2 A 2026 IP-recognition profile gives the amount raised as over $740 million.7 Back in 2022, STAT reported that insitro had raised more than $640 million in under four years, including a $400 million Series C.10

Partner deals keep growing in importance. According to the Observer, Bristol Myers Squibb expanded its ALS collaboration twice in a year: it paid $20 million in October 2025 and $10 million in March 2026, when it nominated two more targets from the Virtual Human platform. That brought the partnership to three ALS targets and more than $2 billion in potential milestones.1 insitro's first big deal, with Gilead, also focused on fatty liver disease, with $15 million upfront against about $1 billion in potential milestones.1 The AstraZeneca deal brings the company back to liver biology with a bigger partner and more upfront cash.

Koller has described this as a deliberate hedge. Partner revenue funds the company while it advances a smaller set of programs it majority-owns, so it can share in the upside if a drug works.1 She has also said insitro chose not to in-license outside assets, on the grounds that success with someone else's molecule says nothing about whether her platform works.1

That makes the timing of insitro's own clinical entry important, and here the reporting differs. InfluenceAsia reported in mid-2026 that the lead wholly owned MASH candidate, CTRO-1013, was expected to start human testing during 2026. The program targets IRS1, which the company says emerged as the strongest of 480 genetic signals linked to liver fat.2 PharmExec, citing insitro's release this week, says the company's liver-fat and fibrosis programs are expected to enter the clinic in early 2027.12 On these sources, the first-in-human timeline appears to have moved by at least a quarter. That is common in biotech but worth noting. PharmExec also pointed out that the release doesn't explain how insitro's internal liver programs relate to the targets AstraZeneca will nominate.4 That question matters for investors trying to separate the value insitro owns from the value it is sharing.

The design angle: from targets to molecules

Most coverage treats this as a target-discovery deal. But AstraZeneca's own description also covers designing potential treatments, and it calls the platform a combination of large-scale human and cellular data with AI-supported target identification and medicine design.13 That fits a capability insitro has added recently. The Observer reports that in January insitro acquired CombinAbleAI, an Israeli biologics-design startup, and folded it into a new platform called TherML. TherML is meant to extend target discovery into molecule design and manufacturability across small molecules, oligonucleotides and antibodies.1

That matters for the protein-design side of AI drug discovery. The agreement has insitro advancing targets all the way to a development candidate, not just handing over a ranked list of genes.11 So AstraZeneca appears to be paying for an engine that covers both target finding and molecule design. Antibody and other biologics design would logically be part of that, though no source says which modalities this collaboration will use. The CTRO-1013 program shows the approach in practice: the company says AI helped link IRS1 to fibrosis and then design a liver-directed silencing therapy.2

Koller is careful about how much this can speed things up. She told the Observer that throwing more GPUs at the problem won't make cells grow faster. She also described a case where her team found that an early AI-generated target ranking rested on reasoning that didn't hold up, and caught it before the program moved forward.1 Her a16z essay this summer, titled "Drug Discovery Has No Magic Wands," makes the same argument at length.9

What to watch

No candidates, efficacy data or safety data came with the announcement.12 Pharmacally stresses that this is a discovery-stage deal, not a clinical program, and its significance depends on whether AI-derived insights turn into validated medicines.14 PharmExec adds that the same heterogeneity that justifies the AI approach could complicate trial design later.12

Overall, the coverage points to a measured conclusion. AstraZeneca's commitment is a real endorsement of insitro's focus on causal biology, and the $40 million upfront is meaningful cash for a company paying for its own clinical push. But the headline $1.5 billion is mostly contingent, and the deal doesn't settle whether insitro's platform works. That answer will come from patients. InfluenceAsia argued that CTRO-1013's first trial must show liver-directed exposure, acceptable safety and target engagement.2 If it does, deals like this one with AstraZeneca will look like early bets. If it doesn't, the milestone figures in these deals will be hard to justify.