AI Drug Discovery: AstraZeneca Pays insitro $40M for MASH Targets

What AstraZeneca and insitro agreed

AstraZeneca has signed a drug discovery and licensing agreement with insitro, the machine-learning biotech based in South San Francisco. insitro will use its AI platform to find new drug targets and design treatments for metabolic dysfunction-associated steatohepatitis (MASH) and for cardiomyopathies, a group of diseases of the heart muscle.11 The deal was announced on October 15, 2026. insitro will apply its Virtual Human platform to find the biology that causes both conditions. It will then take targets that AstraZeneca nominates from discovery to the development-candidate stage.11

The split of work is clear. insitro does the early science. AstraZeneca leads clinical development, regulatory work and commercialization, and holds exclusive worldwide rights to the programs it selects.1315 According to Reuters' reporting, AstraZeneca will also run and pay for all later clinical trials.5

The money: $40 million certain, the rest conditional

The fixed number is the same everywhere: insitro gets $40 million upfront. It can also earn up to about $1.46 billion in development and commercial milestones, plus single-digit royalties on global sales.11

The headline value is where coverage splits. Bloomberg reported the deal as worth up to $1.86 billion.1 Fierce Biotech, Pharmaceutical Executive and The Wall Street Journal, as cited by Quartz, put it at roughly $1.5 billion.21517 Quartz pointed out that the disclosed upfront and milestone figures add up to about $1.5 billion, not $1.86 billion.12 Pharma Journal said neither press release states a total, and it could not find the basis for the higher figure.

The $1.5 billion figure is the better-supported one. Even so, it overstates what changes hands now. Pharmaceutical Executive calculated that the upfront payment is less than 3% of the combined headline value. Neither company has said how many targets are involved, when programs are expected, or when payments fall due.15 There is also a small difference between the two press releases. AstraZeneca describes development and commercial milestones, while insitro's release also includes discovery milestones.15 For now, the money AstraZeneca has committed for certain is $40 million. Everything else depends on programs getting through stages that most programs never clear.

That structure is standard for platform deals. A fixed payment of this size isn't a vote of no confidence. It reflects where the work starts: no candidate, no efficacy data and no safety data were announced alongside the deal.15

Why the bet is on targets, not molecules

For anyone tracking AI drug discovery, the main point is which step AstraZeneca is paying to improve. Sharon Barr, who heads BioPharmaceuticals R&D at AstraZeneca, called choosing the right target one of the most critical decisions in R&D.11 Speaking to Reuters, she said that pairing AI with experienced scientists means testing fewer molecules in the lab and doing less guesswork.5 Reuters also quoted her on timing. Once a pathway or target appears in the major journals, she said, it is public and no longer new. Finding those links early and testing them in cells is what puts a company ahead.18

insitro's CEO, Daphne Koller, described the company's place in the field the same way. She told Bloomberg that most AI drug discovery companies work on what happens after a target has been found. insitro concentrates on the earlier step of understanding disease biology.12 That sets it apart from much of the attention in AI biotech, where generative chemistry and AI protein design have drawn the most interest. Designing better molecules against the wrong target still fails. AstraZeneca is paying to reduce that risk before any molecule is made.

Design is still part of the deal. AstraZeneca describes the Virtual Human platform as combining large-scale human and cellular data with AI-based target identification and medicine design.11 Reuters reporting describes it as lab-grown cell models combined with machine learning.5 Pharma Journal noted that insitro's release names a second platform, TherML, which designs medicines against the causes that Virtual Human identifies. The release does not say whether TherML is part of the AstraZeneca work. Whichever tool is used, insitro's role ends at a development candidate, before any human trial. This deal pays for upstream causal biology. Design work is part of the package, but the target finding is what AstraZeneca is buying.

Why MASH and cardiomyopathies

The two diseases fit the approach. Both companies call them complex and heterogeneous, meaning the underlying biology differs a lot from one patient to another.1114 Koller described them as common, underdiagnosed diseases where detailed, data-rich measurements carry a strong signal about what drives disease.11 If a causal-inference platform is going to show it can beat conventional discovery, these are reasonable places to test it.

AstraZeneca's track record in both areas adds weight. Quartz reported that its drug Wainua failed to show benefit in a heart muscle disease trial this year, and that the company stopped developing a mid-stage MASH candidate last year.17 Pharmaceutical Technology named that candidate as the antisense drug AZD2693, dropped after disappointing Phase II results. AstraZeneca is now testing AZD2389, a small-molecule FAP inhibitor, in Phase II.10 The same outlet noted that Farxiga, AstraZeneca's best-selling SGLT2 inhibitor, is losing sales to generics after its US exclusivity ended earlier this year. That puts more pressure on the company's cardiovascular, renal and metabolism business.10

The MASH market is also getting crowded. Madrigal's Rezdiffra and Novo Nordisk's Wegovy are already in use, and analysts expect more incretin drugs and antifibrotics to follow.6 A new drug in that market will need a mechanism that existing drugs don't cover. That makes new, causally validated targets the point of this deal.

On timing, Koller told Bloomberg that insitro has made considerable progress in MASH and expects to identify targets relatively soon. The heart program will take longer.12 insitro already has its own liver work. Pharmaceutical Executive reported that the company has nine programs, including genetic regulators of liver fat and fibrosis expected to reach the clinic in early 2027. The release does not explain how those internal programs relate to targets AstraZeneca nominates.15 That overlap is worth watching.

A pattern across AstraZeneca and insitro's other deals

For AstraZeneca, this is the latest in a series of AI moves. Fierce Biotech listed its immunology target partnership with Algen Biotechnologies, its acquisition of Modella AI and an agentic AI licensing deal with Owkin.2 Pharma Journal described the common pattern: partners provide the target-finding platform, AstraZeneca keeps the expensive and risky late stages, and payments are small upfront and large only if programs succeed. Crypto Briefing linked the deal to AstraZeneca's goal of about $80 billion in revenue and 20 new medicines by 2030.16

For insitro, the deal adds to a list of big-pharma partners. Gilead signed in 2019 and Bristol Myers Squibb in 2020, and BMS has extended and expanded its deal since.2 Reuters reported recent work with BMS on ALS and with Eli Lilly on AI models that predict how small molecules behave in living organisms.57 Fierce noted that these deals pay mostly later, but they still bring in cash while insitro builds its own pipeline.2 Pharmaceutical Executive reported that the company has more than $750 million in capital.15

The verdict: a real bet whose size is easy to overstate

The reporting agrees on the substance: the two diseases, the platform, the $40 million and the division of labor. It differs mainly on how big the deal sounds. The $1.86 billion figure doesn't match the disclosed terms. The reasonable reading is a deal capped at about $1.5 billion, paid out over years and only if things work.

It is still significant. A top-tier drugmaker has agreed to rely on an outside AI platform for the step its R&D chief called among the most critical in drug development, in two areas where it has recently had setbacks. That shows large companies treating AI target discovery as practical infrastructure. Pharmaceutical Executive noted that no independent check of the platform's performance in these diseases is available.15 Any target it finds will still have to be validated preclinically and in trials.15 As Pharma Journal pointed out, the first concrete sign of progress would be AstraZeneca nominating targets and insitro delivering a development candidate. Until that happens, the deal tests whether AI can find better targets. It doesn't yet show that it can.