AstraZeneca and machine-learning drug discovery company insitro announced a deal worth up to $1.5 billion on October 15, under which AstraZeneca will identify promising disease targets and insitro will apply its machine-learning platforms to validate and develop them into potential drug candidates. The agreement is one of several AstraZeneca has struck with AI-focused biotechs this year, joining a separate deal with Tremont AI to screen drug candidates for toxicity before human trials.
How the Deal Is Structured
Under the arrangement, AstraZeneca picks the disease areas and therapeutic targets it wants pursued, while insitro applies its proprietary machine-learning models — trained on large-scale genomic, cellular and clinical datasets — to predict which targets are most likely to translate into viable drugs before either company commits significant money to wet-lab validation. That division of labor reflects a pattern increasingly common in pharma-AI partnerships: the pharmaceutical company supplies domain expertise and disease focus, while the AI company supplies the predictive modeling meant to cut years off early-stage discovery timelines.
Part of a Bigger Spending Wave
The AstraZeneca-insitro deal lands amid what research firm DealForma estimates was $47.1 billion in AI-based target and drug discovery licensing deals in just the first half of 2026 — a figure that dwarfs the AI drug discovery spending of just a few years earlier and reflects how central machine learning has become to pharma’s research strategy. Eli Lilly reportedly followed with its own drug-delivery deal days later, worth up to $1.1 billion, and GSK separately widened its existing AI drug design partnership with Chai Discovery in the same month.
The Catch Nobody Can Ignore
Despite that torrent of dealmaking, no AI-designed drug has yet won FDA approval — a gap that looms over every headline deal value in this space. The billions being committed are, in effect, bets on a technology category that has not yet produced a single validated commercial success, which makes the eventual outcome of deals like AstraZeneca’s with insitro an important real-world test of whether machine-learning target discovery actually shortens the famously long and expensive drug development pipeline, or merely shifts cost and risk earlier in the process.
Why insitro Needed This
For insitro, landing an AstraZeneca partnership of this size provides both near-term capital and a credibility signal to other pharmaceutical companies evaluating AI discovery partners. The company has spent years building what it describes as “dataset-centric” machine learning — generating its own large, consistent biological datasets rather than relying solely on public or partner-supplied data — an approach it argues produces more reliable predictive models than competitors working from fragmented data sources.
The Skeptical Read
Critics of the AI-pharma dealmaking boom point out that headline deal values are typically structured as milestone payments contingent on specific drug development successes, meaning the $1.5 billion figure represents a ceiling AstraZeneca would only pay if every target insitro identifies clears successive stages of preclinical and clinical development — an outcome far from guaranteed given the drug industry’s historically high failure rates even for conventionally discovered compounds. Skeptics also note that machine-learning target identification addresses only the earliest stage of a drug’s roughly decade-long path to market, leaving the costliest and riskiest clinical trial stages untouched by the technology.
What’s Next
The real signal of whether this deal justifies its billion-dollar structure will come years from now, when AstraZeneca either advances insitro-identified targets into clinical trials or quietly lets the partnership lapse. In the nearer term, expect more pharma companies to announce similar AI-discovery partnerships, given how clearly AstraZeneca, Eli Lilly and GSK’s recent deals have established the going rate and structure for this category of agreement.
Photo: Tima Miroshnichenko / PEXELS via Pexels
