AstraZeneca has struck a $1.86 billion partnership with Insitro, a machine-learning drug discovery startup, deepening a wave of pharmaceutical industry bets on artificial intelligence to speed up the notoriously slow and expensive process of finding new medicines. The deal, disclosed in mid-October, pairs AstraZeneca’s clinical development infrastructure with Insitro’s AI models trained on large-scale biological and genomic datasets.
What Insitro Brings to the Partnership
Insitro, founded by former Stanford computer scientist and Google Brain researcher Daphne Koller, builds machine learning models trained on lab-generated biological data designed to predict which drug candidates are most likely to succeed before they ever reach human trials. The company’s approach differs from some AI drug discovery rivals by generating much of its own training data in-house through automated wet-lab experiments, rather than relying solely on publicly available datasets, a strategy intended to produce higher-quality predictions for complex diseases where existing biological data is sparse.
How This Fits a Broader Pharma-AI Dealmaking Wave
The AstraZeneca-Insitro agreement is one of several large AI drug discovery partnerships announced in 2026. Eli Lilly previously expanded its own AI drug development agreement with Insilico Medicine in a deal worth up to $2.75 billion, while Takeda separately signed a $600 million AI drug discovery agreement with the same company. Industry-wide, AI drug discovery deals have topped $47 billion in the first six months of 2026 alone, according to tracking by healthcare investment analysts, reflecting how aggressively major pharmaceutical companies are betting that machine learning can shorten the decade-plus timelines traditionally required to bring a new drug to market.
The Uncomfortable Fact Looming Over Every Deal
Despite the enormous sums changing hands, no drug designed primarily through AI has yet won full FDA approval, a gap that has become a persistent talking point among skeptics of the AI drug discovery boom. Drug development remains a multi-stage process spanning target identification, preclinical testing, and three phases of human clinical trials that typically take eight to twelve years in total, and AI’s contribution so far has mostly accelerated the earliest stages, such as identifying promising molecular targets, rather than eliminating the lengthy and expensive clinical trial process itself.
The Bull Case From Industry Executives
AstraZeneca and other pharma executives argue that even incremental improvements in target identification and candidate selection can save hundreds of millions of dollars by weeding out drug candidates likely to fail clinical trials before significant capital is spent on them. Supporters point to the fact that the traditional drug discovery success rate, historically estimated at less than 10% from first clinical trial to approval, leaves enormous room for AI-driven improvements in candidate selection to generate outsized returns even without directly automating clinical trials.
The Skeptical Counterpoint
Critics, including some biotech investors who lived through earlier waves of computational drug discovery hype in the 2000s and 2010s, caution that AI models trained on existing biological data may simply be rediscovering patterns already known to experienced drug developers, rather than generating genuinely novel insights. There is also a reproducibility concern: biological systems are famously difficult to model computationally, and a drug candidate that looks promising in an AI-driven simulation can still fail for reasons the model never captured, a risk that has sunk computational drug discovery efforts before.
What’s Next
The real test for the AstraZeneca-Insitro partnership, and for the broader AI drug discovery sector, will be whether any of the candidates generated through these platforms successfully clear Phase 2 and Phase 3 clinical trials in the coming years, the stage where most drug candidates historically fail regardless of how they were discovered. Until an AI-originated drug wins FDA approval, the sector’s enormous deal values will remain a bet on future potential rather than a proven return on investment, making the next 24 to 36 months a critical proving ground for the entire category.
Photo: Edward Jenner / PEXELS via Pexels
