AI drug discovery is the use of artificial intelligence to identify what a new medicine should target in the body and to design the molecule that will act on it. Traditionally this is the slowest and most expensive part of making a drug: bringing a single medicine to market has been estimated at around ten years and well over a billion dollars, with most candidates failing somewhere along the way. The promise of AI is to compress the earliest stages — finding a target and designing a molecule — from years down to months.
The approach builds directly on the kind of capability AlphaFold demonstrated. Once a system can model how proteins are shaped and how molecules might bind to them, it can propose drug candidates far faster than trial-and-error chemistry in a lab. In 2025, the first drug both discovered and designed using generative AI — rentosertib, for a serious lung disease called idiopathic pulmonary fibrosis — posted positive results from a mid-stage human trial, published in a peer-reviewed medical journal.
That milestone deserves its caveats, and they matter. It was one drug, an early-stage trial, a small study whose own authors asked for caution. Discovering a candidate faster is not the same as proving it safe and effective: that still requires years of clinical trials, most of which fail, regardless of how the molecule was found. AI has compressed the front of the pipeline, not the regulated human testing that follows.
The reason the field draws so much attention is the shape of what is changing. The slow, human, decades-long step of deciding what to make is the one being handed to something that does not work at human speed. Whether that translates into a wave of new cures depends on everything downstream holding up — but the bottleneck has demonstrably moved, and that is what makes this more than hype.
