Eight years of AI drug discovery, and still no approved molecules

117 assets designed with AI have entered interventional trials at 63 companies, 60 have completed a phase I and eight a phase II. The capital has exceeded $60 billion since 2019.

Nearly eight years after the onset of the enthusiasm for the discovery of artificial intelligence-assisted drugs, no treatment discovered or designed by these methods has been authorised by a regulator in a major market. What the field has accumulated in the summer of 2026, sums up The Pharma Letter, has three components: capital, clinical trial launches, and a thin first layer of human data.

Identifying this pipeline is more difficult than appears to be, with estimates varying according to the criteria used. An analysis presented this year at the American Society of Clinical Oncology conference identifies 117 AI-related assets, distributed among 63 companies, which have entered human interventional trials. Of these, 60 completed Phase I and eight completed Phase II. Other observatories suggest a figure closer to 175 programs, for about $60 billion invested since 2019, and estimate that 15 to 20 candidates could enter this year's pivotal phase III.

It is precisely this stage that will serve as a justice of the peace. As long as no molecule has passed a decisive trial, the promise of a shortcut to the drug remains a funded hypothesis, not a demonstration.