AI-driven drug discovery could substantially reduce the cost and time required to develop a drug candidate. Rentosertib is the clearest example so far. According to Insilico Medicine, the program moved from hypothesis to a preclinical candidate in under 18 months. Rentosertib’s preclinical work came in near $2.6 million against a conventional $15-25 million. Insilico published peer-reviewed Phase IIa results in Nature Medicine. One successful program cannot establish the economics of an entire platform. Rentosertib targeted idiopathic pulmonary fibrosis, a well-characterized disease with established clinical endpoints. It's still too early to know whether this approach can deliver similar timelines and costs in more complex disease areas. About $230B in branded U.S. drug sales are at risk of generic or biosimilar competition between 2025 and 2030. Partnerships may become a more natural route for pharmaceutical companies seeking access to AI-driven discovery without taking on the risk of building the entire stack themselves. If other programs produce similar results, more of the value in drug discovery could move toward full-stack platforms that license their capabilities to pharma.