Bernstein: About 70 AI drug discovery programs have entered clinical trials; definitive answers will still take 3 to 5 years

According to Chaoxiang Research, a Bernstein report dated August 26, 2026, indicates that approximately 70 AI-driven drug development projects have entered clinical pipelines, with the vast majority still in early stages, leaving the true test for the next three to five years. The CEOs of DeepMind and Anthropic recently claimed that “most diseases will be cured within 5 to 10 years,” but Bernstein regards such statements more as defenses of AI’s societal value, noting they lack grounding in the practical realities of drug development. Every layer—from target discovery to clinical trials to healthcare system capacity—presents a bottleneck. Since biological mechanisms must ultimately be validated in humans, this timeline cannot be compressed by algorithms. Bernstein argues that AI’s core value in pharmaceuticals lies not merely in accelerating processes, but in enhancing decision-making quality. A 20% improvement in success rates generates R&D returns that far outweigh equivalent reductions in time and costs. Conversely, applying AI to treat refractory diseases where biological understanding remains limited can yield even greater value, even if the overall success rate is relatively low. Major pharmaceutical companies exhibit significantly different AI strategies: Eli Lilly leads in external partnerships (25 initiatives), Roche invests the most in computing power, and Amgen boasts the deepest data moat backed by deCODE Genetics. AI has yet to produce a measurable impact on FDA approval metrics; the average annual approval of 48.5 new drugs over the past five years better reflects regulatory modernization and the rise of biopharmaceuticals. AI has already begun playing a role in molecular design, virtual screening, and target discovery. However, “curing all diseases” requires surmounting multiple bottlenecks—a challenge that algorithmic iteration alone cannot resolve.