[Bernstein: Around 70 AI drug development projects have entered clinical trials, real answers will take 3 to 5 years] Deep潮 TechFlow reports, according to潮向研究, Bernstein's August 26, 2026 research report indicates that approximately 70 AI-driven drug development projects have entered clinical pipelines, with the vast majority still in early stages. The real test will come in the next 3 to 5 years. Recently, DeepMind and Anthropic CEOs claimed that 'most diseases will be cured within 5 to 10 years,' but Bernstein believes these statements are more like a defense of AI's societal value, lacking realistic support for drug development. From target discovery to clinical trials to healthcare system capacity, every layer presents bottlenecks, and biology ultimately needs to be validated in human bodies—a pace that algorithms cannot compress. Bernstein argues that AI's core value in the pharmaceutical field lies in improving decision-making quality, not merely accelerating processes. A 20% improvement in success rates generates far greater returns on research and development than equivalent savings in time and costs. If AI is applied to tackle difficult-to-treat diseases with insufficient biological understanding, even if overall success rates remain low, the value created could be even greater. Major pharmaceutical companies show significant differences in their AI strategies: Eli Lilly has the widest range of external collaborations (25 projects), Roche has the largest investment in computing power, and Amgen benefits from the strongest data moat through deCODE Genetics. AI has yet to visibly impact FDA approval data, with an average of 48.5 new drugs approved annually over the past five years, reflecting more on regulatory modernization and the rise of biologics. AI has already played a role in molecular design, virtual screening, and target discovery, but 'curing all diseases' requires overcoming multiple bottlenecks—this is not a problem that algorithmic iteration alone can solve.