Pandemic preparedness is undermined by fragile, thin late-stage pipelines in high-risk viral families. Using a machine learning model trained on 11,600 trials and 560,000+ documents, we predict phase-transition success to identify clinical vulnerabilities before failure occurs. Our analysis shows misaligned portfolios and dangerous gaps. Proactive clinical trial intelligence is essential to forecast risk, guide diversification, and ensure deployable vaccines and therapeutics within 100 days of the next outbreak.