ARPA-H Launches $98.5 Million Initiative to Revolutionize Rare Disease Care with AI

According to the U.S. Advanced Research Projects Agency for Health, more than 350 million people around the world are living with one of over 10,000 rare diseases — a vast, often silent tapestry of conditions that medicine has, until now, struggled to see clearly. ARPA-H, an agency within the Department of Health and Human Services, is now turning to artificial intelligence to do what the human eye alone cannot: shorten the long, wearying path from first symptom to a name, and from a name to a treatment.

The diagnostic odyssey, quantified

For these patients and their families, the average wait for an accurate answer stretches six years — and sometimes decades — marked by misdirected care, irreversible disease progression, and costs that accumulate quietly in the background of ordinary life. Only about 5% of rare diseases currently have an approved therapy; what is missing, in other words, is not only medicines, but the data infrastructure and patient populations required to design clinical trials in the first place. Delayed diagnoses make the problem worse, because they make it harder to find enough patients to recruit for studies that might, in time, yield therapies.

What the new program will build

ARPA-H has announced up to $98.5 million in contract awards through its Rare Disease AI/ML for Precision Integrated Diagnostics program, known as RAPID, to be invested over four and a half years. Funded teams will work across the full AI stack — building technologies that can learn from sparse and complex rare-disease data, and assembling the foundational datasets those tools depend on. The aim, in the agency's framing, is earlier diagnosis, a deeper understanding of how these diseases progress, and a faster route from target discovery through clinical trial design and endpoint selection. RAPID Program Manager Scott Gorman has described rare diseases as one of the greatest unmet needs in medicine and one of the most important frontiers for AI; the program, he has said, aims to build AI-ready datasets and infrastructure that improve outcomes for patients while advancing precision medicine more broadly.

Why this moment feels different

What we are watching, here, is a quiet restructuring of the diagnostic landscape rather than a single dramatic breakthrough. Genomic sequencing has already shown that timely access can lift diagnostic rates for many rare diseases — yet many genetic results remain inconclusive without richer clinical context, and many conditions have no known genetic cause at all. AI is uniquely suited to knit these disparate signals together, provided the underlying datasets are robust, representative, and shared across institutions that have historically worked in isolation. The practical questions for researchers, clinicians, and patient communities are now concrete: which consortia end up funded, how quickly the AI-ready datasets become accessible to outside investigators, and whether the tools that emerge can be validated against the rarest of the rare — the conditions that have, until now, been invisible by design.