Mayo Clinic · Healthcare and medical research · United States (Rochester, Minnesota)
Mayo Clinic: a DGX SuperPOD with DGX B200 for pathology foundation models
Mayo Clinic deployed an NVIDIA DGX SuperPOD with DGX B200 systems in July 2025 and says its first work will be building foundation models for pathology, drug discovery and precision medicine. Mayo states the system cuts four weeks of slide analysis and model work to one; no measurement is published.
- Status
- In production
- Status as of
- 28 Oct 2025
- NVIDIA technologies
- NVIDIA DGX, NVIDIA Mission Control
The challenge
Pathology is still largely manual and slow, according to NVIDIA. Mayo Clinic holds about 20 million digitized whole-slide images with about 10 million associated patient records, and wanted to use that archive to build foundation models for pathology, drug discovery and precision medicine.
Whole-slide images are very large, so training on them needs a lot of GPU memory and compute; NVIDIA points to the 1.4 TB of GPU memory per DGX Blackwell system as a fit for this data.1
What was implemented
In January 2025 NVIDIA and Mayo Clinic announced a collaboration on pathology foundation models, with Mayo planning to deploy DGX Blackwell systems and to use MONAI, the open medical imaging framework that NVIDIA and King's College London started. Later steps named in that release (Cosmos Nemotron vision language models and NIM microservices) were plans, not deployments.
In July 2025 Mayo Clinic said it had deployed an NVIDIA DGX SuperPOD with DGX B200 systems. Mayo says the system's first work will be building foundation models for pathomics, drug discovery and precision medicine. Mayo builds on Atlas, a pathology foundation model developed with Aignostics and trained on more than 1.2 million whole-slide images.
In October 2025 NVIDIA described the system as a Mayo Clinic AI factory built on DGX SuperPOD with DGX B200 and managed with NVIDIA Mission Control.12345
Reported outcomes
Mayo states the infrastructure cuts four weeks of pathology slide analysis and foundation model work to one2
4 weeks to 1 week (stated at deployment, not measured)
Projected, not measured
Each outcome links to its source; the label there says whether the company, NVIDIA or a third party reported it.
What others can learn, and the limits
What others can learn: a large, well-labeled internal data archive (here, 20 million digitized slides) is what makes owning training infrastructure worth considering; Mayo pairs it with an external model partner (Aignostics) rather than building everything alone. Limits: this is an infrastructure case, and the only stated result is Mayo's own four-weeks-to-one statement, made at deployment, with no workload, baseline system or method given. The cited sources report no clinical outcome data. Most hospitals do not have the data volume or research workload to keep a SuperPOD busy; cloud GPU capacity or smaller on-premises systems are the usual starting point.
Explore a similar project
Use the AI Factory Efficiency Lab with your own assumptions. Results are independent of this case.
Sources
- NVIDIA Partners With Industry Leaders to Advance Genomics, Drug Discovery and Healthcare (opens in a new tab)
- Mayo Clinic deploys NVIDIA Blackwell infrastructure to drive generative AI solutions in medicine (Mayo Clinic release via Newswise) (opens in a new tab)
- NVIDIA and Partners Build America's AI Infrastructure and Create Blueprint to Power the Next Industrial Revolution (opens in a new tab)
- Project MONAI: About (opens in a new tab)
- London AI Centre and NVIDIA launch MONAI, a new AI framework for healthcare (opens in a new tab)
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.