Particle.news

Download on the App Store

Mayo Clinic Deploys On-Premises Nvidia DGX SuperPOD to Accelerate Clinical AI

Housed in Brooklyn Park with air-cooling, the 128-GPU system triples compute power to convert four-week model training into one-week workflows, driving Mayo’s precision medicine advances.

Image

Overview

  • Operational since July 28, the Brooklyn Park DGX SuperPOD integrates 128 Nvidia Blackwell GPUs to deliver three to four times the compute performance of Mayo’s previous infrastructure.
  • The system compresses foundation model training for pathomics, drug discovery and precision medicine from four weeks to one week to speed up diagnostic and research timelines.
  • Over 12 million digitized pathology slides now fuel the SuperPOD’s high-resolution image analysis for AI-driven insights into disease staging and progression.
  • Hosting the supercomputer on-site ensures scalable, cost-effective performance and tighter governance of sensitive patient data compared with cloud-based alternatives.
  • Air-cooling technology and optimized performance-per-watt design minimize energy use, aligning the deployment with Mayo’s sustainability goals and broader AI partnerships.