Washington – September 22, 2026 -- Internet2 has added Cisco AI PODs at the University of Nebraska–Lincoln and Sacramento State to expand shared GPU capacity on the National Research Platform (NRP), a federally backed computing network supporting thousands of researchers nationwide.
New AI PODs let researchers burst into shared GPU capacity when local resources fall short
The Cisco AI PODs are pre-validated, full-stack infrastructure units built on Cisco Validated Designs, combining UCS servers and GPU nodes, Nexus networking, and Cisco Intersight management. Deployed as a single unit, they reduce the engineering and integration work institutions face when adding new GPU nodes to the NRP's distributed Kubernetes-based fabric, known as Nautilus.
NRP scales through pooled contributions rather than a centralized supercomputer
The National Research Platform is supported by the National Science Foundation and more than 70 participating organizations, pooling computing, storage, and AI resources across U.S. institutions instead of relying on one central facility. That model has created a recurring compatibility challenge for organizations trying to integrate GPU nodes into the shared, multi-region cluster — a gap the standardized AI POD deployment is designed to close.
Internet2 network infrastructure underpins data movement across the platform
Internet2 operates six dedicated Open Science Data Federation nodes — in Boise, Denver, Houston, Indianapolis, Jacksonville, and Kansas City — that keep datasets closer to compute resources and researchers. The organization also runs 41 performance assurance service nodes serving a hybrid NRP function, contributing 1,200 CPU cores, 6.4 terabytes of memory, and 80 terabytes of storage, alongside thousands of perfSONAR instances used for cross-domain performance monitoring.
Executive says the deployment links an NRP need directly to industry capacity
"Today, these two new AI PODs provide researchers access to additional GPU capacity when local resources aren't enough,