At MIT Technology Review’s EmTech AI conference, Hewlett Packard Enterprise’s VP of HPC & AI Customer Solutions Chris Davidson and Oak Ridge National Laboratory division director Mallikarjun Shankar debated how enterprises can own data while delivering trustworthy AI at scale. Their joint experience spans biotech, cloud services, and national research.

The panel argued that AI Factory pipelines for model training, validation, and deployment enable organizations to reuse high‑quality datasets without exposing them to external clouds. By keeping compute on‑prem or within sovereign clouds, firms reduce latency and regulatory risk while achieving exascale performance on HPE’s Cray systems. Customers report faster model iteration and lower total cost of ownership.

Shankar highlighted the National Center for Computational Science’s role in bridging massive scientific campaigns with AI‑driven analytics, leveraging the same sovereign infrastructure. His dual appointment at the University of Tennessee and senior memberships in IEEE and ACM underscore the interdisciplinary push needed to mainstream AI in high‑performance environments. The discussion underscored that controlling data pipelines is now as strategic as raw compute power for security.