HPE's AI Factory lead and Oak Ridge's compute chief outline how governments and enterprises are building sovereign, large-scale AI infrastructure outside hyperscaler control.
Chris Davidson (VP, HPE HPC & AI) and Arjun Shankar (Division Director, Oak Ridge National Laboratory) are presenting a joint perspective on operationalizing AI at national and enterprise scale. HPE is positioning its AI Factory and Cray exascale systems as the backbone for sovereign AI deployments. Oak Ridge brings the scientific HPC credibility, having deployed Frontier — the world's first exascale system. Together they represent a push to move serious AI workloads off hyperscaler clouds and onto sovereign, on-premise or nationally controlled infrastructure.
This isn't about a new API or model release — it's about the physical compute layer that runs large-scale AI training for governments and research institutions. Developers building on top of public cloud are insulated from this shift for now, but those working on HPC-adjacent problems — scientific ML, large model training at national lab scale — should track HPE's Cray and AI Factory stack as a deployment target.
If you're training models above 70B parameters and spending over $50K/month on cloud GPU, request an HPE AI Factory sizing brief this week to benchmark on-prem TCO against your current AWS/GCP spend.
Tags