HPE and Oak Ridge National Laboratory used the EmTech AI stage on May 1, 2026 to argue that sovereign AI factories — national-scale compute clusters owned and governed by the country running them — are now the central infrastructure question for governments and large enterprises. Chris Davidson, Vice President of HPC and AI Customer Solutions at Hewlett Packard Enterprise, and Arjun Shankar, Division Director for the National Center for Computational Science at Oak Ridge, framed data control as a strategic imperative rather than a compliance checkbox. The session was hosted by MIT Technology Review.
The pitch is straightforward. Companies and governments want models tuned to their own data, run on hardware they control, with audit trails they can defend. The hard part is keeping that ownership intact while still moving high-quality data through the system fast enough to produce useful output.
Davidson leads HPE's global strategy for AI Factory solutions and what the company calls Sovereign AI, working with governments, enterprises, and research institutions to stand up secure national- and enterprise-grade capabilities. His remit also covers Product Management and Performance Engineering across HPE's HPC and AI portfolio, including large-model training platforms and Cray exascale systems. He has been at HPE for nine years, with prior stints in biotech and medical diagnostics.
Key facts
- 01HPE and Oak Ridge National Laboratory shared the EmTech AI stage on May 1, 2026 to pitch sovereign AI factories.
- 02Chris Davidson leads HPE's global strategy for AI Factory and Sovereign AI customer solutions across governments and enterprises.
- 03Arjun Shankar runs Oak Ridge's National Center for Computational Science and holds a joint faculty seat at the University of Tennessee's Bredesen Center.
- 04Davidson's portfolio at HPE covers large-model training platforms and Cray exascale systems, built over nine years at the company.
- 05The session was hosted by MIT Technology Review at its EmTech AI conference.
Shankar's seat at the table is the public-sector counterpoint. Oak Ridge's National Center for Computational Science is one of the United States' premier scientific computing facilities, and Shankar holds a joint faculty appointment at the University of Tennessee's Bredesen Center. He is a senior member of both the IEEE and the ACM, and his research sits at the intersection of computer science and large-scale scientific campaigns that depend on scalable compute and data infrastructure.
“HPE is positioning its Cray exascale systems and AI Factory line as the substrate for sovereign national compute, with Davidson nine years into building the portfolio.”— Jaeden Schafer
Sovereign AI as a category has hardened over the past two years. Governments in Europe, the Gulf, and Asia have committed to building domestic compute rather than renting capacity from US hyperscalers, and HPE has spent that window positioning its Cray-derived systems as the alternative platform. The AI Factory framing — a turnkey stack of training hardware, inference capacity, networking, and governance tooling — is HPE's answer to Nvidia DGX Cloud and the hyperscaler bundles.
The EmTech AI conversation centered on a tension that gets glossed over in vendor decks. Sovereign control over data is only useful if the data can actually flow into models cleanly and quickly. National labs like Oak Ridge have spent decades solving that problem for scientific workloads. Commercial AI factories are now trying to inherit that playbook on a much shorter timeline.
Governance was the other through-line. Davidson and Shankar both emphasized that scale, sustainability, and governance have to be designed in from the start rather than bolted on once a cluster is live. Power draw, water use, and chain-of-custody for training data are now line items in procurement conversations that used to focus only on FLOPS per dollar.
The skeptic's case is that sovereign AI is mostly branding. Most national programs still depend on US-designed accelerators, US-trained foundation models as starting points, and a small number of vendors — HPE among them — to integrate the stack. True sovereignty in that picture is partial at best, and the marketing language outruns the engineering reality.
For HPE, the bet is that the partial version is still a large enough market to anchor the next decade of HPC revenue. Every government that decides it cannot run its citizens' data through a foreign hyperscaler becomes a customer, and the AI Factory line is the product built to catch them. The pitch at EmTech AI was less about a new technology than about claiming the lane before Dell, Lenovo, and the cloud providers' on-prem offerings get there.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
