The push for sovereign AI is reaching university research labs, where control over data, models and compute now shapes what science gets done.

Universities sit at an unusual crossroads in the current technological shift, as their work spans open education and tightly controlled federal research. Learning sits at the heart of both missions, making AI a core supporting technology for expanding research and education alike, according to Vince Kellen (pictured, center), chief information officer of the Texas A&M University System.

“AI is a platform for thought. It’s a platform for discovery,” Kellen noted. “AI kind of undergirds all of what we’re going to be doing in the future — certainly what we’re doing in the present now.”

Kellen and Casey Bryan (right), chief information officer of the Texas A&M Engineering Experiment Station, spoke with Krista Case (left) during theCUBE’s coverage of the AI Cybersecurity & Resilience Summit. They discussed how Texas A&M is building a secure, controlled foundation for AI-driven research. (* Disclosure below.)

Building a sovereign AI foundation for research at scale

The foundation of Texas A&M’s approach is IGNITE, short for Innovative Growth in Next Generation AI Technology Ecosystem. It is a state-funded AI and high-performance computing platform led by TEES, as the experiment station is known. Dell Technologies Inc. is building the system with Advanced Micro Devices Inc. graphics processing units inside an Equinix Inc. data center in Dallas. That investment will give research teams the computing power their existing systems lack, Bryan noted.

“Part of our current challenge is that existing resources, while they are performing some of that work, are severely underpowered,” he said. “Research is getting done but could be helped by capacity and scale, which I think IGNITE’s going to bring to the table.”

But the economics matter as much as the capability. Public attention has centered on neoclouds and cloud providers, yet on-premises data centers are growing significantly because of both security controls and cost, according to Kellen. High retail prices for AI computing resources are now limiting the ability of institutions to conduct scientific research.

“There’s no way we could take the science to your classic neocloud provider today and conduct the same thing,” Kellen said. “It’s 5 to 10x the cost.”

Controlling those assets significantly reduces costs and strengthens security, Kellen said. Sovereignty spans four pillars — data, infrastructure, capability and model, and governance — with Equinix supplying physical protection for the compute itself, Bryan added. As a state agency, TEES follows National Institute of Standards and Technology frameworks 800-53 and 800-171, as well as Texas Administrative Code 202, and it wants IGNITE to be ready for Cybersecurity Maturity Model Certification Level 2.

“There’s a lot of controls we have to inherit, but our researchers shouldn’t have to think about all those,” Bryan said. “It should be absolutely seamless when they go to use the capabilities.”

That principle shaped the IGNITE build. When TEES sat down with Dell’s implementation team to schedule workshops, security came first, and the project’s architecture review board has since affirmed that direction, according to Bryan. For institutions starting their own sovereign AI projects, that sequencing is the lesson.

“The idea was to build that security foundation with which all of the other operational components could be built and referenced back to,” Bryan said. “At its core, start with security early — get it embedded in the process earlier on — and it’s much easier to keep it baked in at the beginning than have to worry about adding that at the end.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the AI Cybersecurity & Resilience Summit:

(* Disclosure: TheCUBE is a paid media partner for the AI Cybersecurity & Resilience Summit 2026 event. Neither Dell, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE