CRA Report Maps How Four US States Are Building Shared AI Computing

Four states, four approaches

According to the CRA, access to the computing resources needed for AI remains uneven. Universities and colleges face growing demand for advanced AI infrastructure, yet many lack the resources to develop and sustain it on their own. The report looks at how states and regions are responding and what others can learn from them.

It was written by Erik Deumens of the University of Florida, Ian Fisk of the Simons Foundation, Manish Parashar of the University of Utah, Barr von Oehsen of the Pittsburgh Supercomputing Center, and Helen Wright of CRA. It builds on a vision advanced by NVIDIA cofounder Chris Malachowsky for a 50-state AI computing initiative. That vision stresses making advanced computing and AI education accessible across higher education, rather than concentrating resources at a few institutions.

The four case studies each take a different route:

• Florida: the University of Florida's HiPerGator initiative combines a major philanthropic investment with university and state support to create a centralized resource for research and education across the state.
• Pennsylvania: the emerging Keystone AI + Quantum Factory brings together seven research universities in a collaborative model, with shared governance and access beyond the founding institutions.
• Utah: the University of Utah's Redtail AI Factory is a public-private partnership that uses vendor-managed infrastructure to support statewide research, education and innovation.
• New York: Empire AI is a state-supported consortium of public and private research institutions that pool resources and expertise to share advanced AI computing capacity.

The report concludes that there is no single model. Successful initiatives must respond to local circumstances while setting up sustainable approaches to funding, governance and access.

More than hardware

A central finding is that expanding AI computing capacity requires more than buying advanced hardware. The report points to the need to invest in people, partnerships and institutional structures. These include technical support staff, faculty development, educational programs, sustainable funding and governance models that reflect the needs of participating institutions.

Across the four cases, the report identifies several common requirements: a sustained catalyst for collaboration, institutional commitment, reliable funding and trusted governance.

Deumens, who directs UF Information Technology Research Computing at the University of Florida, said in the CRA announcement: "technology is only the starting point." He added that the goal is to ensure capacity "translates into meaningful opportunities for students, faculty, and researchers."

Malachowsky, NVIDIA co-founder and fellow, said the findings reinforce his experience in Florida. In his words, "the institution has to truly own the mission, and the technology is only as good as the people supporting it."

Link to the national conversation and what it means

The report also draws on the CRA-National Science Foundation (NSF) Industry/Philanthropy Summit on Potential NAIRR State Hubs for Educational Transformation, held in February 2026. That meeting brought stakeholders together to look at state and regional partnerships that could expand AI computing access and support educational transformation. The new report adds practical lessons from initiatives already underway.

For states and institutions weighing their own AI infrastructure plans, the report offers a starting point rather than a template. Its examples suggest that funding sources, governance and partnerships will differ by region, and that spending on staff, training and educational programs is as important as spending on computing capacity. By extending access beyond major research universities, the CRA says, such initiatives can help more students, faculty and institutions take part in AI education and research.