On 2026-08-04, the U.S. National Science Foundation announced its State and Regional AI Infrastructure Hubs programme. The programme is designed to connect researchers and higher-education institutions to shared compute, data, software, and expertise through state or multi-state consortia.
What is actually funded
The solicitation says approximately $100 million is available and anticipates 10 awards in an award cycle. Typical hub proposals may request between $4 million and $12 million over 5 years, with a full-proposal deadline of 2026-11-04. Those are programme expectations, not awards already made.
A crucial design detail is the split of responsibility. Consortia are responsible for funding new or expanded computing, data, and AI resources, whether on-premises or cloud-based. NSF funding supports consortium coordination, AI-infrastructure workforce development, faculty training, and coursework. The programme therefore treats access, skills, and institutional coordination as infrastructure components rather than assuming hardware alone creates usable capacity.
Independent context
NVIDIA announced its participation on 2026-08-04 and described planned contributions of infrastructure, software, educational resources, and technical support. That description is a participant's statement, not evidence that every proposed hub will receive the same resources.
What organisations can learn
Shared AI infrastructure needs an operating model: eligibility, scheduling, cost allocation, data handling, support, and outcome measurement. Regional consortia should define those controls before capacity arrives. Applicants should use the NSF solicitation as the authoritative source and distinguish programme availability from a future award decision.