• The US National Science Foundation has launched a $100m programme expected to support up to ten state or regional AI Infrastructure Hubs
  • NSF funding cannot be used to acquire computing, networking, storage or cloud services, leaving each consortium to fund those resources separately

The fact

The US National Science Foundation has launched a $100m State and Regional AI Infrastructure Hubs programme to expand access to AI resources for scientific research and education. NSF expects to support up to ten hubs, normally one for each selected state or multi-state region. Typical proposals are expected to request between $4m and $12m over five years.

NSF funding will support consortium coordination, workforce development, faculty training and mechanisms for widening access to AI infrastructure. It cannot be used to acquire computing hardware, networking, storage, cloud services, data or other AI infrastructure. Each consortium must secure those resources separately through universities, industry, philanthropy, state or local governments or other sources. The infrastructure may be on-premises, cloud-based or a combination of both.

The assessment

The funding model separates hub operations from infrastructure procurement. NSF can pay for the people, partnerships, and training needed to run a hub, but each consortium must separately secure the computing, storage, networking and other resources that users will access.

Each consortium therefore has to show that its infrastructure is available or being acquired and can be funded throughout the five-year award. Because hubs may use different mixes of on-premises systems, cloud services and partner-provided capacity, an NSF award will not represent a standard amount of compute across regions. For BTW readers, the $100m is programme funding, not a national compute build. The useful measure will be how much computing capacity each winning consortium can actually make available and sustain.

What to watch

Full proposals are due on 4 November 2026. Watch which consortia can document five years of infrastructure funding, what computing resources they commit and whether those resources are on-premises, cloud-based or hybrid. After awards, the clearest measure of delivery will be when that capacity becomes available to researchers and how much each hub can provide.