NSF AI Datasets 2026: grants up to $5 million to make science data ready for AI
NSF grants of up to $5 million to make existing science datasets usable by AI. Open to US universities, labs and companies. Deadline November 4, 2026.
The NSF AI Datasets program (solicitation NSF 26-512) funds work that makes existing scientific datasets ready for use by AI tools. NSF expects to spend $60 million to $100 million on 25 to 50 awards, subject to available funds.
There are three kinds of award. Impact awards give up to $2 million over up to 3 years, with 10 to 20 expected. Flagship awards give up to $5 million over up to 3 years for larger challenges, with 5 to 10 expected. Planning grants give up to $200,000, with 10 to 20 expected, for communities that need time to plan.
NSF names three goals: use AI for feature extraction, metadata generation and the joining of several datasets; build data pipelines for automated analysis by AI tools; and enrich or harmonize existing datasets so AI pipelines can use them. Proposals should also cover dataset security, integrity and how the research community can contribute to the data.
US universities and colleges, US non profit research organizations and US based companies with strong research capabilities can apply. There is no limit on the number of proposals per organization or per PI. Proposal titles must start with "AIDatasets:Impact", "AIDatasets:Flagship" or "AIDatasets:Planning".
Full proposals are due on November 4, 2026, by 5 p.m. in the local time of the submitting organization, through Research.gov or Grants.gov. After this round, the deadline is the first Wednesday in November each year. Questions go to aidatasets@nsf.gov.
Quick answers
Can I use the money to collect new data?
No. Research to collect new datasets is out of scope. The program funds work that makes existing scientific datasets more useful for AI.
Is there a limit on how many proposals I can send?
No. The solicitation sets no limit per organization and no limit per PI or co PI.