NSF AI grant $2M: "AISuper" — AI-driven design of superconducting materials for magnets (NSF–DFG / DMREF)
The NSF awarded about $2M to "AISuper," which accelerates the discovery of new superconducting materials by combining AI, quantum theory, and experimental synthesis. Graph neural networks predict superconducting properties; generative AI designs synthesizable new materials; and density functional theory (DFT) plus experiments validate them in a closed discovery loop.
Grant overview (primary data)
- Award amount$2,000,000
- RecipientUniversity of Florida（FL）
- ProgramOFFICE OF MULTIDISCIPLINARY AC, DMREF
- Period2025-10-01 〜 2029-09-30
- FunderU.S. National Science Foundation (NSF) / NSF
Key points
- Integrates AI, quantum theory, and experimental synthesis to accelerate superconductor discovery
- Graph neural networks predict properties; generative AI (stochastic flow matching) designs synthesizable materials
- DFT plus high-throughput experiments validate candidates in a closed discovery loop
- Publishes an open-access dataset of successful and unsuccessful candidates
- About $2M; NSF–DFG (U.S.–Germany) collaboration (DMREF); Florida; from 2025
The U.S. National Science Foundation (NSF) awarded about $2,000,000 to "AISuper" (AI-Driven Design of Superconducting Materials for Magnets) (NSF Award 2522891; program: DMREF / NSF–DFG international collaboration; Florida; starting October 2025).
Per the abstract, the project accelerates the discovery of new superconducting materials through a transformative approach combining artificial intelligence (AI), quantum theory, and experimental synthesis. Superconductors are essential for technologies from MRI systems and high-field magnets to quantum computing and sustainable energy, yet analysis of known compounds suggests only a small fraction of potential superconductors may have been discovered. The project aims to significantly expand the number of known superconductors and identify materials optimized for practical applications — high critical temperatures and fields, ductility for wire fabrication, and three-dimensional electronic structures for enhanced performance.
Technically, it integrates two complementary AI methods. Property-prediction models based on graph neural networks estimate superconducting characteristics — electron–phonon spectral functions, critical temperatures, and critical fields — directly from crystal structures. In parallel, generative AI models using stochastic flow matching design novel, synthesizable materials with targeted superconducting and mechanical properties. Predictions from both are evaluated with density functional theory (DFT) for thermodynamic stability, electronic structure, and superconducting potential; selected candidates undergo targeted synthesis and high-throughput characterization. Both models are iteratively refined with experimental feedback, forming a closed discovery loop that integrates theory, simulation, and validation and yields a comprehensive open-access dataset of successful and unsuccessful candidates. An educational mission trains students in AI-driven materials research and develops hands-on K–12 experiment kits.
Why it matters
A case applying AI as a "search engine" for materials design to the hard problem of superconductors. A useful read on the direction of U.S. (and German-partnered) investment for those tracking AI for Science, generative inverse design, and quantum/energy/medical-magnet materials.
FAQ
Why use AI for superconductors?
How is generative AI used in materials design?
Sources (primary)
Source: NSF Award Search (U.S. National Science Foundation, public domain). Amounts are the obligated amount. For privacy, we do not handle principal investigator names.
- NSF Award (original, official)
- NSF Award ID: 2522891