Entry Track

The CU System Sprint Grant is a small, fast-turnaround internal award supporting faculty who want to develop and implement AI-driven pedagogical strategies within a single course. At up to $20,000 with an explicit allowance for one course buyout, it is designed around the real obstacle facing most faculty experimenting with generative AI in teaching - not equipment or data, but protected time. The single-course framing is a deliberate scoping decision and applicants should respect it: proposals aimed at curriculum-wide or program-level transformation are a poor fit, while a concrete plan to redesign one course around AI tools and measure the effect on student learning outcomes matches what the program is asking for. Eligibility extends to tenured and tenure-track faculty as well as full-time and instructional series faculty across all CU campuses, which is broader than many internal research awards and deliberately includes teaching-focused appointments who are often closest to the pedagogical questions at issue. This is the second of three planned cycles, so faculty who miss the October 16, 2026 deadline should expect a third round. For a grants database this entry is narrower in eligibility than a federal or foundation program - it is open only to CU faculty - but it is a useful example of the institutional micro-grant tier that has grown rapidly as universities respond to generative AI in the classroom, and comparable programs now exist at many university systems. Faculty at other institutions who find this relevant should check whether their own provost or academic affairs office runs an equivalent scheme, since these internal AI teaching grants are frequently under-advertised and undersubscribed relative to external competitions.

The EuroHPC Joint Undertaking's AI Factories give European AI companies free access to some of the largest GPU systems in the world, and the Industrial Innovation access route is the one built specifically for SMEs and startups rather than for academic consortia. It runs three tracks of increasing scale. Playground Access provides limited entry-level resources on a first-in-first-out basis with access granted within two working days and onboarding support for users new to HPC - it exists to let a team benchmark and port before committing to a proposal. Fast Lane Access serves experienced HPC users needing up to 50,000 GPU hours. Large Scale Access covers applications requiring more than 50,000 GPU hours, allocated for up to one year, with cut-off dates running through 4 January 2027. All three are open and free of charge to AI SMEs including startups pursuing innovation, which is the critical distinction from the separate AI for Science and Collaborative Projects mode aimed at publicly funded research; other industrial users can still access the machines but on commercial pay-per-use terms. For a European AI startup, this is close to the largest non-dilutive compute resource available anywhere, and the practical route in is almost always Playground first, then a Large Scale proposal built on the benchmarking data that Playground produced. Applications are submitted through the EuroHPC access portal.

Future CoRe is NSF CISE's consolidated core research solicitation, merging eleven previously separate programmes into a single entry point, and it is the broadest available route to funding for foundational artificial intelligence and machine learning research in the United States. It supports foundational computing research and education with strong emphasis on AI and machine learning - theory, algorithms, systems and applications - alongside the rest of the computing research portfolio. The structural change that matters most to applicants is the move away from hard deadlines. Proposals may be submitted at any time, but target dates have been added to organise review into batches, with target dates of 10 September 2026 and 4 February 2027 for the relevant tracks. In practice a proposal submitted well away from a target date waits longer for a panel, so the target dates function as soft deadlines for anyone who needs a timely decision. The solicitation covers small, medium and large project classes with different ceilings; the figures here reflect the medium class, where most AI and machine learning foundations work is funded. For researchers whose work does not map onto a named NSF AI programme - the AI Institutes, the AI infrastructure hubs, or the geosciences and health collaborations - Future CoRe is the default home, and the eleven constituent programmes within it offer meaningful choice over which review community sees the proposal.