Free compute is out there.
We keep the list current.
A live directory of GPU, token, and HPC programs — split for scientists and startups — with the part nobody else tells you: how to actually use the grant. Get API credits → plug the key into MegaBrain Science as your own model. Get a cloud or HPC allocation → attach the kernel to it over SSH and run there instead of on your laptop.
Browse by who you are
Pick your track, then filter by resource. Sorted newest-first — subscribe below to hear when new listings open.
| Program | What you get | Who · where | Deadline | Listed | Use | |
|---|---|---|---|---|---|---|
| Research GrantsNewNebius | GPU credits + Token Factory tokens (per proposal) | PhD students, postdocs, faculty at academic/nonprofit orgs | Annual · 2026–27 cycle, decisions ~Q3–Q4 2026 | Jul 5, 2026 | SSH → HPC | Apply |
| Regular & Extreme-Scale AccessNewEuroHPC JU | Large node-hour allocations, 1-year | EU / Horizon-Europe researchers & industry | Cut-offs: Sep 4, 2026 (Regular) · Oct 19, 2026 (Extreme) | Jul 4, 2026 | SSH → HPC | Apply |
| AI for Science ProgramNewAnthropic | Up to $20,000 in Claude API credits (6 months) | Institutional researchers · bio & life sciences focus | Rolling · first Monday monthly | Jul 2, 2026 | BYOK | Apply |
| Academic Grant ProgramNewNVIDIA | Up to 30,000 H100 GPU-hours (or hardware tracks) | PhD-granting faculty · must use NVIDIA models/software | Periodic CFP · currently closed | Jun 28, 2026 | SSH → HPC | Apply |
| Modal for AcademicsNewModal | Up to $10,000 in serverless-GPU credits | Researchers & academics | Rolling | Jun 25, 2026 | SSH → HPC | Apply |
| Development & Benchmark AccessEuroHPC JU | Node-hours on LUMI, Leonardo, MareNostrum 5, JUPITER… | Academia & industry in EU / Horizon-Europe countries | Rolling · monthly cut-offs | Jun 22, 2026 | SSH → HPC | Apply |
| Research GrantLambda | Up to $5,000 in cloud GPU credits (B200/H100/A100) | AI researchers | Rolling | Jun 18, 2026 | SSH → HPC | Apply |
| ALCCDOE (US) | 10–30% of NERSC / ALCF / OLCF resources | Researchers on DOE-mission problems | Annual · next call ~late 2026 | Jun 14, 2026 | SSH → HPC | Apply |
| ACCESS AllocationsNSF (US) | Explore ≤400k / Discover ≤1.5M / Accelerate ≤3M credits; Maximize uncapped | US-based researchers & educators · no cost, no NSF award needed | Rolling (Explore/Discover/Accelerate) | Jun 10, 2026 | SSH → HPC | Apply |
| External Researcher Access ProgramAnthropic | $1,000 in Claude API credits | AI safety / alignment researchers | Rolling · first Monday monthly | Jun 3, 2026 | BYOK | Apply |
| INCITEDOE (US) | ~0.5–2M node-hours on Frontier & Aurora | Worldwide · academia, labs, industry (capability computing) | Annual · next call ~April 2027 | May 30, 2026 | SSH → HPC | Apply |
| Cloud Credit for ResearchAWS | Students up to $5,000; faculty/staff uncapped | Faculty, research staff, grad/PhD students | Rolling · 90–120 day review | May 28, 2026 | SSH → HPC | Apply |
| Cloud Research CreditsGoogle Cloud | GCP compute credits (amount set per proposal) | Faculty / PhD / postdoc, ~50 eligible countries | Rolling | May 20, 2026 | SSH → HPC | Apply |
| TPU Research Cloud (TRC)Google Research | Free access to 1,000+ Cloud TPUs (temporary) | Anyone · must share results publicly | Rolling · invitations on a rolling basis | May 15, 2026 | SSH → HPC | Apply |
| Researcher Access ProgramOpenAI | Up to $1,000 in API credits (12 months) | Early-stage researchers, API-supported countries | Rolling · reviewed quarterly | May 12, 2026 | BYOK | Apply |
Never miss a fresh grant
We add new compute programs as they open. Get an email when a listing you care about lands — no spam, unsubscribe anytime.
Know a grant or startup program?
Running a compute grant, or found one we're missing? Send it in and we'll review it for the directory.
A grant is only useful if you can spend it
Won API credits from OpenAI, Anthropic, Google, Mistral, or another supported provider? Enter that key in MegaBrain Science as your model. The agent drives your analysis on the granted budget — your data still stays on your machine, only the model calls go out.
Got GPU credits or a supercomputer allocation (NSF ACCESS, EuroHPC, DOE, TRC, a cloud grant)? Point the MegaBrain Science kernel at it over SSH (MB_KERNEL_RUNTIME=ssh) and run the heavy compute there — the workbench comes to your allocation, not the other way around.
Verified Jul 8, 2026. Programs, amounts, eligibility, and deadlines change — always confirm on the program's own page before applying, and tell us if something here is stale. Amounts are shown only where the program publishes them; "not published" means the figure is set per application.
Win the compute. Do the science.
Download MegaBrain Science, bring your granted key or HPC allocation, and run your first verified analysis.