Most founders write a grant budget the same way they write a pitch deck: optimistic, rounded, and built after the idea rather than before it. Then the award letter arrives with a line you skipped past during the application, the one that says the money can only go toward certain categories, and now you're explaining to a program officer why half your budget was salaries when salaries weren't eligible. This happens constantly, and it's avoidable if you build the budget around the restrictions instead of bolting the restrictions on afterward.
What's usually eligible and what isn't
Grant makers split spending into categories for a reason that has nothing to do with bureaucracy: they're trying to fund the thing you're building, not your general cost of existing. Capital equipment, contractor fees for defined deliverables, and direct project costs are the easiest to get approved because an auditor can point to them and see exactly what was bought and why. Your own founder salary, overhead like rent and software subscriptions, and travel are the categories that get cut or capped, because they're hard to tie to a specific output and easy to pad.
- Usually eligible: contractor and developer payments tied to a milestone, equipment and tooling bought for the project, audits, and direct technical costs like gas fees or hosting for a specific deliverable.
- Often restricted or capped: founder salary (sometimes allowed, often capped at a modest percentage of the award), general overhead, travel and conference costs, and anything that reads as 'operating expenses' rather than 'this grant's work'.
- Programmes that fund pure technical output, like the The Graph grants program or NLnet Foundation Grants, lean hardest toward contractor and development costs and are the least forgiving of salary or overhead line items.
On-chain programmes behave a little differently because the grant itself is denominated in the thing you're building with. Base Builder Grants pay out 1 to 5 ETH, and Starknet Seed Grants pay up to $25,000 in STRK. Spend that on contractor work or infrastructure and nobody blinks. Try to convert a chunk of it to cash and call it a founder draw, and you've just created the exact conversation you don't want to have with the foundation's grants committee.
Reimbursement versus advance: this decides whether you need cash on hand first
The category rules matter less than this single mechanical question: does the programme pay you before you spend, or only after you've already spent and can show the receipt? Reimbursement-based grants require you to float the cost yourself, sometimes for months, which is a real problem if your budget assumed the grant was the cash. Advance-funded programmes, common in the capped, milestone-based crypto grant world, release a tranche up front and the rest on delivery, which is far friendlier to a team with no runway.
Trailblazer 2.0 from the Arbitrum Foundation is a useful example of the milestone structure: the pool is $1M total, capped at $10k per submission, which forces you to scope a deliverable small enough to fund and ship inside one tranche rather than planning a six-month runway against it. See Trailblazer 2.0 for the submission terms. Compare that to SCRA Startup & Acceleration Grants, which pays up to $25,000 at the Startup stage and up to $50,000 at the Acceleration stage, reviewed on a rolling basis, meaning you can time your application to when you actually need the capital rather than waiting for a fixed window.
Unspent money and misspent money are different problems
If you don't spend the full award, most programmes want it back, not quietly absorbed into your next quarter. If you spend it on something outside the approved categories, the consequence ranges from a request to reallocate and resubmit receipts, up to being asked to return the misspent portion and, for repeat offenders, being locked out of future rounds from that same funder. Neither outcome is catastrophic on its own, but both cost you something more valuable than money at the early stage: the relationship with a funder who might fund your next round too.
This is why rolling and no-deadline programmes are more forgiving of a budget that needs revision. Starknet Seed Grants has no application deadline and reviews in rounds on an ongoing basis, and SCRA's rolling basis works the same way: you can go back and adjust scope before the money moves, rather than discovering a mismatch after a fixed-deadline committee has already approved your original numbers.
Writing a budget that survives review
- Match your biggest line item to the programme's stated purpose. A grants committee reading a budget where 70% goes to 'contractor development costs' for a technical grant like SSV Network Grants Program (a $5M pool) will approve it faster than one where 70% is salary.
- Check whether the award tier changes what's eligible. NC IDEA Grants offers $10,000 at MICRO and $50,000 at SEED, both non-dilutive, and the categories reviewers expect to see scale with the tier, not just the total.
- If your work is genuinely capital-heavy, say equipment or lab-grade tooling, look at programmes built for exactly that. Advanced Industries Accelerator Grants from Colorado OEDIT varies by grant type but typically runs up to $250,000, and equipment costs that would get flagged elsewhere are the whole point here.
- Name your deliverables, not your hours. A budget line that says '$8,000 for audit of settlement contract' survives review. A line that says '$8,000 for founder time' invites questions, especially against a cap like BNB Chain Grants' up to USD 200K per project, where reviewers expect the money mapped to specific technical milestones.
- Build in a buffer line and label it, rather than padding every category by 15% silently. Reviewers at NLnet Foundation Grants, which funds between €5,000 and €50,000 for a first grant and has its next deadline on November 3rd 2026 at 12:00 CET, read enough applications to notice an honest contingency line versus padding disguised as categories.
- If the programme's funding mechanism is tiered or undisclosed, like Circle Developer Grants, which ties funding to meaningful usage rather than a published figure, budget conservatively for the first tranche and treat anything beyond that as a stretch goal you earn, not a number you plan against.
The founders who get burned aren't the ones who guessed wrong about amounts. They're the ones who assumed every dollar was the same color. It isn't. Read the eligible-use clause before you build the budget, not after you've been awarded the money, and check whether the programme pays in advance or only reimburses, because that single fact decides whether you can actually afford to win.