The order you claim credits in decides how much you get
Most credit programs are one-time and tier you by what you raised on the day you apply. Sequencing is worth more than volume — here is the order that works.
There is a version of this article that is just a list of programs. It would be useless. Lists of startup credits are everywhere, they all quote the largest number on each provider's marketing page, and founders who work through them still end up with a fraction of what they were eligible for.
The reason is structural. Almost every program on our database has two properties at once: it is one-time per company, and the tier you receive is decided by what you had raised on the day you applied. Put those together and the timing of each application becomes a financial decision, not an administrative one.
The mistake, with numbers
Cloudflare for Startups has three tiers. Bootstrapped teams with under $1M raised get $10,000. Companies under $5M raised that are backed by an affiliated partner get $100,000. Companies at $5M+ with an affiliated partner get $350,000.
Now consider a team six weeks away from closing a $6M Series A, working through a credits checklist on a Sunday afternoon. They apply while still technically unfunded. They receive $10,000. Cloudflare accepts first-time applicants only, so there is no second application after the round closes. That afternoon cost them $340,000 of published ceiling.
The same shape repeats across the database:
- Google Cloud pays up to $2,000 at the bootstrapped Start tier, and up to $350,000 at the AI-First tier for a Seed-to-Series-A company whose core technology is AI.
- HubSpot gives 30% off year one through entrepreneurial organizations, and 90% off year one to companies that raised pre-seed, seed or Series A and can be verified on Crunchbase or PitchBook. Series B and later get nothing.
- Neon gives self-funded teams up to $1,000, and venture-backed teams up to $200,000 in combined Neon and Databricks credits.
In each case the ratio between the tier you get while unfunded and the tier you get shortly after a round is somewhere between 20x and 175x. There is no other lever available to an early-stage company with that kind of multiplier attached to a decision about when to fill in a form.
The rule
If a round is realistically closing within three months, do not claim any tiered program that you could claim afterwards.
That is the whole rule. The complications are in identifying which programs are tiered, which are not, and which have their own clock that makes waiting expensive.
Three buckets
Sort every program you are considering into one of three groups.
Bucket 1: claim now, regardless. Programs that are not tiered by funding, or where the unfunded tier is the only tier. These carry no downside to claiming early, and several of them do useful preparatory work — a NVIDIA Inception membership or a JetBrains license forces you to get your company documentation, domain email and public website in order, which is exactly what the six-figure programs check later.
Typical members: NVIDIA Inception (free, no funding requirement, needs incorporation and a live site), JetBrains for Startups, ElevenLabs Grants (fewer than 25 employees, no funding test), Microsoft's entry tier, Google Cloud Start if you are nowhere near a round.
Bucket 2: claim after the round. Tiered programs where funding moves you up a bracket. Cloudflare, HubSpot, Neon, Modal, Datadog, Perplexity, the six-figure AWS Activate tiers, Google Cloud Scale and AI-First.
Bucket 3: claim when you can actually spend it. Programs whose value is set by consumption inside a fixed window rather than by the headline number. More on this below.
The trap in bucket 3
A credit you cannot burn before it expires is worth nothing, and the windows are shorter than founders assume.
| Program | Window | Detail that catches people |
|---|---|---|
| Cloudflare | 12 months | Hard expiry, no extensions available |
| DigitalOcean | 12 months | Core credits exclude GPU Droplets and dedicated inference |
| Neon | 12 months | Counts from acceptance, not from first use |
| MongoDB | 12 + 12 months | 12 months to activate the code, then 12 to spend |
| ElevenLabs | 12 months | Converts to the free plan when 10,000 credits remain |
If you are pre-product with no traffic, a $100,000 cloud credit that expires in twelve months is not a $100,000 asset. It is however much infrastructure you can genuinely consume in twelve months, which for a three-person team before launch might be $4,000. Meanwhile the same application, made after launch, might be worth the full amount.
MongoDB is the most forgiving design here — you have twelve months to activate and then twelve months to spend, and extensions can be requested case by case. Cloudflare is the least forgiving, and states plainly that extensions are not available.
Do the cheap preparation first
Every program that rejects you does so for one of a small number of boring reasons. Fix these before you apply to anything in bucket 2:
- Company-domain email. A personal address disqualifies you from Notion outright and weakens every other application.
- A live website that describes a product. Cloudflare requires a live, publicly accessible site. NVIDIA requires an official website. A coming-soon page does not count.
- A findable company profile. Cloudflare asks for an active presence on LinkedIn, X or GitHub. MongoDB asks for a company website and LinkedIn profile.
- Not looking like an agency. DigitalOcean excludes service businesses, consultancies and agencies. So do NVIDIA, MongoDB and ElevenLabs. If your website leads with "we build software for clients", you will be read as an agency even if you are not one.
These take a week. Losing a one-shot program takes an afternoon.
Find the door, not the form
The last piece: most of the large numbers are not self-serve at all. AWS routes its six-figure tiers through Activate Providers. HubSpot's 90% tier needs an approved partner or verifiable venture funding. Modal, Datadog and Perplexity run through investor networks. Cloudflare's two upper tiers require funding from an affiliated partner.
Which means the practical question is rarely "where is the application form" and usually "which of my existing relationships is an approved partner". Your investor, your accelerator, and — surprisingly often — your business bank are the doors. A single banking relationship can unlock a bundle of partner perks at once, including AI API credit that has no self-serve path anywhere.
The short version
- Fix the boring qualification signals first: domain email, live product site, findable company, no agency framing.
- Claim everything that is not tiered by funding, now.
- Hold every tiered program until after your round closes, if a round is close.
- Before claiming anything, check the expiry window against what you can realistically consume.
- For the six-figure tiers, work out which partner relationship you already have, rather than hunting for a form.
None of this requires a relationship we have and you do not. It requires reading the terms in the right order, which is what the database is for.