AWS vs Google Cloud vs Azure vs Cloudflare: cloud credits compared
Four programs, four completely different gates. Which one a seed-stage team should apply to first, and why the biggest published ceiling is rarely the best offer.
The four largest cloud credit programs publish ceilings between $150,000 and $350,000. Those numbers are real, and they are also close to irrelevant for deciding where to apply, because the ceilings sit behind gates that differ far more than the amounts do.
Here is what actually separates them.
The published ceilings
| Program | Ceiling | Bootstrapped entry | Window |
|---|---|---|---|
| Google for Startups Cloud | Up to $350,000 | Up to $2,000 (Start tier) | Up to 2 years |
| AWS Activate | Up to $200,000 | Small self-serve tier | Typically up to 2 years |
| Cloudflare for Startups | $10,000 – $350,000 | $10,000, no minimum funding | 12 months, no extensions |
| Microsoft Founders Hub | Up to $150,000 | Entry tier, no funding needed | Typically 1 year per tier |
Read that table twice and the interesting column is not the ceiling. It is the bootstrapped entry.
Google Cloud: the cleanest tier ladder
Three tiers, and unusually transparent about the shape of each.
Start gives up to $2,000 in Google Cloud and Firebase credits to a technology startup founded within the last five years that has no institutional investor and has not taken Google Cloud credits beyond the free trial. Self-serve.
Scale covers 100% of your Google Cloud and Firebase usage up to $100,000 in year one, plus 20% of usage up to another $100,000 in year two. It requires equity funding from an institutional investor, pre-seed through Series A, and if you are Series A the round must have closed in the last twelve months.
AI-First adds up to $150,000 on top of Scale, reaching $350,000 in total, for companies whose core technology is AI. Seed to Series A, founded within the last ten years, and — the condition that catches people — you must not have received more than $5,000 in Google Cloud credits before.
That last clause is the reason to think before claiming Start. Taking $2,000 today does not disqualify you from AI-First later, since the bar is $5,000. But it is the tightest "prior credits" test among the four programs, and it deserves a check rather than an assumption.
Note also what Scale actually is: coverage of usage, not a lump sum. Year two covers 20% of spend up to $100,000, meaning you would have to spend $500,000 to consume the whole second-year allowance. Treat the headline as a ceiling on generosity, not on your bill.
AWS Activate: the widest ceiling, the most gated
Up to $200,000, with credits redeemable on third-party models through Amazon Bedrock and additional credits available for AI startups ready to scale.
The self-serve tier is small but genuinely open: an incorporated company with a company domain and a working product or prototype can claim it. Everything above that runs through Activate Providers — accelerators, incubators, VCs and startup-friendly banks. There is no route from a cold application to the six-figure tier.
So the practical question for AWS is never "should I apply" but "which Activate Provider am I already connected to". If the answer is none, fixing that is the higher-leverage task.
Microsoft Founders Hub: the best bootstrapped on-ramp at scale
Up to $150,000 usable across Azure, GitHub and other eligible Microsoft platforms. The top of that range is reserved for VC-backed companies coming in through Microsoft's investor network.
What makes it distinctive is the entry tier: genuinely self-serve, no funding requirement, needing only a verified incorporated startup that is new to Azure. Among programs with six-figure ceilings, this is the one with the most usable door for a company that has raised nothing. It also bundles AI model access and go-to-market support rather than credits alone.
Cloudflare: the most predictable, and the strictest clock
Three tiers, stated plainly: $10,000 bootstrapped with no minimum funding, $100,000 under $5M raised with an affiliated partner, $350,000 at $5M+ with an affiliated partner. The upper tiers add an account manager, priority support and regular Solutions Architect office hours.
General requirements are equally clear: founded in the last ten years, actively building a technology product, a live public website, an active presence on LinkedIn, X or GitHub, a for-profit business with a company email, and first-time applicants only.
And the hardest constraint in the whole comparison: credits expire after twelve months or when consumed, whichever comes first, and extensions are not available.
The combination of "first-time applicants only" and a 20x tier jump makes Cloudflare the single program where premature application costs the most. If a round is close, wait.
Which to apply to first
If you have raised nothing: Microsoft's entry tier and Cloudflare Tier 3 are the two real options with meaningful money attached, plus Google Cloud Start if $2,000 is useful now and you have confirmed the $5,000 prior-credits ceiling for AI-First. If a round is anywhere on the horizon, take Microsoft and hold Cloudflare.
If you have just closed a round with an institutional investor: Google Cloud Scale or AI-First is usually the strongest single offer, because coverage-of-usage means you cannot leave value stranded the way a lump sum can. Then Cloudflare at the tier your raise unlocks, then AWS through whichever Activate Provider your investor is.
If you are AI-native: the AI-First tier at $350,000 is the largest published number available to a Seed-to-Series-A company anywhere on our database. It is worth arranging your application around, including waiting for the round that qualifies you.
The mistake to avoid
Do not apply to all four in one sitting. They are not lottery tickets. Three of the four are one-time per company, all four tier you by your position on the day you apply, and Cloudflare in particular will hold you to your first attempt forever.
Pick the one that fits your stage, claim it, and put the rest in a calendar entry dated the week after your round closes. The full terms for each are in the database, and the general method is in the sequencing playbook.