The order most founders use them in
- Credits first. Cloud and GPU credits are the fastest to get and cover the bill you already have. They are not cash, and they expire.
- Equity-free grants for people. Small, fast grants to individuals can fund the months before you have a company or a product.
- SBIR and STTR. The largest non-dilutive source for US small businesses, but slower, and only for genuinely risky R&D.
- Accelerators and equity. Usually alongside or after the above, once you know what you are building.
For the full picture, including tax relief and how the pieces interact, see the AI startup funding stack.
NSF SBIR and STTR
NSF's America's Seed Fund says it awards $200+ million a year to about 400 startups across the United States, takes no equity, and leaves founders owning their company and intellectual property [source]. It funds work across nearly all technology areas, but not straightforward engineering or incremental product development: the technical feasibility of the work has to be unproven at the start [source].
NSF describes the programme as up to $2 million over 42 months or more. Phase I is up to $305,000 over 6 to 18 months, for small businesses with an official Project Pitch invitation, and Phase II is up to $1,250,000 over 24 months for Phase I awardees [source] [source].
The eligibility rules that matter most for AI startups: fewer than 500 employees, at least 51% owned and controlled by US citizens or permanent residents, the principal investigator more than 50% employed by the company during the project, and all R&D performed in the United States [source]. Businesses majority-owned by venture capital operating companies, hedge funds or private equity firms are not eligible [source].
You start with a short Project Pitch, which needs no government registrations. Read our full NSF SBIR guide and how to write the proposal.
NIH SBIR and STTR for health and clinical AI
If your AI product is in health, the NIH is the bigger door. NIH says it sets aside more than $1.4 billion of its research and development funding for its small business programmes [source]. Its published guideline amounts, which are not a promise of what any one institute will award, are $335,937 for Phase I (six months to two years) and $2,239,578 for Phase II (one to three years) [source].
NIH's eligibility rules follow the same small-business pattern, including US ownership and US-based work [source]. See NIH SBIR and STTR for health and clinical AI for the details.
Cloud and GPU credits
The large cloud programmes are open to startups worldwide, including US companies, and are the quickest money you will see:
- Google for Startups Cloud Program: AI track: Up to $350,000 in Google Cloud credits over 2 years (AI-first startups).
- AWS Activate Credits: Up to $5,000 (self-funded) or up to $200,000 (with an Activate Provider Org ID); invite-only $200K+ for AI startups at scale.
- Microsoft for Startups: Up to $150,000 in Azure credits.
Headline caps are maximums, and they often unlock in stages. Read the funding stack guide for how credits sit next to grants, and our compute-credits guides for the detail on each programme.
Equity-free money for individuals
Before there is a company, a few programmes fund people directly:
- Thiel Fellowship: $250,000 grant over two years, plus support from the foundation's network. The foundation says fellows skip or stop out of college to build something [source].
- Emergent Ventures: fellowships and grants for ambitious projects. Its apply page says it no longer accepts AI-written applications [source].
Z Fellows is close to this but is not equity-free: Optional $10,000 investment ($1B valuation cap); one week with founders of billion-dollar companies.
These are small next to SBIR, but they are fast and need no company. Both Thiel and Emergent Ventures were taking applications when we checked on 18 November 2026, with no deadline published.
If you are not a US company yet
NSF's ownership and work-location rules mean a company that is not majority owned by US citizens or permanent residents, or that does its R&D abroad, will not qualify, whatever the product [source]. Setting up a US entity is a legal and tax decision, not a shortcut around those rules. If you are deciding where to incorporate, get legal and tax advice before you rely on any US-only funding route, and see the grants list for calls open to other countries.
