We went to The Exchange in San Diego this week and came home with a slide that stopped us cold.
The U.S. government invests $4 billion per year in startups through the SBIR/STTR programs. They take zero equity. Zero IP. They fund the R&D, you keep everything.
We did the research. Here’s what we found.
The Landscape
Eleven federal agencies participate in SBIR (Small Business Innovation Research) and STTR (Small Business Technology Transfer). Each runs its own program with its own focus areas and funding levels. The SBA coordinates. Congress just reauthorized the whole thing through 2031.
The programs fund about 4,000 companies per year. Phase I is proof of concept - typically $100K-$300K over 6-12 months. Phase II is full development - $750K-$1.8M over 24 months. Phase III is commercialization, funded by the market or by federal contracts.
Every dollar is non-dilutive. You keep your cap table clean.
The Programs That Matter for AI
NSF SBIR - The Best Fit
The National Science Foundation runs “America’s Seed Fund” - the most startup-friendly of all the SBIR programs.
- Phase I: Up to $300,000
- Phase II: Up to $1.25 million
- Budget: $174 million/year, ~400 companies funded
NSF funds nearly all technology areas. AI is an explicit topic with eight sub-categories:
- Conversational AI Technologies - AI that talks you through something
- Language-Based AI Technologies - AI that reads and explains
- Technologies for Trustworthy AI - safe, fair, transparent, explainable
- Novel AI Hardware - edge devices, neuromorphic computing
- Sustainable AI for Low Resource Environments - AI that works without massive compute
They also fund Advanced Systems for Scalable Analytics - machine learning, decision support, optimization, data management, and visualization.
The application process starts with a Project Pitch - about 3 pages, submitted online anytime. NSF responds in 1-2 months. If they invite you, you submit a full proposal. Funding decision comes ~6 months later.
Eligibility is straightforward: Small business (<500 employees), at least 50% U.S.-owned, all work in the U.S., not controlled by VC/PE. The principal investigator needs to be employed by the company at least 20 hours a week.
NIST SBIR
The National Institute of Standards and Technology runs a smaller program through the Department of Commerce.
- Phase I: ~$100,000
- Phase II: Up to $400,000
- Budget: $15 million/year
NIST’s focus is measurement science and standards - AI safety, AI measurement, semiconductors, cybersecurity. They awarded $18 million to small businesses in 2025, with AI as a priority. The FY2026 solicitation hasn’t been released yet.
The angle here is narrower: AI that measures something. AI that establishes standards. If your technology touches trustworthiness, accuracy, or quality measurement, NIST cares.
Department of Defense / DARPA
The largest SBIR budget by far - $2.3 billion per year. But it’s oriented toward national security.
- Phase I: $50,000-$250,000
- Phase II: $800,000-$1.83 million
- Contract-based (not grants)
Priority areas: AI/autonomy, cybersecurity, 5G, quantum, hypersonics, space. DARPA funds “dual-use” technology - commercial AI with defense applications. AFWERX (Air Force) is the most startup-friendly branch.
The fit is narrower unless your AI serves a defense mission. But the sheer scale means there are opportunities for AI analytics, decision support, and data systems that have both commercial and government use cases.
→ sbir.gov
DOE SBIR/STTR
The Department of Energy runs a substantial program - $315 million/year - focused on energy and scientific computing.
- Phase I: $200,000-$250,000
- Phase II: $1.1-$1.6 million
DOE just announced $293 million for the “Genesis Mission” - AI for national science challenges. But the focus is energy infrastructure, data centers, grid management, clean energy. Unless your AI serves an energy mission, this isn’t the right door.
How the Money Flows
| Program | Phase I | Phase II | Annual Budget |
|---|---|---|---|
| NSF | $300K | $1.25M | $174M |
| NIST | $100K | $400K | $15M |
| DOD | $50-250K | $800K-1.83M | $2.3B |
| DOE | $200-250K | $1.1-1.6M | $315M |
| HHS/NIH | $275K+ | $1.83M | $1.2B |
| NASA | $150K | $1M | $174M |
| USDA | $125-175K | $600K | $42M |
| Dept of Ed | $250K | $1M | $10M |
All non-dilutive. All 0% equity.
What This Means for Us
We’re building an AI system that reads a company’s financial data, generates an explainable forecast, and guides a human through reviewing every assumption in conversation. The technology sits at the intersection of several NSF topic areas - conversational AI, language-based AI, trustworthy/explainable AI, and decision support systems.
We’re a bookkeeping and CFO firm with 500 companies on the platform. The IFM framework has been validated across all of them. The Charting Tool already ingests and visualizes financial models. The next step - automated forecasting with conversational review - is exactly the kind of early-stage R&D that SBIR was designed to fund.
We’re drafting our NSF Project Pitch — for a $100/month product with free setup that makes financial forecasting accessible to every small business. We’ll share the pitch publicly, because we think the process is worth documenting. If you’re building AI and you haven’t looked at SBIR, start at sbir.gov.
Getting Started
If any of this applies to you:
- Take the NSF Project Fit Assessment - seedfund.nsf.gov/apply/the-basics/ - 10 minutes
- Register on SAM.gov - required for any federal grant, takes weeks, start now
- Sign up for newsletters:
- Browse open solicitations at sbir.gov
The money is there. The process is documented. The hardest part is knowing it exists.