SBIR vs. STTR for AI Startups: Which Program Fits

The two programs, defined

The Small Business Innovation Research (SBIR) program funds R&D performed primarily by the small business itself. The small business is the prime recipient and must perform the substantial majority of the work (specifically, at least 67% in Phase I and at least 50% in Phase II). SBIR's goal is to move promising technology from research through development to commercialization, with the small business as the principal vehicle.

The Small Business Technology Transfer (STTR) program funds cooperative R&D between a small business and a qualified research institution — typically a university, a federally funded research and development center (FFRDC), or a non-profit research institution. The small business is still the prime recipient, but STTR specifically requires that the research institution perform at least 30% of the work and the small business perform at least 40%. The remaining 30% can be split between them as the team chooses.

STTR exists to pull technology out of university labs into commercial hands via a formal small-business partnership. SBIR assumes the small business already has the technology and needs funding to develop and commercialize it. The difference in intent matters for how reviewers read proposals.

Advantages of SBIR for AI startups

Full control over the work

The small business does the majority of the work, controls the engineering, and owns execution risk end-to-end. For an AI startup where the founders want to move fast and iterate tightly on product, SBIR fits the normal operating rhythm. No research-partner coordination overhead. No university procurement calendar. No sub to manage who lives on a different cadence than the startup.

Larger pool of topics and agencies

Every SBIR agency runs a Phase I and Phase II cycle. Total topic count across all eleven participating agencies runs into the hundreds per year. STTR topic counts are smaller at each agency, typically about 10-15% of the SBIR volume. If your technical area is narrow, SBIR offers more targets to hit.

Simpler proposal logistics

One entity, one PI, one scope. No IP agreement to negotiate with a university tech-transfer office. No research partner's indirect-cost structure to absorb. No faculty schedule to coordinate around. First-time proposers in particular benefit from the reduced coordination complexity.

Faster commercialization signal

Reviewers reading SBIR proposals expect to see clear commercialization intent — real customers, a real path to revenue, a real business model. For AI startups already hunting commercial product-market fit, SBIR's commercialization emphasis aligns naturally with what the startup is already trying to prove.

Full founder-driven IP

IP created under SBIR is fully the small business's, subject only to the government license. No university tech-transfer office has rights. For a startup planning to raise venture capital, SBIR produces a cleaner IP stack than STTR joint-IP arrangements.

Advantages of STTR for AI startups

Access to university research depth

If your work builds on university research — a specific lab's methods, a professor's algorithms, specialized equipment — STTR is the natural vehicle. You get formal funded access to the research partner, which is hard to replicate with a standalone SBIR.

Credibility boost on technical review

A named research partner with relevant publications and credentialed faculty adds technical weight to the proposal. Reviewers tend to read STTR proposals with somewhat different eyes — more "is the science sound" and somewhat less "does this team ship." For AI work with genuine research novelty, STTR can win proposals that would lose on SBIR review.

Built-in talent pipeline

Graduate students and postdocs contributing to STTR work often become hires to the startup later. The research partnership becomes an early-access recruiting pipeline that is harder to build any other way.

Smaller, less competitive pool

STTR topic counts are smaller but so are submission volumes. If you can match a topic cleanly, the competition per topic can be less intense than on a hot SBIR topic. This is not universal — top topics at top agencies are competitive — but at the aggregate level STTR tends to be less crowded.

Federal lab partnerships under STTR

STTR allows FFRDCs and federal research labs (Los Alamos, Argonne, NREL, MITRE, Lincoln Lab, and more) to serve as research partners. For AI work with direct mission relevance, partnering with a federal lab through STTR creates a warm relationship with the agency that funds both the lab and the STTR — a relationship that converts to Phase III far more reliably than a cold academic partnership.

Side-by-side comparison

Which agencies run STTR

STTR participation is narrower than SBIR. Only five agencies operate STTR programs: DoD, HHS (primarily NIH), NSF, DOE, and NASA. The other SBIR agencies (DHS, DOT, ED, EPA, USDA, DOC) do not run STTR. If your target agency is outside the STTR-participating five, the SBIR path is your only SBIR/STTR option at that agency.

Within the five participating agencies, STTR volume is uneven. NIH has the largest STTR program by count of awards. NSF runs a well-organized STTR alongside its SBIR. DoD's STTR is smaller than its SBIR but still substantial across components. DOE STTR is smaller but tightly integrated with national-lab research. NASA STTR is the smallest but often tied to specific mission directorates.

Decision framework for an AI startup

Choose SBIR when...

• Your AI work is engineering and commercialization of known techniques, not research novelty.
• Your founders can execute the technical scope themselves without university infrastructure.
• You want IP that is fully clean for a future equity raise.
• You want to move fast without research-partner coordination overhead.
• Your target agencies include DHS, DOT, USDA, EPA, ED, or DOC (no STTR option).
• Your topic pool at the target agency is broader than the STTR topic pool.

Choose STTR when...

• Your technology builds on specific university research, faculty expertise, or specialized infrastructure.
• You already have a research relationship with a qualified institution.
• You want formal pipeline access to graduate students or postdocs as future hires.
• Your target agency is a good STTR agency (NIH, NSF, DoD, DOE, NASA) and offers relevant topics.
• Reviewer emphasis on research rigor would favor your proposal over pure commercialization-focused SBIR review.
• You can negotiate a clean IP agreement with the research partner up front.

Run both when...

Nothing prevents a firm from running SBIR and STTR proposals simultaneously at different agencies — or even at the same agency if topic fit justifies it. Some of the strongest federal AI portfolios mix both programs deliberately. Most first-time firms should pick one and execute cleanly before attempting both.

Practical pitfalls

The wrong-fit research partner

The worst STTR proposals name a research partner whose publications have nothing to do with the topic, listed only to satisfy the 30% workshare requirement. Reviewers see through this in the first read. If the research partner is not a genuine technical contributor, SBIR is the better vehicle.

The undocumented IP agreement

STTR requires an IP agreement between the small business and the research institution. Many first-time proposers leave this for after award and discover that the university tech-transfer office moves on a different timeline than the funding clock. Negotiate the IP agreement before submission. Attach it if the solicitation allows. This is the single biggest post-award failure mode for STTR teams.

The over-budgeted research partner

University indirect cost rates can consume a material chunk of the research-partner workshare if not managed. Negotiate a reduced IDC rate where the institution permits, or structure the partner's scope to minimize IDC-eligible labor. Many research institutions will negotiate; they do it routinely.

The phantom commercialization plan

STTR reviewers read both for research rigor and for commercialization. Academic partners sometimes pull the proposal toward a research-only tone. The small business must protect the commercialization narrative — what market, what revenue path, what Phase III transition looks like. Strong STTR proposals balance both lenses.

The underutilized Phase III path

Both SBIR and STTR Phase IIIs are sole-sourceable to the small business without a competitive acquisition. This is the biggest strategic asset of either program. Firms that fail to plan for Phase III from Phase I forward leave enormous value on the table.

Our recommendation for most AI startups

For an AI startup whose founders are capable engineers commercializing known techniques on mission-specific data, SBIR is the default recommendation. Lower coordination cost, broader agency pool, cleaner IP, faster rhythm. Most of the federal AI wins we see in 2026 are SBIR-driven.

STTR is the right call when research depth is a real differentiator — when the startup's edge comes from methods, algorithms, or infrastructure that a specific lab or faculty member actually contributes to. In that situation, STTR outperforms SBIR because the reviewer reading the proposal sees the research provenance and gives technical credit the SBIR framing would not support.

Both programs can be stacked: a startup can win a Phase I SBIR at one agency and a Phase I STTR at another, running in parallel. The founder's time is the binding constraint, not the SBA's rules.

For more on win-rate math and agency selection, see our agency-level analysis. For the broader federal AI market picture, our trends article covers the full picture.

FAQ

What is the key difference between SBIR and STTR?

STTR requires a formal partnership with a nonprofit research institution that performs at least 30% of the work. SBIR allows the small business to do all or nearly all of the work, with no mandatory research partner.

Are the dollar amounts different?

Practically no. Phase I and Phase II ranges are similar across SBIR and STTR at the same agency. Some agencies size STTR slightly larger to accommodate research-partner IDC, but the difference is modest.

Is STTR easier to win?

Win rates are similar at the aggregate. Topic specificity, proposal quality, and incumbency matter more than SBIR vs. STTR. The STTR pool is smaller but reviewers expect stronger research rigor.

Does an AI startup need a research partner?

Only if the research partner is a genuine technical contributor. If the startup's edge comes from engineering and commercialization, SBIR fits. If it comes from university-anchored methods or infrastructure, STTR fits.

Who owns the IP?

In both, the small business holds primary rights with a government license for federal use. STTR requires an IP agreement with the research institution up front. Done well, it clarifies rather than complicates.

Can one firm do both?

Yes. Many firms run SBIR and STTR proposals concurrently at different agencies. The binding constraint is founder time, not program rules.