SBIR vs STTR: a decision tree for AI firms

The core structural difference

SBIR and STTR run in parallel and look similar from the outside. Same phases, similar dollar ranges, similar solicitations. The structural difference is STTR's mandatory research partner: at least 30% of the STTR award must go to a qualified research institution (university, federally funded R&D center, or nonprofit research lab). In SBIR, the small business can execute 100% in-house.

That 30% minimum creates a very different project structure. STTR is built for technology transfer from academic research labs into commercial small businesses. When the 30% partner brings distinctive IP, data, or methods, STTR is leverage. When the 30% partner is a formality, STTR is overhead.

Quick decision tree

flowchart TD
    A([Your AI firm has a technical project]) --> B{Genuine university lab partnership?}
    B -->|No — pure in-house execution| C[SBIR]
    B -->|Yes — they bring real IP or data| D{Partner brings distinctive value?}
    D -->|No — partner is a formality| C
    D -->|Yes — unique dataset or algorithm| E{PI primarily employed by university?}
    E -->|Yes — faculty PI| F[STTR — dual affiliation OK]
    E -->|No — industry PI| G{30% budget to partner workable?}
    G -->|Yes| H[STTR — negotiate IP first]
    G -->|No| C
    style C fill:#3b82f6,color:#fff,stroke:#3b82f6
    style F fill:#7c3aed,color:#fff,stroke:#7c3aed
    style H fill:#d97706,color:#fff,stroke:#d97706
  

When STTR helps AI firms

STTR is distinctly advantageous in several scenarios:

When STTR hurts AI firms

Topic-level considerations

Some solicitations publish both SBIR and STTR versions of similar topics. In those cases the firm can choose. Other topics are exclusively one or the other. Check the topic designation before investing in proposal development.

DoD, NIH, NSF, and DOE all run both SBIR and STTR programs. The dollar ranges are similar: STTR Phase I typically $150-200K, STTR Phase II typically $1-1.7M. Competition rates for STTR are sometimes lower than SBIR at the same agency, though this varies by cycle.

Data rights in STTR

STTR data rights follow the same basic structure as SBIR — 20-year protection period, limited rights for the government. The wrinkle is that the research institution has negotiated rights to the IP as well. The STTR awardee and the research institution must have a written IP agreement that allocates rights between them. This agreement should be finalized before the proposal is submitted, not after.

The IP allocation agreement

The allocation agreement typically covers: who owns foreground IP (usually the small business for commercialization, with the university retaining rights for research and teaching), how background IP is handled (each party retains their own), how licensing revenue is split (often the small business commercializes and pays the university a royalty or license fee), and how publication is managed (university retains publication rights, often with a delay window for the small business).

Getting this right before submission saves months after award. Getting it wrong can tie up the Phase I in IP negotiations while the calendar runs.

PI affiliation

In SBIR, the PI must be primarily employed by the small business (more than half time). In STTR, the PI can be employed by either the small business or the research institution. This is the single most important practical difference for faculty-led startups: STTR allows a faculty member to remain the PI while continuing their academic position, which SBIR does not.

Bottom line

For a pure AI integration firm with no academic partnership, SBIR is the default. For a firm with a distinctive university partnership — shared IP, unique dataset, faculty founder — STTR is often the better structure. For most AI companies at the Precision Federal profile (senior industry PI, no academic affiliation), SBIR is the right choice.

Frequently asked questions