NIH SBIR is not one program
The National Institutes of Health runs the largest dollar-volume civilian SBIR program in the federal government. It is also the most organizationally complex. NIH is actually 27 institutes and centers (ICs), each with its own director, budget, study sections, program officers, and funding philosophy. NIH SBIR is run centrally by the Office of Extramural Research, but topics, review panels, and funding decisions live inside each IC. For a small firm, the question is not "should I submit to NIH" but "which IC matches my work and how do I engage it."
Each NIH institute funds SBIR independently. Largest funders: NCI, NHLBI, NIMH, NIAID. Phase I ceiling: $300K. Phase II ceiling: $2M. Grant sizes are materially larger than DoD SBIR.
NIH SBIR uses an omnibus grant solicitation with broad institute-specific priorities and periodic specific Program Announcements and Funding Opportunity Announcements (FOAs). The mechanism resembles R&D grantmaking more than DoD contracting: peer review by named study section reviewers, score-based funding decisions, and substantial negotiation around scope.
The big-dollar institutes
National Cancer Institute (NCI)
Largest IC, largest SBIR portfolio. NCI funds diagnostics, therapeutics, imaging, informatics, and digital health for cancer. AI-relevant topic areas: radiology and pathology ML, clinical trial matching, genomics informatics, cancer surveillance data systems.
National Heart, Lung, and Blood Institute (NHLBI)
Cardiovascular, pulmonary, hematologic, sleep. AI topics around ECG/echo analysis, sleep study automation, lung imaging, and remote monitoring ML.
National Institute on Aging (NIA)
Aging biology, Alzheimer's, dementia. AI topics around cognitive assessment, digital biomarkers, in-home monitoring, dementia prediction, and caregiver decision support.
National Institute of Mental Health (NIMH)
Mental health. AI topics around digital phenotyping, suicide risk prediction, therapy engagement tools, and clinical decision support.
National Institute of General Medical Sciences (NIGMS)
Basic biomedical research. AI topics around biomedical data science infrastructure, omics analysis, and computational biology tools.
National Institute of Allergy and Infectious Diseases (NIAID)
Infectious disease, immunology. AI topics around pathogen surveillance, vaccine informatics, antimicrobial resistance prediction.
National Institute of Neurological Disorders and Stroke (NINDS)
Neurology. AI topics around imaging, seizure prediction, neuromodulation, and clinical trial platforms.
National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK)
Diabetes, kidney, liver, GI. AI topics around closed-loop glucose systems, CKD prediction, and MASH imaging.
National Eye Institute (NEI)
Ophthalmology. Strong imaging AI tradition — retinal disease detection, glaucoma screening, and OCT analysis all have active SBIR volume.
National Library of Medicine (NLM)
Biomedical informatics and NLP. Smaller budget but directly AI-focused. Topics around clinical NLP, literature mining, and biomedical knowledge graphs.
Other substantive ICs
NICHD (child health), NIDA and NIAAA (substance use), NIDCD (deafness and communication), NIAMS (arthritis, musculoskeletal), NCATS (translational), AHRQ, CDC (separate but administratively linked SBIR), FDA (small program).
Review culture: study sections and impact scores
NIH review is fundamentally different from DoD review. Applications go to named study sections composed of outside subject-matter experts. Reviewers score on Significance, Investigator(s), Innovation, Approach, and Environment — five criteria with an overall Impact score that drives the funding decision. Summary statements are written and shared with the applicant after review. This transparency is unusual in federal contracting and is a gift: you learn exactly what the reviewers thought.
The practical consequence: NIH is the SBIR agency most friendly to iterative improvement. A firm that submits, reads the summary statement, and resubmits in the next cycle often wins on the second or third attempt.
Program officers are the single most important contact
At NIH, the program officer (PO) assigned to your IC is the relationship to cultivate. POs advise on fit, help you understand which IC and which FOA fits best, and can be consulted on concept before the full application is written. Every IC publishes PO contact information organized by SBIR program. A 20-minute phone call with the right PO saves weeks of wrong-direction work.
Award amounts and mechanics
NIH SBIR Phase I in 2026 is typically up to 306 thousand dollars over up to one year (higher than DoD). Phase II is up to 2.05 million over up to two years. NIH accepts higher Phase I and Phase II ceilings than the default for topics the Director has flagged as strategically important. The Fast Track mechanism allows Phase I and Phase II to be submitted together for expedited review — useful for firms with clear technical risk and strong preliminary data.
Direct-to-Phase-II (DP2) is available with documented prior technology maturation. NIH also runs commercialization readiness pilot programs and post-Phase-II extensions.
Where AI/ML firms should target in 2026
- NCI — largest, most AI-friendly, deepest informatics pipeline. Imaging, trials, and surveillance.
- NIA — strong digital biomarker and in-home monitoring interest tied to Alzheimer's and dementia initiatives.
- NHLBI — cardiology AI, sleep AI, pulmonary imaging.
- NIMH — digital phenotyping and clinical AI.
- NEI — imaging AI with a clear regulatory pathway tradition.
- NLM — pure biomedical informatics and NLP.
Commercialization and Phase III
NIH emphasizes commercialization throughout. The Phase II application requires a serious commercialization plan. Phase III for biomedical AI often means FDA clearance (510(k) or De Novo), licensing to a larger medical device or pharma company, or direct SaaS sales to health systems. The NIH SBIR program does not typically fund Phase III directly — Phase III is commercial revenue or follow-on private/payer capital.
For firms without prior regulatory experience, NIH provides resources (NIH SEED, regulatory consultations). A Phase II commercialization section that shows the firm understands FDA pathway, reimbursement, and clinical adoption is a strong differentiator.
Practical steps for a first NIH SBIR
- Identify two ICs whose priorities match your capability. Do not try to submit to five.
- Read each IC's current funding priorities and program announcements.
- Contact a program officer in one IC. Introduce the firm, describe the capability, ask about fit.
- If the PO suggests a specific FOA, write to that FOA. If they suggest an omnibus submission, that is also fine.
- Write with study section reviewers in mind: Significance first, then Approach, with preliminary data wherever available.
- Plan the commercialization/regulatory path from day one. Phase II without a FDA strategy (for device/diagnostic work) is a weak application.
Bottom line
NIH SBIR in 2026 is the largest civilian SBIR opportunity and the most transparent review process of any agency. For a small AI firm with biomedical applicability, NCI, NIA, NHLBI, NIMH, NEI, and NLM are the natural entry points. Program officer conversations are the single highest-leverage pre-submission activity. Study section reviews are iterative — plan on two cycles to land a first award, and the firm that takes the written critique seriously will score higher on every subsequent submission.
Frequently asked questions
NIH SBIR runs across 27 institutes and centers, each with its own director, budget, study sections, and priorities. CDC, FDA, and AHRQ also participate in linked programs.
Phase I is up to 306 thousand dollars over up to one year. Phase II is up to 2.05 million over up to two years. Higher ceilings apply to topics flagged as strategically important.
NCI is the largest. NIA, NHLBI, NIMH, NEI, and NLM all have active AI SBIR portfolios. NLM focuses purely on biomedical informatics and NLP.
NIH uses named study sections, five scored criteria, and written summary statements shared with applicants. The transparency makes iterative resubmission practical — most winners resubmit at least once.
Very. Program officers advise on IC fit, FOA selection, and scope. A 20-minute pre-submission call saves weeks of wrong-direction work.
NIH does not typically fund Phase III. For medical AI, Phase III means FDA clearance, licensing, or direct commercial revenue. The Phase II commercialization plan should describe that path.