Where SBIR concentration actually lives
TL;DR. At the agency level, the SBIR program is not as concentrated as the "SBIR mill" critique suggests: DoD's top 10 Phase I winners share 11.2% of its Phase I awards over the last decade; NSF's top 10 share is under 1%. Concentration shows up one level down — inside specific topic families, where the same firms repeatedly win. NASA's largest mission directorates send 72.2% of awards to firms with 3+ prior wins in the same family. The reauthorization's new per-firm annual caps (50 Phase I / 20 Phase II) would have displaced only 55 awards across the entire 10-year window — all at DoD, all from four firms. The caps target a phenomenon the data only weakly shows.
Methodology
Data source: the official SBIR.gov bulk award extract, restricted to fiscal years 2014–2023 — the last full decade available, covering the post-2014 reauthorization regime through to the most recent complete year in the public file.
Firm names are normalized by stripping standard corporate suffixes (Inc, LLC, Corp, Co, Ltd) and whitespace before comparison, which collapses common near-duplicates ("Acme Inc." and "Acme Corporation") but does not deduplicate genuine name changes or subsidiaries.
Topic families are derived from the SBIR.gov Topic Code field, taking
the leading agency prefix (e.g. AF241-001 → AF,
NIH-NIAID-24-001 → NIH). Families with fewer
than 50 awards in the window are excluded from the concentration tables —
HHI is too noisy below that.
Sample window covers 61,685 awards from
the 207,731-row bulk file (older awards filtered out).
Replication code:
scripts/research/topicscout/compute_repeat_winner_concentration.py.
1. Agency-level concentration is modest
Aggregating across all topic families inside each agency, the top-10 firms capture between 0.7% (NSF) and 12.1% (Commerce) of Phase I awards. DoD — the headline SBIR program — sends about 11% of Phase I dollars to its top 10 firms, across a base of 5,500+ distinct winners.
| Agency | Phase I awards | Distinct firms | Top-10 firm share |
|---|---|---|---|
| DoD | 17,702 | 5,520 | 11.2% |
| HHS | 9,375 | 5,002 | 3.3% |
| DOE | 3,775 | 1,519 | 10.3% |
| NASA | 3,671 | 1,253 | 8.8% |
| NSF | 3,229 | 3,234 | 0.7% |
| USDA | 800 | 646 | 7.9% |
| DOC | 355 | 299 | 12.1% |
That is concentration, but it is not what the press shorthand "SBIR mills" implies — most agency programs have a long tail of one- and two-award firms doing the bulk of the work.
DoD — top 10 Phase I winners
| # | Firm | Phase I awards (2014–2023) | Share of agency Phase I |
|---|---|---|---|
| 1 | Physical Optics | 273 | 1.5% |
| 2 | Triton Systems | 272 | 1.5% |
| 3 | Intelligent Automation | 256 | 1.4% |
| 4 | Physical Sciences | 222 | 1.3% |
| 5 | Charles River Analytics | 204 | 1.2% |
| 6 | Lynntech | 172 | 1.0% |
| 7 | Creare | 151 | 0.9% |
| 8 | Luna Innovations | 147 | 0.8% |
| 9 | Toyon Research | 144 | 0.8% |
| 10 | CFD Research | 135 | 0.8% |
HHS — top 10 Phase I winners
| # | Firm | Phase I awards (2014–2023) | Share of agency Phase I |
|---|---|---|---|
| 1 | Physical Sciences | 46 | 0.5% |
| 2 | Lynntech | 35 | 0.4% |
| 3 | Minnesota Healthsolutions | 33 | 0.4% |
| 4 | Electronic Biosciences | 31 | 0.3% |
| 5 | Epicypher | 31 | 0.3% |
| 6 | Celdara Medical | 27 | 0.3% |
| 7 | Microbiotix | 27 | 0.3% |
| 8 | Giner | 25 | 0.3% |
| 9 | Larix Bioscience | 25 | 0.3% |
| 10 | Affinergy | 25 | 0.3% |
DOE — top 10 Phase I winners
| # | Firm | Phase I awards (2014–2023) | Share of agency Phase I |
|---|---|---|---|
| 1 | Radiation Monitoring Devices | 75 | 2.0% |
| 2 | Radiabeam Technologies | 47 | 1.2% |
| 3 | TDA Research | 44 | 1.2% |
| 4 | Physical Sciences | 40 | 1.1% |
| 5 | Radiasoft | 35 | 0.9% |
| 6 | Luna Innovations | 33 | 0.9% |
| 7 | Radiabeam Systems | 30 | 0.8% |
| 8 | Mainstream Engineering | 29 | 0.8% |
| 9 | Tech-X | 29 | 0.8% |
| 10 | Advanced Cooling Technologies | 27 | 0.7% |
NASA — top 10 Phase I winners
| # | Firm | Phase I awards (2014–2023) | Share of agency Phase I |
|---|---|---|---|
| 1 | CFD Research | 47 | 1.3% |
| 2 | Creare | 47 | 1.3% |
| 3 | Nanosonic | 36 | 1.0% |
| 4 | Cornerstone Research Group | 31 | 0.8% |
| 5 | Physical Sciences | 30 | 0.8% |
| 6 | Advanced Cooling Technologies | 27 | 0.7% |
| 7 | TDA Research | 27 | 0.7% |
| 8 | Alphacore | 26 | 0.7% |
| 9 | Traclabs | 26 | 0.7% |
| 10 | Freedom Photonics | 26 | 0.7% |
NSF — top 10 Phase I winners
| # | Firm | Phase I awards (2014–2023) | Share of agency Phase I |
|---|---|---|---|
| 1 | Farmsense | 3 | 0.1% |
| 2 | Secure Food Solutions | 3 | 0.1% |
| 3 | Kepley Biosystems | 3 | 0.1% |
| 4 | Smartbot360 | 3 | 0.1% |
| 5 | Alchemie Solutions | 2 | 0.1% |
| 6 | Astek Diagnostics | 2 | 0.1% |
| 7 | National Resource Consultants | 2 | 0.1% |
| 8 | Nitrate Elimination | 2 | 0.1% |
| 9 | Pensievision | 2 | 0.1% |
| 10 | Sentire Medical Systems | 2 | 0.1% |
USDA — top 10 Phase I winners
| # | Firm | Phase I awards (2014–2023) | Share of agency Phase I |
|---|---|---|---|
| 1 | Isca Technologies | 17 | 2.1% |
| 2 | Intelligent Optical Systems | 6 | 0.8% |
| 3 | TDA Research | 6 | 0.8% |
| 4 | Giner | 6 | 0.8% |
| 5 | Compact Membrane Systems | 6 | 0.8% |
| 6 | Ocean Era | 5 | 0.6% |
| 7 | Application Insight | 5 | 0.6% |
| 8 | Astalake Biosystems | 4 | 0.5% |
| 9 | Geo-Spider | 4 | 0.5% |
| 10 | Guild Associates | 4 | 0.5% |
DOC — top 10 Phase I winners
| # | Firm | Phase I awards (2014–2023) | Share of agency Phase I |
|---|---|---|---|
| 1 | Aerodyne Research | 8 | 2.3% |
| 2 | Creare | 7 | 2.0% |
| 3 | Toyon Research | 5 | 1.4% |
| 4 | CFD Research | 4 | 1.1% |
| 5 | Synthetik Applied Technologies | 4 | 1.1% |
| 6 | Coastaloceanvision | 3 | 0.8% |
| 7 | Arete Associates | 3 | 0.8% |
| 8 | Atmospheric & Space Technology Research Associates | 3 | 0.8% |
| 9 | Infobeyond Technology | 3 | 0.8% |
| 10 | Nikira Labs | 3 | 0.8% |
2. Topic-family concentration is where the action is
Look one level down — inside specific topic families — and the picture inverts. The same firms repeatedly win the same kinds of work. Across NASA's mission directorates, 60–72% of Phase I awards in each major family go to firms with 3 or more prior wins in that same family.
The headline measure is the Herfindahl-Hirschman Index (HHI), the standard antitrust concentration metric: the sum of squared firm shares, scaled to 10,000. The DOJ treats HHI above 1,500 as "moderately concentrated" and above 2,500 as "highly concentrated." SBIR topic families don't reach those antitrust thresholds, but the repeat-winner share — what fraction of each family's awards go to firms with 3+ prior wins — runs as high as the 70% range in NASA's main program areas.
NASA — most concentrated topic families
| Family | Awards | Distinct firms | HHI | Top firm | Share to repeat winners (≥3 awards) |
|---|---|---|---|---|---|
A | 841 | 257 | 109 | Mosaic ATM (3.8%) | 72.1% |
T | 724 | 296 | 68 | CFD Research (3.0%) | 62.6% |
Z | 871 | 343 | 59 | CFD Research (2.2%) | 62.6% |
H | 1,165 | 443 | 46 | TDA Research (2.0%) | 62.8% |
S | 1,609 | 512 | 45 | Creare (2.2%) | 72.2% |
DoD — most concentrated topic families
| Family | Awards | Distinct firms | HHI | Top firm | Share to repeat winners (≥3 awards) |
|---|---|---|---|---|---|
NON | 59 | 49 | 290 | Physical Sciences (10.2%) | 15.3% |
NGA | 156 | 69 | 277 | Intelligent Automation (9.0%) | 53.2% |
SCO | 60 | 48 | 250 | Soar Technology (6.7%) | 6.7% |
DMEA | 100 | 59 | 232 | Physical Optics (7.0%) | 32.0% |
SF | 100 | 75 | 218 | Physical Sciences (9.0%) | 22.0% |
DHA | 515 | 171 | 217 | Triton Systems (10.5%) | 65.8% |
ST | 105 | 63 | 193 | Stottler Henke Associates (3.8%) | 24.8% |
CBD | 298 | 121 | 142 | CFD Research (3.7%) | 53.4% |
DTRA | 308 | 142 | 137 | Radiation Monitoring Devices (6.5%) | 50.3% |
DHP | 315 | 137 | 134 | Triton Systems (3.8%) | 53.7% |
HHS — most concentrated topic families
| Family | Awards | Distinct firms | HHI | Top firm | Share to repeat winners (≥3 awards) |
|---|---|---|---|---|---|
NIMHD | 61 | 40 | 459 | ISA Associates (14.8%) | 36.1% |
ODCDC | 55 | 37 | 373 | Luna Innovations (7.3%) | 32.7% |
FDA | 52 | 36 | 369 | En Solucion (7.7%) | 26.9% |
NIDILRR | 57 | 37 | 341 | Createability Concepts (7.0%) | 28.1% |
N/A | 57 | 42 | 323 | Touch Graphics (8.8%) | 24.6% |
113 | 51 | 41 | 296 | Litron Laboratories (5.9%) | 17.6% |
500 | 94 | 59 | 242 | Teachley (6.4%) | 23.4% |
172 | 181 | 92 | 228 | Electronic Biosciences (9.4%) | 47.5% |
600 | 73 | 58 | 200 | Transcendent International (4.1%) | 8.2% |
NIDCD | 157 | 88 | 174 | Altec (4.5%) | 38.9% |
DOE — most concentrated topic families
| Family | Awards | Distinct firms | HHI | Top firm | Share to repeat winners (≥3 awards) |
|---|---|---|---|---|---|
DE | 55 | 24 | 466 | N5 Sensors (5.5%) | 65.5% |
1 | 57 | 24 | 440 | Global Bamboo Technologies (5.3%) | 52.6% |
N/A | 54 | 25 | 418 | Advanced Cooling Technologies (5.6%) | 27.8% |
06A | 54 | 33 | 418 | Radiation Monitoring Devices (9.3%) | 38.9% |
01A | 52 | 34 | 414 | Reservoir Labs (11.5%) | 23.1% |
02B | 50 | 31 | 408 | Reservoir Labs (10.0%) | 16.0% |
14A | 52 | 31 | 406 | Radiation Monitoring Devices (7.7%) | 26.9% |
07A | 50 | 35 | 392 | Radiabeam Technologies (10.0%) | 28.0% |
18A | 50 | 35 | 376 | Advanced Cooling Technologies (10.0%) | 16.0% |
01B | 52 | 35 | 355 | Princeton Optronics (7.7%) | 19.2% |
3. The April 13, 2026 reauth caps barely move the needle
The SBIR/STTR Reauthorization Act signed April 13, 2026 introduced per-firm annual caps: 50 Phase I awards and 20 Phase II awards per firm per year. Applied retroactively to the 2014–2023 window:
Applied to 2014–2023 SBIR/STTR Phase I and Phase II awards: 55 awards (0.1%) would have been displaced if the new per-firm annual caps (50 Phase I / 20 Phase II) had been in force.
Across the full 61,685-award window, only 4 firms ever exceeded the new cap in any single year.
| Agency | Awards displaced (cumulative, 10 yrs) |
|---|---|
| DoD | 55 |
Firms that would have been clipped
| Firm | Excess awards (10 yrs) |
|---|---|
| Physical Optics | 38 |
| Physical Sciences | 14 |
| Intelligent Automation | 2 |
| Creare | 1 |
Two readings of this:
- The cap level is set at a threshold almost no firm has ever approached. Even the top-10 DoD firms average ~25 Phase I awards per year at their peak — half the new cap.
- If the policy intent is to redistribute work away from
"repeat winners," the binding constraint isn't firm-level annual
throughput; it's topic-family concentration. A firm that
wins three awards in NASA topic
Hin one year hasn't hit the new cap, but it has captured a meaningful share of that family.
A cap calibrated to family-level concentration — e.g. capping any one firm at N% of a topic family's annual awards — would have more bite than the current per-firm annual ceiling.
About this analysis
TopicScout is a tool that pulls every open SBIR/STTR topic each week and ranks them against a firm's capability statement, so founders skip the manual crawl. We sit on the same SBIR.gov bulk extract used for this analysis, plus the live agency open-topic feeds.
Questions, requests for additional cuts, or interested in the underlying CSV? Email jcavallo@alumni.colgate.edu.