- Security & Defense
- US Defense
- Determining America's Role in the World
Can America's venture-backed defense sector scale? Tracking Department of War contract obligations to roughly eight hundred startups between 2015 and 2026, the authors go beyond three often studied industry leaders—SpaceX, Anduril, and Palantir—to ask: Can the hundreds of smaller firms cross from research dollars into procurement programs? The answer determines whether the arsenal of the future arrives at speed and scale.
Key Takeaways
- Federal funding has arrived for defense R&D: Even excepting the three largest companies (SpaceX, Anduril, and Palantir), Department of War R&D commitments to venture-backed aerospace and defense startups jumped from $33 million in FY2015 to $1.8 billion in FY26.
- Procurement is the real test, and recent signals are positive. By the end of FY26, $859 million in procurement was allocated to this cohort—more than the department committed across the entire prior decade through FY2024.
- A few giants have dwarfed the field since FY2015: SpaceX ($10.2B DOW funding in FY26), Anduril ($2.8B), and Palantir ($4.2B). But the question that matters is whether the hundreds of smaller firms behind them can cross from R&D into programs of record.
- The crossing depends on structural reform. Keeping the bridge standing means accelerating use of agile contracting authorities, transitioning to portfolio versus product-based acquisition, and continuing to clear security, deployment, and accounting burdens that too often have stranded companies in the valley of death.
Cite this essay:
Dan Berkenstock, Maggie Gray, and Walter Manuel, “From R&D to the Arsenal: The Status of Acquisition Progress for America's Emerging Venture-Backed Defense Industrial Base,” Hoover Institution, Technology Policy Accelerator, November 2026.
DATA COMPANION
This interactive data companion tracks Department of War (DOW) contract obligations to 882 venture-backed aerospace and defense companies beginning in FY2007. Contract data are current through late May 2026, because of a ninety-day holdback from award to public release. Company data, including founding years, come from PitchBook as of August 2026. All 2026 figures are partial.
Part One
Who the companies are
Companies founded per year
Founding years of the venture-backed aerospace and defense companies.
Source PitchBook Data Inc. Accessed August 2026. All 882 companies are included, SpaceX and Anduril among them. Founding years before 2000 are grouped, and the 2026 count is partial.
Top ten headquarters states
Where the companies are based.
Source PitchBook Data Inc. Accessed August 2026. Shares are of the companies with a listed headquarters.
Part Two
How much the department obligates to these companies
Total DOW obligations to the three top-of-market firms and the rest of the cohort
SpaceX and Anduril belong to the broader cohort. Palantir, an established firm that went public in 2020, sits outside it and is shown for scale.
Source Hoover Institution analysis of Federal Procurement Data System actions above $100,000 and Department of War (DOW) Other Transaction agreements, FY2015–FY2026. All categories are counted, launch and services included. FY2025 constant dollars. FY2026 is a partial year, with contract data current through late May 2026.
R&D and procurement obligations by fiscal year
The cohort's annual take, split by category, in constant FY2025 dollars.
Source Same contract data as above. Constant FY2025 dollars via CPI-U. Launch excluded. FY2026 partial.
R&D versus procurement, as a share of each year
The two-way split each year, services excluded so R&D and procurement sum to 100%.
Source Same contract data as above. Shares of each year's total are identical in either dollar basis.
The cohort's share of all DOW obligations
How much of DOW’s total research and procurement spend these companies capture.
Source Same contract data as above, with all-DOW totals from USAspending. The cohort lines follow the company set selected above.
Top 25 companies by cumulative obligations
The 25 largest by R&D plus procurement, as in the paper, ordered by total obligations with services included. Each bar is split into R&D, procurement and services.
Source Same contract data as above. Cumulative over all years, FY2007–FY2026, in FY2025 dollars, launch excluded. Services are shown here but sit outside the R&D/procurement split used in most charts on this page.
Part Three
What the money buys, and the channels through which it flows
FAR contracts versus Other Transactions
Which contracting mechanisms move the money.
*FY2026 partial year
Source Same contract data as above. FAR R&D means research product and service codes (PSC A-codes), OT R&D means prototype agreements, and OT procurement means production agreements. FPDS has reported Other Transactions systematically only since about FY2019, so earlier OT columns understate them.
Procurement dollars per R&D dollar
The cohort against all of DOW, on a logarithmic scale.
Source Same contract data as above, with all-DOW totals from USAspending. The scale is logarithmic, so each gridline is a tenfold change. All-DOW figures cover the whole department, not just this cohort.
Top product and service codes
The five largest PSCs on each side, with everything else grouped.
R&D
Procurement
Source Same contract data as above. Top five ranked by cumulative FY2025 dollars. OT production agreements count as procurement regardless of the PSC reported.
Where the dollars go, by technology domain
The 865 companies fall into eighteen primary domains. The twelve largest are shown, and the rest are grouped.
All 865 companies
| Technology domain | Obligations | Share |
|---|---|---|
| Total | 100.0% |
Source Same contract data as above. Each company is assigned to its primary technology domain, so all of its obligations sit in a single slice. The twelve named domains are the paper’s top twelve. The grouped domains are Human Performance & Training, Computing & Semiconductors, Quantum, Bio & Health, Logistics & Supply Chain, Other, and companies without an assigned domain. As in the paper, procurement shares leave out the $2.6M of procurement not assigned to a domain. Over FY2007–FY2026 SpaceX and Anduril account for $13.45B of the $25.14B total, SpaceX $10.65B (Space, of which $9.59B is launch and services) and Anduril $2.80B (Robotics & Autonomy), so only those two slices change between company sets. Counted from FY2015, as the paper does, SpaceX’s total is $10.2B in FY2025 dollars.
Hardware, software, and the companies that do both
The same dollars, categorized by what each company builds. The cohort includes 492 hardware companies, 192 software companies, and 181 that build both.
All 865 companies
| Company profile | Obligations | Share |
|---|---|---|
| Total | 100.0% |
Source Same contract data as above. Each company carries one profile, so its full obligations sit in a single slice. Both excluded companies build hardware, SpaceX on its own and Anduril alongside software, so the software slice is identical under both company sets.
Part Four
Which organizations do the buying
Obligations by awarding agency
Who is writing the checks, by category and year.
R&D
Procurement
Source Same contract data as above. Air Force includes the Space Development Agency.
The three services, side by side
R&D against procurement within each service, on a shared scale.
Air Force
Army
Navy
Source Same contract data as above. Shared vertical scale across the three panels.
Service drill-down
The three-services chart above compares them at a glance. This chart shows the detail behind each, including the companies they buy from. The Defense Innovation Unit (DIU) and US Special Operations Command (SOCOM) are joint organizations rather than services, but they award enough to include here.
Obligations by fiscal year
Top 12 companies by cumulative obligations from this service
Source Same contract data as above. Every awarding organization, the services plus DIU and SOCOM, is tabulated from the same FPDS actions and DOW Other Transaction agreements on the identical basis. Air Force includes the Space Development Agency. Each organization is drawn on its own vertical scale so the smaller ones stay readable. Use the three-services chart above for a like-for-like comparison of the three services. FY2025 dollars, launch excluded, company set as selected above. Bars run from FY2007, and cumulative totals and shares cover FY2015–FY2026, as in the paper’s table 6.
Part Five
How fast new companies reach the money
Average cumulative obligations per company, by number of years since founding
Each line follows a fixed set of companies from its founding year forward, so the curves are strictly cumulative.
Source Hoover Institution analysis of Federal Procurement Data System actions above $100,000 and Department of War (DOW) Other Transaction agreements, FY2007–FY2026. FY2025 constant dollars. FY2026 is a partial year. Cohorts are balanced panels, meaning a company enters only where its full observation window falls inside the data period.
Share of each cohort clearing $1 million cumulative
Both panels share one scale, which rescales to the selected agency. Research funding and procurement funding are not close to comparable at these ages.
Nobody in any cohort clears $1 million of procurement inside two years — research funding is the only door in. What has changed is the lag behind it: the 2021–2026 cohort crosses the procurement threshold by year four at 2.8% against 0.9% for 2015–2020, a roughly threefold compression in the time it takes to convert research dollars into production work.
Source Same contract data as above. Denominators shrink at longer horizons, where the observation window has not yet elapsed. Counts are given in the table below.
Cohorts reaching $1 million cumulative, with counts
The averages in the cumulative-obligations chart above are lifted by a small number of outliers. These rates ask the same question of every company in the cohort.
| Cohort | Within 2 yrs | Within 4 yrs | Within 6 yrs | Within 8 yrs |
|---|---|---|---|---|
| Research & development obligations | ||||
| Pre-2015 | 0.7% 1/146 | 4.8% 7/146 | 10.3% 15/146 | 20.5% 30/146 |
| 2015–2020 | 3.1% 10/321 | 24.3% 78/321 | 39.6% 127/321 | 40.9% 110/269 |
| 2021–2026 | 7.9% 25/317 | 30.9% 55/178 | 42.1% 24/57 | — |
| Procurement obligations — goods only | ||||
| Pre-2015 | 0.0% 0/146 | 0.0% 0/146 | 0.7% 1/146 | 1.4% 2/146 |
| 2015–2020 | 0.0% 0/321 | 0.9% 3/321 | 3.4% 11/321 | 4.8% 13/269 |
| 2021–2026 | 0.0% 0/317 | 2.8% 5/178 | 5.3% 3/57 | — |
Source Same contract data as above. Rates are shares of all companies in the cohort with an observable window, including those that never received an award.
Years from founding to the first $1M
Median years to cross the threshold, among companies that crossed it.
Source Same contract data as above. Medians are among the companies that reached the threshold, and the finding above reports how many reached it for the selected dollar basis and company set. Many companies in the newer cohorts have not yet had time to cross, so their medians are floors. This chart groups companies by this page’s observation windows, so its counts differ slightly from the paper’s. The paper reports R&D medians of seven, four, and two years (58 of 145, 149 of 314, and 71 of 326 crossed) and procurement medians of ten, six, and three years.
Definitions and caveats
- Universe
- 882 venture-backed aerospace and defense companies, of which 856 have an established founding year (from the source list, PitchBook, or web research). The other twenty-six are left out of the founding-cohort figures. The technology-domain and hardware/software charts cover the 865 companies classified by technology domain.
- Research & development
- Product and Service Codes beginning
A, plus Other Transaction agreements typedPROTOTYPE. $9.77 billion for all 882 companies, FY2007–FY2026. - Procurement
- Numeric Product and Service Codes for goods, plus Other Transaction agreements typed
PRODUCTION. $3.49 billion for all 882 companies, FY2007–FY2026. Launch and services are excluded from the split, and software as a service has no separate category. - Excluded from both
- $11.87 billion falls in neither category, so R&D and procurement do not sum to the total. That remainder covers launch services under
V126, software as a service and IT (D-series and7030), and other services such as engineering support, maintenance, and construction. - Observation window
- Award data begins in FY2007, so a company enters a calculation only where its full window falls inside FY2007–FY2026. This is why the pre-2015 cohort is limited to companies founded 2007–2014 and why the newest cohort's line ends at year four.
- Concentration
- Median cumulative obligations are zero through nearly every year in every cohort. The averages therefore describe cohort dollars per company rather than a typical company. The five largest recipients hold 82%, 53%, and 55% of their respective cohorts' dollars.
Hoover Institution · Technology Policy Accelerator · November 2026