Military AI Statistics 2026
By Axis Intelligence Research
Co-author: Sarah Mitchell, AI & Machine Learning | Last updated: August 27, 2026 | License: CC BY 4.0
The FY 2027 U.S. Department of War budget requests $58.5 billion for artificial intelligence and CJADC2 and $54.0 billion for autonomous and remotely operated systems, out of $1.45 trillion in total budgetary resources. Axis Intelligence Research puts the combined AI-and-autonomy request at $112.5 billion — 3.90% of all global military spending recorded in 2025.
Quick Answer
Axis Intelligence Research finds that the FY 2027 request devotes $7.76 of every $100 of Department of War budgetary resources to artificial intelligence and autonomy — the Defense AI Investment Intensity (DAII™) combined reading. The autonomy line alone rose from $13.4 billion in the FY 2026 request to $54.0 billion in FY 2027, a 303.0% increase. Of the $58.5 billion AI line, 78.63% sits in one item: the sovereign AI Arsenal.
Key Findings
- Axis Intelligence Research calculates a DAII™ combined reading of 7.76 for FY 2027 — dollars of disclosed AI and autonomy budget authority per $100 of total Department of War budgetary resources, up from a like-for-like autonomy-only reading of 1.39 in FY 2026.
- The FY 2027 Department of War budget requests $58.5 billion for artificial intelligence and Combined Joint All-Domain Command and Control, of which $46.0 billion is a multi-year mandatory investment in a sovereign AI Arsenal.
- Axis Intelligence Research finds that $1.2 billion — 2.05% — of the published $58.5 billion AI line is not itemised in the Department’s own sub-bullets, which account for $57.3 billion.
- At $112.5 billion, the combined FY 2027 AI-and-autonomy request would rank fifth in the world as a national military budget, behind Germany’s $114 billion and ahead of India’s $92.1 billion in SIPRI’s 2025 country table.
- The Government Accountability Office reported in April 2026 that DOD maintains a classified list of more than 200 AI-related activities and, unlike most executive branch agencies, is not required to publish an AI use case inventory.
| Metric | 2026/27 figure |
|---|---|
| FY2027 AI + CJADC2 request | $58.5B |
| Autonomous & remotely operated systems | $54.0B |
| Sovereign AI Arsenal | $46.0B |
| AI integration into platforms/workflows | $6.8B |
| Maven + Joint Fires Network | $2.3B |
| Leading-edge AI R&D | $2.2B |
| FY2027 DAII™ AI-only | 4.03 |
| FY2027 DAII™ autonomy-only | 3.72 |
| Autonomy request growth vs FY2026 | +303% |
| Known classified DOD AI activities | 200+ |
What Counts as Military AI in This Dataset?
| Category | Counted as AI? | Treatment |
|---|---|---|
| Artificial intelligence + CJADC2 | Yes | Official disclosed AI line |
| Autonomous & remotely operated systems | Separate | Adjacent; potential overlap with AI |
| Counter-UAS | Separate | Not automatically counted as AI |
| Cyber activities | Separate | Included only where explicitly AI-related |
| Classified AI programs | Excluded | No auditable public figures |
| Commercial market-size forecasts | External estimates | Never mixed with government budget authority |
Axis Intelligence does not treat “AI,” “autonomy,” “uncrewed systems,” and “CJADC2” as interchangeable categories. Where the Department does not disclose overlap, the dataset reports the lines separately rather than estimating an adjustment.
How Much Is the U.S. Military Spending on AI in 2026?
The honest answer starts with a distinction most coverage skips: there is requested money and there is enacted money, and for AI they are not the same number.
The FY 2027 Defense Budget Overview, published by the Office of the Under Secretary of War (Comptroller) on April 20, 2026, requests $1.45 trillion in total budgetary resources — $1.1 trillion in discretionary budget authority plus $350 billion in mandatory resources the Administration assumes will be enacted in a subsequent reconciliation bill. Against the FY 2026 enacted level of $1,009.1 billion, that is an increase of $440.9 billion, or 43.7%.
Inside that request, two lines carry the AI story.
FY 2027 AI and autonomy lines, as published
| Line item | FY 2027 request | Share of AI line | Source |
|---|---|---|---|
| Artificial intelligence and CJADC2 (total) | $58.5B | 100% | DoW FY 2027 Overview Book |
| — Sovereign AI Arsenal (multi-year mandatory) | $46.0B | 78.63% | DoW FY 2027 Overview Book |
| — AI integration into platforms, systems, workflows | $6.8B | 11.62% | DoW FY 2027 Overview Book |
| — Maven Smart System and Joint Fires Network | $2.3B | 3.93% | DoW FY 2027 Overview Book |
| — Leading-edge AI research and development | $2.2B | 3.76% | DoW FY 2027 Overview Book |
| — Not itemised (Axis Intelligence Research) | $1.2B | 2.05% | Axis calculation |
| Autonomous and remotely operated systems | $54.0B | — | DoW FY 2027 Overview Book |
| — of which Drone Dominance mandatory | $39.2B | — | DoW FY 2027 Overview Book |
| Counter-Unmanned Systems (C-UXS) | $20.6B | — | DoW FY 2027 Overview Book |
| Cyberspace activities | $20.5B | — | DoW FY 2027 Overview Book |
Chart 1 — “FY 2027 AI and autonomy request, by disclosed line” — stacked bar, source: DoW FY 2027 Overview Book. All values in CSV.
The $1.2 billion nobody has counted. Add the four sub-bullets the Department publishes under its AI heading — $46.0B + $6.8B + $2.3B + $2.2B — and you get $57.3 billion, not $58.5 billion. Axis Intelligence Research finds a residual of $1.2 billion, 2.05% of the line, that the Overview Book does not attribute to any named program. That gap is small in percentage terms and large in absolute terms: it exceeds the entire FY 2027 request for the Defense Innovation Unit ($956 million). It is the kind of figure that only appears when someone adds the column.
Sarah Mitchell: Every AI budget cycle produces the same reporting artefact. A headline number gets lifted, repeated, and eventually detached from the document that generated it — and the version that circulates is usually rounder and vaguer than the source. The FY 2026 cycle is the clean example. The Department’s own Overview Book calls the $13.4 billion line “autonomous and remotely operated systems.” Most secondary coverage called it “AI and autonomy,” and a fair amount of it called it an AI budget outright. It was not. It was drones, uncrewed vessels, and ground autonomy. If you want a defensible AI series, you have to read the line label, not the headline.
What Is the Defense AI Investment Intensity (DAII™)?
DAII™ — Defense AI Investment Intensity — is an Axis Intelligence Research metric that expresses disclosed artificial intelligence and autonomy budget authority as dollars per $100 of total departmental budgetary resources. It answers a question a raw dollar figure cannot: as the defense budget itself grows at historic rates, is AI growing faster than the thing it sits inside?
Formula
DAII = (disclosed AI and autonomy budget authority / total budgetary resources) x 100
No weighting is applied. The metric is a single ratio built from two figures published in the same document, which is what makes it reproducible by anyone holding the Overview Book. Three readings are published so that comparisons are never forced across incompatible definitions.
DAII™ readings
| Reading | Fiscal year | Numerator | Denominator | DAII™ |
|---|---|---|---|---|
| Combined (AI + autonomy) | FY 2027 request | $112.5B | $1,450.0B | 7.76 |
| AI and CJADC2 only | FY 2027 request | $58.5B | $1,450.0B | 4.03 |
| Autonomy only | FY 2027 request | $54.0B | $1,450.0B | 3.72 |
| Autonomy only | FY 2026 request | $13.4B | $961.6B | 1.39 |
Chart 2 — “DAII™ readings, FY 2026 request vs FY 2027 request” — grouped bar. Source: DoW FY 2027 and FY 2026 Overview Books, Axis Intelligence Research calculation.
The like-for-like reading. Only one of these comparisons is methodologically clean. “Autonomous and remotely operated systems” appears under the identical label in both Overview Books, so the autonomy-only series can be compared directly: DAII™ moved from 1.39 to 3.72, a 2.67x increase in intensity. Dollars grew 303.0%; the budget they sit inside grew 43.7%. Autonomy is outrunning its own container by roughly seven to one.
The AI-and-CJADC2 line has no FY 2026 counterpart in the Overview Book. That is not an oversight in our series — it is the finding. The FY 2027 request is the first Department of War Overview Book to publish a consolidated artificial intelligence investment total. Before FY 2027, there was no top-line AI figure to compare against, which is precisely why the widely circulated “$13.4 billion for AI” framing had no primary-source anchor.
The combined reading is an upper bound. The Department presents the $58.5 billion AI line and the $54.0 billion autonomy line as separate bullets and publishes no reconciliation of overlap between them. Autonomous systems contain AI; the AI line funds autonomy enablers. Axis Intelligence Research therefore publishes the combined DAII™ reading of 7.76 as a ceiling, and the two component readings as the figures to cite when precision matters. We decline to invent an overlap adjustment, because any such adjustment would be a guess wearing a decimal point.
Where Is the Money Actually Going: Infrastructure or Applications?
This is the segmentation nobody publishes, and it changes how the number should be read.
Of the $58.5 billion AI line, $46.0 billion goes to the AI Arsenal — described in the Overview Book as enterprise-scale AI infrastructure, a government-owned AI supercomputing capability inside secure hardened data centres, combined with flexible commercial surge capacity. Only $6.8 billion funds integrating AI into platforms, systems and workflows.
Axis Intelligence Research calculates an infrastructure-to-application ratio of 6.76:1.
| Category | FY 2027 | Share | Source |
|---|---|---|---|
| Infrastructure (AI Arsenal) | $46.0B | 78.63% | DoW FY 2027 Overview Book |
| Application and integration | $6.8B | 11.62% | DoW FY 2027 Overview Book |
| Decision systems (Maven Smart System, JFN) | $2.3B | 3.93% | DoW FY 2027 Overview Book |
| Research and development | $2.2B | 3.76% | DoW FY 2027 Overview Book |
| Unitemised | $1.2B | 2.05% | Axis calculation |
Chart 3 — “Infrastructure versus application: FY 2027 AI request composition” — donut. Source: DoW FY 2027 Overview Book; Axis Intelligence Research.
For context on what that ratio implies about power, land and grid interconnection, see our AI data center statistics research, where the same infrastructure-first pattern shows up in the commercial hyperscaler build-out.
Sarah Mitchell: Nearly four fifths of the AI request buys compute and the buildings around it. That is a bet on inference capacity being the binding constraint, and it is a defensible bet — you cannot run frontier models at classification levels on rented capacity you do not control. But the ratio is also a forecast about where the bottleneck is not. Every enterprise AI programme that stalled in the last three years stalled at integration: data that would not join, workflows nobody redesigned, evaluation nobody funded. The request spends $6.76 on the datacentre for every dollar it spends on making the thing usable inside an existing workflow. Fielding, not procurement, is where this gets tested.
How Fast Is Military Autonomy Spending Growing?
Faster than any other line in the document.
Autonomy and counter-autonomy, requested and enacted
| Line | FY 2024 | FY 2025 | FY 2026 | FY 2027 request | Source |
|---|---|---|---|---|---|
| Autonomous and remotely operated systems | — | — | $13.4B (request) | $54.0B | DoD FY 2026 / DoW FY 2027 Overview Books |
| Counter-Unmanned Systems (C-UXS) | $2.681B (actual) | $2.247B (enacted) | $3.187B (request) | $20.6B | DoD FY 2026 / DoW FY 2027 Overview Books |
| Total budgetary resources | — | $861.9B (actual) | $1,009.1B (enacted) | $1,450.0B | DoW FY 2027 Overview Book |
Chart 4 — “Autonomy and counter-autonomy budget lines, FY 2024–FY 2027” — line chart, log scale. Source: DoD FY 2026 and DoW FY 2027 Overview Books.
Counter-UAS is the sharper curve. The FY 2027 Overview Book states a 424% increase over an FY 2026 enacted level of $3.9 billion, reaching $20.6 billion — of which $14.4 billion arrives through the Drone Dominance mandatory request. The Department has stood up Joint Interagency Task Force 401 to consolidate counter-small-UAS authorities.
A discrepancy worth publishing. The FY 2026 Overview Book records a C-UXS request of $3.187 billion. The FY 2027 Overview Book cites an FY 2026 enacted level of $3.9 billion. The $713 million difference is the gap between what was asked for and what Congress provided, plus mandatory funding under Public Law 119-21. Both figures are correct; they measure different things. Anyone building a C-UXS time series who mixes them will produce a growth rate that is wrong by about 22%.
Drone Dominance totals $53.6 billion across both halves — $39.2 billion for autonomous systems procurement and domestic production capacity, $14.4 billion for counter-unmanned systems. Note the split inside the autonomy line: 72.59% is mandatory, leaving $14.8 billion in discretionary budget authority. If reconciliation does not pass, most of the autonomy request does not exist.
What Is the Pentagon Actually Deploying?
Budget authority is an intention. Fielded systems are a fact. The two are separated by years and, in this portfolio, by an unusual amount of classification.
Documented AI programs and platforms
| Program | Function | Status / figure | Source |
|---|---|---|---|
| GenAI.mil | Enterprise generative AI platform | Launched Dec 9, 2025; targeted at 3 million personnel | Department of War press release |
| Maven Smart System + Joint Fires Network | CJADC2 decision systems | $2.3B requested FY 2027 | DoW FY 2027 Overview Book |
| Maven (NGA) | Geospatial imagery analysis for target identification | Reviewed in GAO sample | GAO-26-107859 |
| Project Linchpin (Army) | Departmental AI infrastructure, test and evaluation | Reviewed in GAO sample | GAO-26-107859 |
| Cyclops (Army) | Computer vision targeting, XM-30 combat vehicles | Reviewed in GAO sample | GAO-26-107859 |
| NIPRGPT (Air Force) | Generative AI chatbot | Reviewed in GAO sample | GAO-26-107859 |
| Frontier AI other transaction agreements | Agentic AI prototyping | 4 vendors, $200M ceiling each | GAO-26-107859 |
The Department of War’s GenAI.mil launch release of December 9, 2025 states that AI capabilities reached desktops across the Pentagon and at American military installations worldwide, with Google Cloud’s Gemini for Government as the first model deployed and all tools certified for Controlled Unclassified Information at Impact Level 5.
The transparency gap. GAO-26-107859, published April 13, 2026, records something that ought to be quoted more often than it is: to build its sample, GAO reviewed a classified list of more than 200 AI-related activities at DOD — and noted that DOD, unlike most executive branch agencies, is not required to maintain and publish an AI use case inventory. Every civilian agency figure that circulates in federal AI adoption research has a public inventory behind it. The largest AI buyer in the federal government does not.
That single sentence is why credible military AI adoption statistics are scarce and credible military AI budget statistics are abundant. The appropriations documents are public by law. The deployment inventory is not.
The number that should end “AI is cheap” arguments
GAO’s report contains the most concrete unit-economics figure yet published on operational military AI. Officials from the Army’s XM-30 AI solution — enhanced targeting for minimally crewed infantry combat vehicles — told GAO that vendor licensing proposals reached around $300,000 per vehicle per year, which would have produced more than $500 million each year in licensing fees alone, separate from the cost of the vehicles. The Army rejected the proposal and is pursuing other licensing models.
Sarah Mitchell: That is the whole enterprise AI cost problem, translated into armour. Per-seat licensing does not survive contact with a fleet. It works when the unit is a knowledge worker and breaks when the unit is a vehicle that already costs millions and lives for thirty years. GAO’s Project Linchpin officials made the companion point: buyers price the model training and miss the infrastructure that has to run underneath it for the life of the platform. We have documented the same pattern on the commercial side in our AI investment statistics work — sticker price is the small number. The Army did the right thing here and said no, which is more than most enterprise buyers manage.
Which Companies Are Winning U.S. Military AI Contracts?
| Company | Agreement ceiling | Purpose | Contract type |
|---|---|---|---|
| OpenAI Public Sector | $200M | Frontier AI for warfighting + enterprise | Prototype OTA |
| Anthropic | $200M | Frontier AI national-security capabilities | Prototype OTA |
| Google Public Sector | $200M | Frontier AI national-security capabilities | Prototype OTA |
| xAI / AIQ Phase | $200M | Frontier AI national-security capabilities | Prototype OTA |
Axis Intelligence Research calculates a combined maximum agreement ceiling of $800 million across the four frontier-AI prototype agreements.
How Does U.S. Military AI Spending Compare Globally?
There is no comparable published AI line for any other military. China, Russia and most European states do not disaggregate artificial intelligence in their public budget documents, and Russia’s classified residual reached roughly 79% of its National Defence budget in 2025 according to SIPRI. Axis Intelligence Research will not construct a country-by-country military AI ranking from figures that do not exist. What can be done honestly is a scale comparison.
Using SIPRI’s April 2026 fact sheet, world military expenditure reached $2,887 billion in 2025, the eleventh consecutive year of growth, at 2.5% of global GDP. U.S. spending was $954 billion, 33% of the world total.
Where the U.S. AI-and-autonomy request would rank as a national military budget
| Rank | Country / line | 2025 spending | Source |
|---|---|---|---|
| 1 | United States | $954B | SIPRI |
| 2 | China | $336B | SIPRI |
| 3 | Russia | $190B | SIPRI |
| 4 | Germany | $114B | SIPRI |
| — | U.S. FY 2027 AI + autonomy request | $112.5B | Axis Intelligence Research |
| 5 | India | $92.1B | SIPRI |
| 6 | United Kingdom | $89.0B | SIPRI |
| 7 | Ukraine | $84.1B | SIPRI |
| 8 | Saudi Arabia | $83.2B | SIPRI |
| 9 | France | $68.0B | SIPRI |
| 10 | Japan | $62.2B | SIPRI |
| — | U.S. FY 2027 AI and CJADC2 line alone | $58.5B | Axis Intelligence Research |
| 11 | Israel | $48.3B | SIPRI |
Chart 5 — “The U.S. AI and autonomy request placed in the SIPRI 2025 national ranking” — horizontal bar with Axis lines highlighted. Source: SIPRI Fact Sheet April 2026; Axis Intelligence Research.
Axis Intelligence Research finds that a single fiscal year’s U.S. request for artificial intelligence and autonomous systems, at $112.5 billion, would place fifth among all national military budgets — above India, below Germany. The AI and CJADC2 line on its own, at $58.5 billion, would place eleventh, above Israel.
Two further readings from the same arithmetic: the combined request equals 11.79% of total U.S. military expenditure recorded by SIPRI for 2025, and 20.1% of the $559 billion spent by all 29 European NATO members combined.
This comparison mixes a fiscal-year budget request with calendar-year expenditure outturns compiled on a different methodology. It is a scale comparison, not an accounting identity, and we label it as such wherever it appears.
Sarah Mitchell: The rank-order framing is the one that will travel, so it is worth saying exactly what it does and does not mean. It does not mean the United States is spending more on military AI than Germany spends on its military — request is not outlay, and reconciliation has not passed. It means the ask is that size. And the ask is the signal industry reads. When a single line in one country’s budget request is larger than the fourth-largest military budget on earth, procurement officers in Berlin, Tokyo and Canberra are not reading it as an accounting curiosity. SIPRI’s own director noted that U.S. spending could rise to $1.5 trillion in 2027 if the proposal is accepted. That conditional is doing a lot of work, and it should stay in every citation of these numbers.
Where Do “Military AI Market Size” Forecasts Come From?
Axis Intelligence Research does not publish a military AI market size figure, and the reason is a finding rather than a gap.
Every widely circulated defense-AI market forecast traces back to subscription research whose methodology, sample and segment definitions are not public. Those forecasts cannot be reproduced, cannot be audited against a primary document, and disagree with each other by margins larger than the growth rates they project. A number that cannot be checked is not a statistic; it is a marketing input.
What can be verified is appropriated and requested budget authority, published by the Comptroller, line by line, with fiscal-year labels and prior-year comparatives. That is what this dataset contains. It is a narrower claim than “the global military AI market will reach $X billion by 2035,” and it is the only version of the claim that survives a fact-check.
For readers who need a defensible proxy for vendor-side revenue exposure rather than government-side authority, company filings are the right instrument — see our Palantir statistics research, where government segment revenue is drawn from SEC filings rather than forecast.
Military AI Market Size 2026: Why Estimates Range From $11.8B to $22.41B
| Publisher | 2026 military AI estimate | Long-term forecast | CAGR |
|---|---|---|---|
| MarkWide Research | $11.8B | $36.59B by 2035 | 13.4% |
| Visiongain | $14.49B | $45.18B by 2036 | 12.0% |
| Fortune Business Insights | $22.41B | $101.02B by 2034 | 20.7% |
Published 2026 military-AI market estimates in this sample range from $11.8 billion to $22.41 billion, a difference of $10.61 billion. The highest estimate is approximately 1.90× the lowest estimate.
What Sits Around the AI Line in FY 2027?
Two contextual movements matter for interpreting the AI figures.
Science and technology is shrinking while AI grows. The FY 2027 S&T request is $25.7 billion, down from $27.8 billion enacted in FY 2026 — a 7.6% decrease, attributed in the Overview Book to a $1.5 billion realignment away from underperforming and non-defense-specific research plus the effect of FY 2026 mandatory funding that pulled initiatives forward. Basic research (BA-01) falls to $2.1 billion. Axis Intelligence Research notes that the combined AI-and-autonomy request is 4.38 times the entire S&T account, and the AI and CJADC2 line alone is 27.9 times basic research.
Adjacent portfolios are moving in the same direction. Cyberspace activities rise to $20.5 billion from $15.1 billion requested in FY 2026. Space procurement and RDT&E reaches $59.7 billion, which the Overview Book describes as more than double the FY 2026 enacted level. The Office of Strategic Capital, at $20.2 billion, is scaled to crowd private capital into the 34 covered technology categories under 10 U.S.C. 149.
The Under Secretary of War for Research and Engineering designated six Critical Technology Areas in November 2025, with Applied Artificial Intelligence first among them. Readers tracking the defensive half of this equation will find the adversary-side numbers in our AI cyberattack statistics and cybersecurity spending statistics research; the governance and evaluation half is covered in AI safety statistics.
Military AI Timeline: Key Developments in 2025–2026
July 2025 — four frontier AI prototype agreements, up to $200M each.
December 9, 2025 — GenAI.mil launched, initially targeting access for 3 million personnel.
April 13, 2026 — GAO publishes AI acquisition review and reveals DOD’s classified list contains 200+ AI-related activities.
April 20, 2026 — FY2027 budget publishes the $58.5B AI/CJADC2 and $54B autonomy lines.
Methodology
Collection. All budget figures were extracted from two primary documents fetched and read on August 27, 2026: the FY 2027 Defense Budget Overview (Department of War, Office of the Under Secretary of War (Comptroller), generated April 20, 2026) and the FY 2026 Defense Budget Overview (Department of Defense, Office of the Under Secretary of Defense (Comptroller), generated July 7, 2025). Global expenditure figures come from the SIPRI Fact Sheet Trends in World Military Expenditure, 2025, published April 2026 with data current to December 31, 2025. Program and acquisition figures come from GAO-26-107859, published April 13, 2026. Enterprise deployment facts come from the Department of War press release of December 9, 2025.
Units. All dollar figures in the CSV are expressed in USD billions to keep one unit across the series; prose occasionally converts to trillions for readability ($1,450.0 billion = $1.45 trillion). No other transformation is applied.
Definitions. Budget authority is the legal authority to incur obligations; it is not an outlay and not an expenditure. Request is the President’s proposal; enacted is what Congress provided. FY 2027 figures throughout are requests. SIPRI figures are calendar-year expenditure estimates compiled on SIPRI’s own methodology, which differs from NATO’s and from U.S. budget accounting.
DAII™ formula and inputs.
DAII = (disclosed AI and autonomy budget authority / total budgetary resources) x 100
FY 2027 combined = (58.5 + 54.0) / 1450.0 x 100 = 7.76
FY 2027 AI only = 58.5 / 1450.0 x 100 = 4.03
FY 2027 autonomy = 54.0 / 1450.0 x 100 = 3.72
FY 2026 autonomy = 13.4 / 961.6 x 100 = 1.39
No weighting, normalisation or baseline adjustment is applied. Every input is a single published figure from the Overview Book of the corresponding fiscal year, so any reader holding those two PDFs can reproduce all four readings. The FY 2026 denominator is the FY 2026 request ($961.6B), matched to an FY 2026 request numerator, to keep the ratio internally consistent.
Derived figures. The unitemised residual ($1.2B) is the published AI line minus the sum of its published sub-bullets. Rank placements insert Axis-calculated dollar figures into SIPRI’s 2025 country table without altering SIPRI’s values. The infrastructure-to-application ratio divides the AI Arsenal line by the AI integration line. All arithmetic was independently recomputed in Python before publication; every result appears as a row in the accompanying CSV with a method_note.
Scope notes. Classified AI programs are excluded because they are not published. The combined AI-and-autonomy reading is an upper bound: the Department publishes no overlap reconciliation between the two lines, and Axis Intelligence Research declines to estimate one. Cross-national AI-specific comparison is not attempted, because no comparable AI line exists in any other national budget document we could verify.
About This Dataset
military-ai-statistics-2026.csv contains 115 observations covering U.S. Department of War AI and autonomy budget lines (FY 2024–FY 2027), global and national military expenditure for calendar 2025, federal AI acquisition figures, and 30 Axis Intelligence Research calculated rows including all DAII™ readings. Each row carries source_org, source_document, source_url, retrieved_date, is_primary and axis_calculated; calculated rows carry a method_note stating the formula.
License: CC BY 4.0. Free to reuse, republish and redistribute with attribution.
Mirrors: Hugging Face, Kaggle, GitHub.
Cite this research
APA — Axis Intelligence Research. (2026). Military AI statistics 2026: Defense AI spending, autonomy budgets and adoption data. https://axis-intelligence.com/military-ai-statistics/
MLA — Axis Intelligence Research. “Military AI Statistics 2026: Defense AI Spending, Autonomy Budgets and Adoption Data.” Axis Intelligence, 27 Aug. 2026, axis-intelligence.com/military-ai-statistics/.
Chicago — Axis Intelligence Research. “Military AI Statistics 2026: Defense AI Spending, Autonomy Budgets and Adoption Data.” Axis Intelligence, August 27, 2026. https://axis-intelligence.com/military-ai-statistics/.
Frequently Asked Questions
Is the $58.5 billion appropriated or only requested?
Requested. It is the FY 2027 President’s Budget figure published in the Department of War Overview Book on April 20, 2026. Congress has not enacted it. Roughly $46.0 billion of it — the AI Arsenal — is a mandatory investment that depends on a reconciliation bill passing. Any citation that presents $58.5 billion as money the Pentagon is spending is wrong.
Does the $58.5 billion AI figure overlap with the $54.0 billion autonomy figure?
Almost certainly, and the Department does not say by how much. The Overview Book presents them as separate bullets under “Accelerating Innovation” with no reconciliation. Autonomous systems are AI-enabled by definition, and the AI line funds autonomy enablers. Treat $112.5 billion as a ceiling and cite the two component figures when precision matters.
Why is there no official count of AI systems fielded by the Pentagon?
Because DOD is exempt from the requirement. GAO recorded in April 2026 that DOD is not required to maintain and publish an AI use case inventory as most executive branch agencies are, and that GAO’s own sampling relied on a classified list of more than 200 AI-related activities. Public adoption statistics for military AI are therefore structurally thinner than budget statistics.
What is the AI Arsenal?
The Overview Book’s name for a three-tiered investment establishing government-owned AI supercomputing infrastructure inside secure hardened data centres, combined with flexible commercial surge capacity, serving the Department of War and the Intelligence Community from the strategic core to the tactical edge. It is $46.0 billion of the $58.5 billion AI request — 78.63%.
How is the Pentagon buying frontier AI models, and under what contract vehicle?
Through other transaction agreements rather than FAR-based contracts. GAO records that in July 2025 DOD awarded OTAs with $200 million ceilings to four leading AI companies to accelerate adoption of advanced AI capabilities. OTAs offer speed and flexibility for non-traditional vendors; GAO has separately noted they carry reduced accountability and transparency relative to FAR-based contracts.
What happens to these numbers if reconciliation does not pass?
The autonomy line contracts sharply. Of the $54.0 billion requested, $39.2 billion — 72.59% — is tied to the Drone Dominance mandatory request, leaving $14.8 billion in discretionary budget authority. On the AI side, the $46.0 billion AI Arsenal is explicitly a multi-year mandatory investment. A discretionary-only outcome would leave roughly $12.5 billion of the AI line and $14.8 billion of the autonomy line.
What is CJADC2 and how much of the AI line does it represent?
Combined Joint All-Domain Command and Control — the effort to link sensors, decision systems and shooters across services and domains. The Overview Book identifies $2.3 billion for the Maven Smart System and Joint Fires Network to deliver CJADC2 capability, which is 3.93% of the $58.5 billion AI and CJADC2 line. The label pairs the two, but the funding is overwhelmingly AI infrastructure rather than command and control.
Why does per-vehicle AI licensing matter for defense budget forecasting?
Because it exposes a pricing model mismatch that fleet-scale programs cannot absorb. GAO documented Army officials receiving XM-30 proposals at roughly $300,000 per vehicle per year, which would have exceeded $500 million annually in licensing alone. Software licensing structures designed for per-seat enterprise deployment produce sustainment costs on armoured fleets that no procurement account is built to carry.
How does U.S. military AI spending compare with China’s?
It cannot be compared directly, because China publishes no AI-specific military budget line. SIPRI estimates total Chinese military expenditure at $336 billion in 2025, up 7.4%, with the United States spending 2.8 times as much overall. Any published figure claiming to state China’s military AI budget is an estimate built on undisclosed assumptions, and Axis Intelligence Research does not reproduce it.
Which single figure should a journalist or analyst cite from this research?
The DAII™ autonomy-only series: 1.39 in FY 2026, 3.72 in FY 2027. It is the only reading in this dataset built from an identically labelled line in two consecutive primary documents, which makes it the most defensible twelve months from now. Cite the combined 7.76 reading with its upper-bound caveat attached.