The United States Department of Defense has been funding artificial intelligence research since before most people knew the term. Its first large named program came in April 2017, when Deputy Secretary of Defense Robert Work created Project Maven. DoD has never reported a single "AI" total, so this article uses only figures that appear in published sources and says what each one covers.

Understanding where this money flows, which organizations control it, and which companies capture it is not merely an academic exercise. It is the central intelligence question for any investor, analyst, or policy professional operating in the defense technology sector in 2026.

Project Maven: Where It Started and Where It Went

Project Maven, formally the Algorithmic Warfare Cross-Functional Team, was established in April 2017 under Deputy Secretary of Defense Robert Work. Its initial mandate was operationally specific: apply computer vision AI to the processing of drone surveillance footage collected by MQ-1 and MQ-9 platforms in the Middle East theater. The Intelligence Community and DoD were drowning in raw video data — analysts estimated that only a fraction of collected ISR footage was ever reviewed by human analysts. Maven's task was to use machine learning to automatically identify objects of interest: vehicles, people, structures.

Google was awarded an initial contract under Project Maven in 2017. The partnership became public in early 2018 and triggered a significant employee rebellion inside Google, with over 3,000 employees signing an open letter demanding the company withdraw. Google declined to renew the contract in 2018. The episode became a landmark in the nascent debate over technology company participation in autonomous weapons programs — and it opened the door for a generation of defense-native AI startups explicitly built to serve the market Google vacated.

Project Maven did not end with Google's departure. The program migrated to Palantir Technologies, which took over the computer vision and data pipeline work, and later expanded to include contributions from Booz Allen Hamilton and a consortium of smaller specialized contractors. By 2022, Project Maven had evolved beyond its original drone-footage scope to encompass a broader AI-enabled targeting intelligence architecture, with operational deployments in multiple theaters. Current Maven iteration involves natural language processing for signals intelligence fusion and multi-source target development workflows.

$13.4BFY2026 request for AI and autonomy, the first year DoD listed autonomy as its own budget section (DefenseScoop, June 2025)

The CDAO: Organizational Architecture of AI Spending

The organizational chaos of early DoD AI spending — fragmented across service branches, DARPA, the Intelligence Community, and multiple OSD offices — prompted a consolidation effort. In 2022, the Chief Digital and Artificial Intelligence Office (CDAO) was established under the Office of the Secretary of Defense, merging the Joint Artificial Intelligence Center (JAIC), the Defense Digital Service, the DoD Chief Data Officer, and Project Maven into a single entity reporting directly to the Deputy Secretary of Defense.

Much AI spending sits in the services' own accounts rather than in one central line: the FY2026 autonomy request alone was spread across 348 budget lines, according to Inside Defense. The CDAO's mandate includes:

The CDAO's creation reflected a recognition that AI investment fragmentation was producing capability duplication, interoperability failures, and procurement inefficiencies. It also created a single procurement interface that has substantially benefited larger contractors with the organizational capacity to engage enterprise-scale contracts.

The Replicator Initiative: Autonomous Systems at Industrial Scale

Announced by Deputy Secretary of Defense Kathleen Hicks in August 2023, the Replicator Initiative represents the most explicit statement of DoD intent to field autonomous systems at scale in the current budget environment. The program's stated objective is to field "multiple thousands of autonomous systems across multiple domains" within 18 to 24 months — a procurement timeline that is extraordinary by Pentagon standards.

The impetus for Replicator is explicit: the lessons of Ukraine combined with the growing assessment of Chinese military buildup toward a potential Taiwan contingency. Pentagon planners assessed that in a high-intensity Pacific conflict scenario, the attrition rate for expensive, low-volume precision platforms (aircraft, ships, missiles) would be unsustainable without a complementary force of cheap, expendable autonomous systems that could impose costs on adversary forces without requiring commensurate investment.

Replicator received about $500 million in FY2024, with a similar amount budgeted for FY2025 — about $1 billion over two years, according to then-Deputy Secretary Kathleen Hicks (DefenseScoop (Mar 11, 2024)). It is not a single program of record: programs that have their own budget lines fall under the initiative. The program has focused on three domains:

The contractors benefiting most directly from Replicator awards include Anduril Industries, whose Roadrunner-M interceptor and Altius loitering munition systems were among the earliest Replicator awardees, and multiple small UAS manufacturers including Shield AI and Area-I (an Anduril acquisition).

The Contractor Landscape: Primes vs. Tech Natives

The most significant structural shift in DoD AI contracting over the past decade is the ascendance of technology-native defense startups at the expense of traditional prime contractors' AI programs. This shift is not absolute — Lockheed Martin, Northrop Grumman, L3Harris, and Raytheon all operate substantial AI divisions — but the growth rates, valuations, and contract win rates increasingly favor companies built around software-first architectures.

Anduril Industries

Founded in 2017 by Palmer Luckey, Anduril is a venture-backed defense technology company. Its Lattice AI platform serves as the integrating operating system for its hardware products — the Sentry Tower surveillance system, the Ghost autonomous reconnaissance aircraft, the Altius-600M loitering munition, and the Roadrunner-M interceptor. Anduril won a significant portion of SOCOM's counter-UAS contract in 2022 and has been awarded positions on multiple classified programs.

Palantir Technologies (NYSE: PLTR)

Palantir's AI Platform (AIP) and its government business span intelligence community contracts, the Army's TITAN ground station program (a key Maven successor architecture), and the Maven Smart System itself. Palantir's AI Platform for defense applications ingests multi-source data and produces decision-support outputs for commanders at operational and tactical levels.

Shield AI

Shield AI has built its business on autonomous flight software — specifically the Hivemind AI pilot that allows aircraft to operate without GPS, communications links, or human input during contested missions. Hivemind has been demonstrated on the F-16, the MQ-35A V-BAT drone, and other platforms. Shield AI's Nova unmanned aerial system won multiple military contracts for building clearance and indoor autonomous navigation.

Traditional Primes: AI Integration Status

The traditional defense primes have responded to the AI-native challenge with varying degrees of agility:

DARPA's AI Weapons Portfolio

The Defense Advanced Research Projects Agency remains the frontier of DoD AI weapons research, with several programs currently advancing toward potential acquisition:

ACE: Air Combat Evolution

The Air Combat Evolution program has been the most publicly visible DARPA AI weapons initiative. ACE funds the development of AI agents capable of executing within-visual-range air combat maneuvering autonomously. In a landmark 2020 simulation, DARPA's AI (developed by Heron Systems) defeated a human F-16 pilot 5-0 in simulated dogfights. The program has since progressed to live-aircraft testing, with an AI-controlled F-16 conducting actual flight test sorties at Edwards Air Force Base under the X-62A VISTA program. The AI pilot completed dogfighting maneuvers against a human-piloted F-16 in 2023 — the first such real-world test in history.

Gremlins

The Gremlins program, led by Dynetics (an Leidos subsidiary), develops air-launched, air-recoverable unmanned aircraft that operate as swarms. The concept enables a single manned aircraft to release multiple autonomous UAS that conduct ISR or EW missions and are subsequently recovered mid-air by a C-130 transport aircraft. Gremlins addresses the cost-of-attrition problem differently from expendable platforms: the drones are designed for 20+ reuses, amortizing manufacturing cost across multiple sorties.

LongShot

LongShot is DARPA's program to develop an air-launched autonomous vehicle capable of carrying and releasing air-to-air munitions. By creating an intermediate platform between the launching aircraft and the target, LongShot extends the effective engagement range of air-to-air missiles while keeping the manned aircraft outside adversary threat envelopes. General Atomics and Northrop Grumman have both received development contracts. LongShot represents the next evolution of the "loyal wingman" concept toward genuine autonomous weapons employment.

Investment Implications: Where the Money Actually Goes

For investors seeking exposure to DoD AI spending, the landscape breaks into three tiers with different risk/return profiles:

PLTR Palantir Technologies Direct AI gov exposure
NOC Northrop Grumman B-21, autonomous systems
LMT Lockheed Martin F-35 AI integration
LHX L3Harris Technologies EW / counter-UAS AI
LDOS Leidos Holdings Gremlins / intel AI
RTX RTX Corporation AI seeker tech

The private market holds the highest-growth positions: Anduril, Shield AI, and Sarcos Technology (exoskeleton AI) are all pre-IPO with significant defense contract backlogs. Secondary market access through platforms such as Forge Global or direct venture allocation through defense-focused funds (including the emerging category of NATO-aligned defense VC funds) represents the primary institutional path to these positions before public listings.

Market Assessment

The structural shift from traditional primes to tech-native defense contractors mirrors the broader transition in the U.S. economy from hardware-intensive to software-intensive value creation. Defense prime contractors are not disappearing — they manufacture the platforms. But the premium valuations are accruing to the software companies that make those platforms intelligent. Palantir's market cap exceeding Northrop Grumman's (at peak) in 2025 was the market's verdict on where AI-era defense value accretes.

FY2027: What the Budget Request Shows

The FY2027 request is far larger than any earlier year, but sources count it differently. The Congressional Research Service lists $53.6 billion in mandatory funding for the new Defense Autonomous Warfare Group (DAWG) and $29.5 billion for an "Artificial Intelligence Arsenal" initiative. Greenberg Traurig cites nearly $55 billion for DAWG and $58.5 billion for AI plus joint command-and-control; immixGroup cites $54.6 billion for DAWG and $63 billion for autonomous and unmanned systems overall. The totals differ because of scope, mandatory versus discretionary money, and multi-year amounts. These are requests, not appropriations.

What the money will buy, and how much Congress will provide, is not settled. The largest FY2027 items are proposed mandatory funds, which depend on legislation.

Key Takeaways // Intelligence Summary
  • The FY2026 request listed $13.4 billion for AI and autonomy, the first year autonomy had its own budget section. The FY2027 request is far larger, but sources count it differently (see above).
  • Project Maven migrated from Google to Palantir after the 2018 employee revolt — creating the template for defense-native tech companies as primary beneficiaries of DoD AI contracts.
  • The Replicator Initiative (about $500M in FY2024 and a similar amount budgeted for FY2025) is DoD's most explicit stated commitment to fielding autonomous systems at scale, motivated by lessons from Ukraine and planning for a China contingency.
  • Anduril, Palantir and Shield AI are among the technology-native companies with prominent defense AI work.
  • DARPA reported in April 2024 that an AI-flown X-62A (a modified F-16) flew within-visual-range dogfights against a human-piloted F-16 in 2023.