The United States Department of War just proposed giving the Defense Autonomous Warfare Group a $54.6 billion budget for fiscal year 2027. That is nearly the size of the entire United States Marine Corps budget, which stands at $57.2 billion in FY26. A group that has existed for less than a year, stood up with $225 million and a portfolio of inherited Replicator contracts, is now being positioned to command more resources than a service branch that has been fighting the nation's wars since 1775.
This 24,166% year-over-year increase matters not only because of its size, but because of what it signals. This is not an incremental funding adjustment. It is a structural declaration about how the Pentagon intends to fight future wars. It is a structural declaration that the Pentagon views autonomous warfare as a distinct warfighting domain.
For those unfamiliar with DAWG, the Defense Autonomous Warfare Group is the successor to the Replicator initiative, housed under U.S. Special Operations Command and was led until recently by Marine Corps LtGen Frank Donovan. Its focus is on larger autonomous platforms across air, surface, and subsurface domains, including one-way attack drones and unmanned surface vessels. It operates alongside two parallel efforts: the Drone Dominance program, which handles Group 1 FPV drones and small UAS at scale, and Joint Interagency Task Force 401, which leads counter-drone training and strategy.
More importantly, this budget tells us where the Pentagon wants to go. It suggests the institution is finally internalizing lessons from Ukraine, albeit later than it should have. The harder question is whether the structure behind that number can deliver what it promises, or whether $54.6 billion produces a new warfighting command or just a very expensive program office.
What DAWG Inherited and What It Replaced
DAWG did not emerge from a clean sheet. It inherited the ambitions and unfinished business of the Replicator initiative, which launched in 2023 under Deputy Secretary of Defense Kathleen Hicks with approximately $1 billion in funding and a two-year timeline to field 'thousands' of autonomous systems across multiple domains. The ambition was real and the goal was necessary. As with many Pentagon initiatives, however, execution did not match the rhetoric.
Congressional Research Service found that Replicator delivered hundreds, not thousands, of systems by its August 2025 target. The shortfall was not just numerical. The portfolio also exposed recurring technical and integration problems, from autonomous vessel mishaps to unreliable multi-vendor coordination software and platform demonstrations that failed to inspire confidence.
Replicator was designed to move around the traditional acquisition system and leaned heavily on nontraditional firms, with 75 percent of participants coming from outside the legacy defense industrial base. In the end, it ran into the same frictions it was meant to bypass: vendor coordination problems, fragmented testing opportunities, and no clear operational home for the systems once they were delivered.
The Pentagon moved the entire Replicator portfolio from the Defense Innovation Unit to DAWG, newly formed under SOCOM. LtGen Donovan, a Marine General with deep operational credibility and experience as SOCOM's vice commander, was named director. Autonomous warfare needs an operator at the helm, and LtGen Donovan understands better than most in DoW how these systems would be employed in a fight.
DAWG's lane has been carved alongside two complementary but separate efforts. The Drone Dominance Program, run through DIU under Owen West, focuses on Group 1 FPVs and small UAS at scale, with a stated goal of fielding more than 300,000 domestically produced, weaponized drones. JIATF-401, led by the Army, coordinates counter-drone training, strategy, and evaluation. Conceptually, DAWG occupies the space above small drones and below traditional manned platforms: larger one-way attack drones, unmanned surface vessels, and critically, the orchestration layer that ties autonomous systems together into something a commander can actually use.
This distinction is important, as it reflects how the Pentagon is planning on segmenting autonomous warfare into three functional pillars, each with its own center of gravity. For small drones, the primary challenge is production and distribution at scale. For C-UAS, it is training, doctrine, and integration. For larger autonomous systems, it is command and control. All three share overlapping demands in logistics, industrial base, and workforce development, but the defining problem for each shapes how it must be organized and led. DAWG sits at the intersection of all three pillars, but its primary challenge is the third: how to make autonomous fleets useful to commanders in contested, degraded or denied C2 and GNSS, and chaotic battlefield environments.
Emil Michael, the Pentagon's senior technology and research official, framed it clearly at the Reagan National Defense Forum: "We have to be dominant in both," referring to small and large drones. "What we've learned from the Ukraine-Russia war is that the front lines of a conflict over territory are robot on robot now." His statement captures the operational logic behind the segmentation: small drones are the close fight, and larger autonomous systems, the ones DAWG is responsible for, are the deep fight, the shaping fight, and the fight for logistics and sustainment that determines whether forces can stay in the field.
This Budget Points to a Command, Not a Group
There is no precedent in recent defense budgeting for a working group receiving $54.6 billion in a single fiscal year. That level of spending authority is usually reserved for service branches or operational commands with established doctrine, organic forces, training pipelines, and acquisition authority. For context, the proposed DAWG budget accounts for 15 percent of the Pentagon's $350 billion reconciliation budget proposal for FY27. It falls just short of the entire Marine Corps budget in FY26. Taken together, the topline and the absence of a public spending breakdown suggest DAWG is being positioned for a mission far larger than its current name implies.
Aviation Week's analysis notes the spending proposal, combined with the organizational signals coming from senior Pentagon leadership, points toward the creation of a new Autonomous Warfare Command. Such a command would be a joint organization focused on orchestrating autonomous systems across domains, operating drone and unmanned vessel fleets in scenarios like the Pacific "Hellscape" concept that envisions swarms of autonomous platforms denying Chinese naval forces freedom of maneuver in contested waterways.
The reality is this is not a revolutionary concept for a modern military. Ukraine stood up its Unmanned Systems Force as a dedicated military branch for autonomous warfare, built under the pressure of conflict, and has evolved from an improvised collection of volunteer drone units into a formalized force with its own chain of command, training infrastructure, and direct ties to the defense industrial base. In similar fashion, Russia developed its Rubicon complex to serve a parallel function, integrating drone operations across its military services based on lessons absorbed from three years of intensive combat.
Senator Roger Wicker, chairman of the Senate Armed Services Committee, provided additional clarity during a March 5 hearing focused on the small drone industry and the U.S. military. Wicker differentiated DAWG's role from the Drone Dominance program, noting that "while the Drone Dominance efforts focus on smaller, Group 1 first-person-view drones, the DAWG will lead on larger one-way attack drones and small unmanned boats." This portfolio is not a niche, but rather spans air, surface, and subsurface domains. It does not fit neatly under any existing service, and it requires its own organizational home lest it fail before it even gets started.
A useful comparison is the trajectory Space Command followed from supporting function to command-level priority and eventually to a separate service. Autonomous warfare may not follow the same path, but this budget suggests it is moving in that direction. If Congress approves both the funding and the accompanying authorities, DAWG's transition from working group to joint command will become a question of execution more than concept.
None of this is guaranteed. A $54.6 billion proposal with no publicly available spending breakdown will face serious scrutiny from the Senate and House Armed Services Committees and the appropriators behind them. Members will ask why a SOCOM working group needs a budget approaching a service branch, and whether this duplicates the autonomous programs the services are already running independently. The Navy has its own unmanned surface vessel programs under PEO Unmanned and Small Combatants. The Air Force is developing Collaborative Combat Aircraft. The Army is building its own drone marketplace and has invested heavily in counter-UAS through JIATF-401. Each of these has existing funding, acquisition authorities, and institutional champions who will resist ceding resources or mission space to a new joint organization. DAWG's path to becoming a command will run through these competing equities, not around them.
Another indicator is the $100 million Autonomous Vehicle Orchestrator Prize Challenge, launched jointly by DIU, DAWG, and the Navy in January 2026. This challenge is designed to develop a technology layer that translates a commander's intent, expressed through natural language, into machine execution across mixed fleets of autonomous platforms. You do not invest $100 million in a C2 layer for drone swarms unless you expect to command those swarms as a standing force with persistent operational responsibilities.
The Orchestrator Problem
Buying autonomous platforms is the easy part, but overseeing and commanding them is much harder. The $100 million Autonomous Vehicle Orchestrator Prize Challenge is the clearest admission from inside the Pentagon that they can procure drones and USVs but cannot yet command them at the scale and tempo that modern conflict demands.
LtGen Donovan described the requirement in terms that deserve careful attention: "We want orchestrator technologies that allow humans to work the way they already command, through plain language that expresses desired effects, constraints, timing, and priorities, not by clicking through menus or programming behaviors." This is both an elegant statement of intent, and is also a capability that does not yet exist in any fielded U.S. system.
The Orchestrator must solve several problems simultaneously. It must be vehicle agnostic, meaning it works across platforms from different manufacturers with different communication protocols, autonomy stacks, and sensor suites. It must be domain agnostic, coordinating assets in the air, on the surface, and beneath the water through a single interface. And it must be degradation tolerant, continuing to function when communications are jammed, GPS is denied, link connections are intermittent, and platforms are attriting in real time. These are not software engineering challenges alone, but are doctrine problems wrapped in a complex technological challenge.
What makes the prize challenge particularly interesting is the commercial solutions already building toward significant portions of this capability. JIATF-401 has already selected Anduril's Lattice as its enterprise tactical C2 platform for counter-UAS; the Army chose it for the Integrated Battle Command System Maneuver program after live-fire demonstrations at Yuma Proving Ground produced a four-for-four intercept record; and Anduril's recent $20 billion enterprise contract with the Army consolidates more than 120 separate procurement pathways under Lattice's architecture.
Shield AI's Hivemind has demonstrated rapid integration across a widening range of platforms, from Northrop Grumman's Talon IQ testbed to GA-ASI's MQ-20 Avenger to Destinus's Hornet counter-UAS system. Shield AI's recent $2 billion Series G round at a $12.7 billion valuation reflects the market's confidence that autonomous piloting and multi-ship coordination are near-term operational capabilities.
These two companies represent the leading edge of what commercial industry can deliver right now. Both will almost certainly bid on the Orchestrator challenge itself. The question for DAWG is whether the prize challenge builds on what these platforms have demonstrated or starts from scratch with requirements that ignore the commercial momentum already underway.
The war in Ukraine offers an additional and equally relevant comparison. Ukraine's Delta system solved a version of the orchestration problem by building from the bottom up. Civilian software engineers worked alongside frontline operators to create a platform that was iterative, imperfect, and immediately useful. Delta did not try to unify every system in the Ukrainian military, but focused on connecting what was needed and left out what was not. It worked on commercial tablets and smartphones. It allowed any unit to become a reporting node, feeding real-time data into a shared operational picture that shortened the kill chain to seconds.
Delta succeeded because it was designed for the fight, not to meet acquisition milestones. It embraced simplicity over elegance and speed over completeness and grew through constant use and direct feedback from operators who had immediate reasons to make it work. That model, where the people fighting the war have a voice and are able to shape the tools they use, is the engine that makes battlefield software effective.
The Orchestrator must follow a similar logic. If it becomes a traditional program of record with layered requirements, multiple vendors competing through years of integration testing, and a product that arrives polished but late, it will repeat the pattern that has defined too many Pentagon software initiatives. The prize challenge format, with iterative sprints and down-selection based on demonstrated performance, is designed to prevent this, but whether that structure holds against institutional gravity remains an open question. The Pentagon has a long history of starting fast and finishing slow once program offices, requirements boards, and oversight layers attach themselves to a promising initiative (JADC2 anyone?).
There is also a human dimension to the Orchestrator problem that technology alone cannot solve. Commanding swarms of autonomous systems is fundamentally different from commanding individual platforms or manned units. It requires operators who think in terms of effects and constraints rather than direct control of each asset. A commander using the Orchestrator must be able to express intent ("deny this strait to surface traffic for six hours using available USVs"), set boundaries ("no engagement within two nautical miles of commercial shipping"), and then trust the system to translate that guidance into coordinated autonomous action. That is a training and culture shift, not a software upgrade, and it requires years of development in how operators learn to command well before the Orchestrator reaches full capability.
LtGen Donovan made one additional point that should not be lost in the technical discussion: "We must ensure the human, the ethical decision maker, always maintains a clear understanding of what the system is doing and why." This is not a compliance statement but rather is an operational requirement. Autonomous systems that operate beyond the comprehension of their commanders are liabilities, not force multipliers. The Orchestrator must make the complexity of multi-domain autonomous operations legible to humans who are under stress, operating with incomplete information, and making decisions with strategic consequences.
What the Budget Does Not Fund
Most commentary on DAWG's proposed budget will focus on the topline number or debate whether autonomous warfare deserves its own command. Those are valid discussions, but they skip over the harder question: what happens once the money starts flowing and the platforms start arriving? A $54.6 billion budget creates demand, but it does not, by itself, create the institutional infrastructure required to turn that demand into operational capability.
Logistics and sustainment: An autonomous fleet of the scale implied by this budget must be powered, maintained, updated, repaired, and replaced continuously. Every drone, USV, and UUV in the DAWG portfolio requires batteries, charging infrastructure, diagnostic tools, spare components, firmware management, and trained maintainers. At scale, this becomes an industrial logistics challenge comparable to managing a fleet of manned aircraft, but with a fundamentally different consumption model. Autonomous platforms are attritable and designed to be lost, which means the sustainment system must absorb losses and regenerate capability at a rate that matches operational tempo, not depot timelines.
The budget may fund procurement, but procurement is only the visible tip of the enterprise. Someone still has to build forward charging infrastructure, manage battery logistics across thousands of platforms, and train maintainers for systems designed to attrite at scale.
Testing infrastructure: A command-level organization responsible for autonomous warfare across multiple domains needs dedicated test environments where systems can be evaluated against realistic threats at realistic scale. This includes environments where kinetic counter-UAS engagements are legal and instrumented; where USVs can operate in open water against simulated threats; and where multi-domain orchestration can be exercised without the airspace and safety constraints that limit most domestic ranges.
The deeper problem is that Pentagon's testing infrastructure was largely built during WWII and the Cold War to evaluate individual weapon systems in single domains. It was never designed to test the kind of multi-domain autonomous kill chains that DAWG will be responsible for orchestrating. The result is a testing enterprise that operates episodically, through periodic exercises and vendor demonstrations, rather than as a persistent capability. Systems arrive for evaluation, wait weeks or months for range time, undergo testing against threat profiles that may not reflect current operational conditions, and produce narrative reports rather than structured, machine-readable performance data that can inform rapid acquisition decisions. That cycle is fundamentally incompatible with autonomous platforms that iterate in weeks and attrite in days.
What DAWG needs is persistent, software-enabled test infrastructure that can evaluate autonomous systems against realistic threats, collect structured performance data, and shorten the gap between iteration and fielding. The capability exists. The barrier is adoption and access.
The regulatory environment compounds the problem. Current FAA regulations limit kinetic C-UAS operations to the four federal departments with statutory authority (Defense, Homeland Security, Justice, and Energy), effectively restricting private test ranges to detection and tracking evaluation only. A DAWG portfolio that includes one-way attack drones, autonomous USVs, and counter-UAS interceptors cannot be adequately tested without environments where these systems engage realistic threats under controlled, instrumented conditions. A national network of civilian-run, certified, instrumented test ranges with standing regulatory authorities for kinetic and electronic warfare testing must be a prerequisite for fielding autonomous systems at the speed and scale this budget implies.
Industrial base depth: The domestic production capacity for batteries, motors, sensors, radios, microelectronics, and the thousands of small components required to sustain autonomous fleets is not yet scaled for the demand that a command-level budget implies. Most military-grade batteries are either sourced from or use components manufactured overseas, and supply chains that depend on foreign components become strategic vulnerabilities the moment conflict begins.
The Blue UAS list, originally designed to vet platforms for cybersecurity and supply chain compliance, continues to function as a barrier to entry for startups and smaller manufacturers that cannot afford the lengthy, expensive certification process. The budget creates demand, but if the industrial base cannot meet it with domestic production at speed and scale, the money flows to a small number of established vendors who may not deliver the innovation or cost discipline that attritable warfare requires. Battery production, semiconductor sourcing, motor manufacturing, and sensor fabrication all must be treated as strategic industrial priorities on the same level as shipbuilding or munitions production, and that investment must run in parallel with platform procurement, not as an afterthought.
Leadership continuity: LtGen Donovan was the architect of DAWG and brought enormous operational credibility from his tenure as SOCOM's vice commander. He understood the autonomous warfare problem from a warfighter's perspective and gave the organization its early direction and institutional backing. He has now moved on, confirmed as a four-star and assigned as commander of U.S. Southern Command. His departure creates a gap at a critical moment. DAWG needs a successor with equivalent operational credibility and institutional standing, and it needs that person in place quickly, lest a leadership vacuum during the transition from a $225 million group to a $54.6 billion organization invite exactly the kind of bureaucratic infighting and mission drift that killed momentum in Replicator.
Doctrine and career paths: If autonomous warfare becomes a command, it needs people who build careers in it. That means dedicated occupational specialties for drone operators, USV controllers, autonomy engineers, swarm coordinators, and the maintainers who keep these systems in the fight; training pipelines that produce qualified personnel at the scale the command requires; professional military education that treats autonomous warfare as a discipline, not an elective; and a promotion pathway that rewards expertise in autonomous operations rather than treating it as a detour from traditional career tracks.
Without this investment in people, the command will borrow talent from services that have their own priorities, their own retention challenges, and their own cultural preferences. Ukraine's Unmanned Systems Force recruited heavily from the civilian tech sector and built training programs from scratch under wartime pressure. Services will need to come together and do something analogous, adapted to their own institutional structures but driven by the same recognition that autonomous warfare requires consolidated a dedicated workforce, not a collection of temporarily assigned personnel.
The Reciprocal View
If the United States stands up an Autonomous Warfare Command, it will not do so in isolation. China is already building its own autonomous warfare infrastructure, with heavy investment in autonomous swarm development, USV fleets, AI-enabled targeting, and the manufacturing base to produce these systems at scale. A U.S. Autonomous Warfare Command is unlikely to slow Chinese investment and may instead reinforce the competitive cycle already underway. The same applies to Russia, which has used the pressure of combat in Ukraine to build and refine its Rubicon complex for integrated drone operations.
The symmetry here is structural, not coincidental, as both sides in a future fight will field large autonomous fleets. Both sides will target the other's logistics nodes with unmanned systems designed to delay, disrupt, and degrade sustainment. Both sides will struggle with the same C2 challenges: how to direct mixed fleets of autonomous platforms across multiple domains in contested, communications-degraded environments. Both sides will face industrial base constraints when attrition rates exceed peacetime production capacity.
The competitive advantage in this environment will not go to the side that buys the most platforms, but rather to the side that has built an adaptive enterprise behind them. An autonomous fleet whose logistics can absorb attrition and keep platforms in the fight will outperform one that fields more systems but cannot sustain them. A command whose software updates faster than the adversary adapts will maintain decision advantage. A force whose C2 layer functions under degraded conditions while the adversary's fractures will control the tempo of the fight. This is an organizational and cultural outcome, not a procurement outcome, and the side that treats autonomous warfare as a buying program will discover that platforms without adaptive sustainment and responsive command systems are expensive inventory, not combat power.
The $54.6 billion FY27 budget proposal is the clearest signal yet that the Department of War views autonomous warfare as a primary warfighting domain, not a supporting capability bolted onto existing force structures. The trajectory from Replicator to DAWG to a potential Autonomous Warfare Command is a fundamental shift in how the United States intends to organize, equip, and fight with unmanned systems across every domain.
Replicator showed what happens when ambition outpaces institutional readiness. It set the right goals but could not overcome the friction of vendor coordination, testing gaps, and the absence of an operational home. DAWG has the chance to learn from that experience, but only if it builds the infrastructure that Replicator lacked: living logistics that sustain autonomous fleets under pressure, an Orchestrator that commands them at the speed of the fight, a testing enterprise that evaluates them against realistic threats, an industrial base that replaces them as fast as they are consumed, and leaders who stay long enough to see the transformation through.
A $54.6 billion budget can buy a large inventory of autonomous systems. It cannot, by itself, build the command architecture, logistics base, testing infrastructure, and trained workforce needed to turn those systems into combat power. That is the real test of whether DAWG is becoming a warfighting command or simply a better-funded program office.