In August 2023, then-Deputy Secretary of Defense Kathleen Hicks stood at the Reagan National Defense Forum and announced a program that, she claimed, would "field attritable, autonomous systems at a scale of multiple thousands, in multiple domains, within the next 18 to 24 months." The program had a name designed to signal ambition: Replicator. The message was unambiguous — the United States, watching the drone war in Ukraine and the growing autonomous weapons buildup in China, was about to demonstrate that it could manufacture and deploy autonomous systems at production scale.
That was August 2023. Now, in early 2026, Replicator has reached two significant milestones: Replicator 1 declared its fielding goal of thousands of uncrewed systems achieved as of August 2025, and Replicator 2's first acquisition was announced on January 11, 2026. The question facing defense analysts — and Congress, which has never been fully briefed on the program's specifics — is what "fielded" actually means, whether the numbers represent genuine military capability, and whether this program is the structural answer to America's autonomous weapons gap or a costly demonstration exercise.
Genesis: Why Replicator Exists
The Replicator initiative was born from a specific threat analysis: the realization that China's PLA, in a Taiwan Strait contingency, could overwhelm American and allied defensive systems through sheer numerical saturation. The U.S. military's conventional weapons inventory — exquisite, expensive, and produced in limited quantities — was designed for the threat environment of the Cold War and the post-Cold War era of uncontested American superiority. Against an adversary capable of fielding thousands of cheap, expendable weapons simultaneously, that inventory is inadequate.
The concept behind Replicator was straightforward: counter mass with mass. If China can flood a battlespace with hundreds of autonomous naval vessels, drone swarms, and loitering munitions, the United States must be able to field comparable numbers of cheap, capable, attritable autonomous systems. The emphasis on "attritable" — meaning systems designed to be lost in combat without catastrophic cost — represents a fundamental philosophical shift from the DoD's traditional preference for highly capable, highly expensive, long-life platforms.
The Ukraine war provided an accelerant. Watching Ukrainian FPV drones costing $400 destroy Russian armor costing millions, watching naval drones costing $500,000 damage warships worth hundreds of millions — the lessons were impossible to ignore. The Replicator program crystallized from a pre-existing set of autonomous systems initiatives and was given a name, a timeline, and political visibility at the Deputy Secretary level.
Replicator 1: The All-Domain Attritable Push
Replicator 1 targeted "all-domain attritable autonomous systems" — a deliberately broad mandate that encompassed air, sea, land, and subsurface platforms. The program sought to field "thousands of uncrewed systems" by August 2025. That deadline has now passed, and the DoD has declared the target met — but with significant caveats.
The systems counted toward the Replicator 1 goal span multiple platforms and multiple services. Air Force contributions include XQ-58A Valkyrie-class attritable combat drones and smaller Group 1–3 UAS platforms. Navy contributions include medium-displacement unmanned surface vehicles and autonomous underwater vehicles. The Army contributed autonomous ground vehicles for logistics and ISR missions. Special Operations Command, which already held a 10-year, $967 million IDIQ contract with Anduril for counter-drone hardware, reported its systems as part of the count.
Critically, the Replicator 1 "fielding" target was not specifically defined in publicly released documents as meaning "combat-ready" or "operationally deployed." Congressional Research Service analysis notes that the program's lack of public specificity about exactly which systems count, in what quantities, and to what readiness standard makes independent verification of the achievement claim impossible. Responsible Statecraft's reporting noted that DoD Secretary Pete Hegseth's congressional testimony referenced "unmanned systems across multiple domains" with "thousands more planned" through the FY 2026 budget — suggesting the achievement is real but the scale may be more modest than the headline implies.
Replicator is one of the least transparently documented major DoD programs in recent history. The Deputy Secretary's announcement in August 2023 provided no specific system types, unit counts, unit costs, or performance requirements. Subsequent Congressional briefings have been largely classified. The GAO has not conducted a public review. Independent analysts are left reconstructing the program's status from contract announcements, service acquisition notices, and congressional testimony fragments. This opacity — unusual for a program of this political visibility — has generated significant criticism from oversight advocates and leaves the program's actual achievements genuinely uncertain from an open-source perspective.
The Industry Ecosystem: Anduril, Shield AI, L3Harris, and General Atomics
Whatever its transparency limitations, Replicator has generated a significant wave of contracting activity that reveals the program's key industrial partners and technical directions. The ecosystem that has emerged around Replicator represents the most significant restructuring of the defense industrial base for autonomous systems since the drone era began.
Anduril Industries: The Replicator Anchor
Anduril Industries, the defense tech startup founded by Palmer Luckey, has emerged as arguably the most consequential Replicator partner. The company's involvement operates across multiple Replicator-related programs simultaneously. Anduril received a prototype contract for the ACT (Autonomous Collaborative Technologies) software program in November 2024 — designed to coordinate thousands of uncrewed systems across the Replicator fleet, providing the networking and command-and-control layer that makes a swarm a swarm rather than a collection of independent platforms.
Additionally, Anduril secured a $642 million Marine Corps contract for its counter-drone system, incorporating the Roadrunner-M missile interceptor and Anvil kinetic kill drone — systems directly relevant to the Replicator 2 counter-small-UAS mission. Anduril also holds the 10-year $967 million SOCOM counter-drone IDIQ. The company's Lattice AI platform — designed as the brain that coordinates autonomous systems in complex environments — positions it as a potential software backbone for the entire Replicator ecosystem.
Anduril was valued at $30.5 billion in June 2025. Its success in penetrating the Replicator program while maintaining a startup's development speed represents exactly the model that DoD acquisition reformers have been trying to enable for two decades — whether the experiment scales is the critical question for the program's future.
Shield AI: Autonomy Without GPS or Comms
Shield AI's core technology — the HIVEMIND AI pilot system — is designed to enable autonomous flight in GPS-denied, communications-denied environments. This capability is specifically relevant to high-end conflict scenarios against adversaries with capable EW systems. An autonomous drone that can only operate with GPS and datalinks is vulnerable to electronic warfare; a drone running Shield AI's HIVEMIND can navigate, form up, and engage targets using onboard sensors alone.
Shield AI received contracts to develop the autonomy layer for the Air Force's Collaborative Combat Aircraft (CCA) program — the crewed-uncrewed teaming initiative running parallel to Replicator. The company's V-BAT autonomous VTOL drone has been acquired by multiple services and is counted among Replicator-relevant systems. Shield AI's approach — prioritizing autonomy at the edge rather than cloud-connected coordination — complements Anduril's Lattice-based network architecture, with the two companies effectively covering different operational scenarios: Anduril when communications are available, Shield AI when they aren't.
L3Harris: The Cognitive EW and Swarm Integration Bridge
L3Harris received a prototype contract for the ACT software program alongside Anduril and Swarm Aero in November 2024. Its role within Replicator spans multiple capability areas: cognitive electronic warfare modules for platform protection, communications systems for swarm networking, and signals intelligence processing for the ISR missions that attritable drones will frequently perform. L3Harris's legacy EW expertise makes it a critical integrator for the cognitive spectrum operations that will be essential in any contested autonomy environment.
General Atomics: Production Capacity and the MQ-20
General Atomics — the manufacturer of the Predator, Reaper, and Avenger drone families — brings production infrastructure that newer entrants like Anduril cannot yet match at scale. GA-ASI received a prototype contract for the Air Force's CCA program (Collaborative Combat Aircraft) alongside Anduril, the only two companies to advance to the prototype build and flight testing phase. GA-ASI's MQ-20 Avenger is the most mature candidate for near-term attritable combat drone production at Replicator scale, with existing production lines and supply chains that can be scaled without building from scratch.
Replicator 2: Counter-Small UAS as the Primary Mission
If Replicator 1 was about fielding offensive and multi-mission autonomous systems, Replicator 2 — announced by then-Deputy Secretary of Defense Kathleen Hicks in August 2024 — addresses the defensive side of the autonomous weapons equation. Replicator 2 focuses specifically on counter-small unmanned aerial systems (C-sUAS), the capability to defend against the mass FPV and small-drone attacks that the Ukraine war demonstrated are now a primary battlefield threat.
The first Replicator 2 acquisition was announced on January 11, 2026. DoD designated the Under Secretary of Defense for Acquisition and Sustainment as the "lead organization to develop capabilities to C-sUAS," providing organizational clarity that the program had lacked in its early phases. The program draws directly on Ukraine war lessons — the recognition that cheap attacking drones require cheap defending drones, and that the economics of kinetic interceptors (missiles costing more than the targets they intercept) are unsustainable against mass UAS attacks.
Anduril's Roadrunner-M and Anvil systems, designed for exactly this mission, are primary candidates for Replicator 2 procurement. The Roadrunner-M is a recoverable, jet-powered interceptor drone that launches and lands vertically, can loiter waiting for threats, and carries a warhead for terminal intercept. The Anvil is a fixed-wing kinetic kill vehicle that uses AI-guided terminal homing to physically intercept and destroy threat drones through direct impact. Both systems are designed around the principle that defending against cheap drones requires systems that are themselves cheap — the cost-exchange problem that plagues Patriot missile defense when faced with $35,000 Shaheds.
The Collaborative Combat Aircraft: Replicator's Air Force Expression
Running in parallel to Replicator but conceptually aligned with it is the Air Force's Collaborative Combat Aircraft (CCA) program — the crewed-unmanned teaming initiative that will pair autonomous drone wingmen with F-35s, F-22s, and B-21s. The CCA is Replicator's most ambitious expression: not cheap expendable drones for saturation attacks, but sophisticated autonomous wingmen capable of complex air combat missions alongside crewed aircraft.
The CCA program awarded its Increment 1 prototype contracts to Anduril (building the YFQ-44A) and General Atomics (building a competing prototype). Northrop Grumman, Boeing, and Lockheed Martin were among the five companies initially considered but were eliminated before prototype funding. In December 2025, autonomy contracts for the CCA's AI pilot system were issued to RTX and Shield AI — confirming that Shield AI's HIVEMIND technology will be at the core of the Air Force's autonomous wingman capability.
The CCA represents a 2026 preview of what Replicator's doctrine of "all-domain attritable autonomy" looks like when translated into high-end air combat. Rather than replacing F-35 pilots, the CCA extends their reach: a single human pilot controls multiple AI wingmen, directing them into threats while remaining in a protected position. The AI executes the resulting combat actions autonomously, at machine speed, in a threat environment that no human pilot could navigate solo. It is the manned-unmanned teaming concept that China's GJ-11 program is simultaneously pursuing — the convergent evolution of the same doctrine driven by the same logic.
The FY2026 Budget: $13.4 Billion for AI and Autonomy
The FY2026 National Defense Authorization Act marks a doctrinal threshold in how the United States funds autonomous weapons development. For the first time, the DoD has included a dedicated, standalone budget line for autonomy and AI systems — totaling $13.4 billion. This is not the total AI and autonomy spend across the defense budget, which is spread across hundreds of program elements; it is a separately tracked, explicitly labeled autonomy budget that enables Congress to exercise meaningful oversight for the first time.
The Replicator program proper was allocated approximately $1 billion across FY2024–2025, and Replicator is not a single program of record with its own budget line. But the $13.4 billion headline figure reflects the much broader ecosystem of autonomous systems, AI decision support, machine learning integration in existing platforms, and next-generation program development that Replicator has helped catalyze throughout the DoD. The program's most significant effect may not be the specific systems it procures but the organizational and cultural change it represents: the explicit commitment by senior Pentagon leadership that autonomous systems are not a future-of-war curiosity but a present-day warfighting priority.
The honest answer, based on available open-source evidence, is: partially. Replicator 1 has clearly accelerated autonomous systems procurement across the services and has generated a productive industrial ecosystem of tech companies, contract mechanisms, and organizational structures that didn't exist in 2022. The January 2026 Replicator 2 acquisition represents institutional momentum. But the program has real weaknesses. The lack of public transparency makes accountability impossible. The "thousands fielded" claim has not been defined with sufficient specificity to evaluate against the Ukraine war standard of what those numbers mean in combat. The CCA program — Replicator's most ambitious expression — won't produce operational aircraft until 2027–2028 at the earliest. Against China's production timelines, the race remains genuinely uncertain.
The ACT Software Problem: Coordinating the Swarm
One of Replicator's least-discussed but most technically challenging components is the Autonomous Collaborative Technologies (ACT) software program — the coordination layer that allows thousands of individually capable autonomous platforms to function as a coherent swarm rather than an uncoordinated mass. Anduril, L3Harris, and Swarm Aero received ACT prototype contracts in November 2024 to address this challenge.
The core ACT problem is: how do you coordinate hundreds or thousands of autonomous systems in a denied communications environment, maintaining coherent tactical behavior without a central controller that becomes a single point of failure? The answer requires distributed AI algorithms — each platform running its own decision-making process, coordinating through local peer-to-peer communications rather than top-down command, and maintaining swarm-level intent through shared behavioral rules rather than centralized direction.
This is exactly the architecture that Chinese researchers have been developing in parallel, as described in the PLA National Defense University papers on autonomous urban warfare swarms. The convergent development — both China and the United States working toward the same distributed autonomous swarm architecture from different directions — reflects the fact that there is a single correct engineering answer to the problem of coordinating attritable autonomous systems in contested environments. The question is which side solves it first, at production scale, and deploys it with effective doctrine.
The Doctrine Gap: Technology Without a Manual
For all of Replicator's technical progress, the program's most significant remaining challenge is doctrinal. The U.S. military knows how to use autonomous systems for ISR, logistics, and simple strike missions. It does not yet have well-developed doctrine for autonomous swarm warfare — for how thousands of attritable drones are coordinated with crewed platforms in complex multi-domain operations, for the rules of engagement governing autonomous lethal decisions, for the command-and-control structure that exercises meaningful human oversight over AI-driven weapons without imposing latency that negates their tactical advantage.
Deputy Secretary Hicks' Replicator announcement included a commitment that all lethal decisions would remain under "appropriate levels of human judgment." But the specific mechanisms for ensuring that commitment at the scale and speed of autonomous swarm operations have not been publicly articulated. At some point in a swarm engagement, the decision to deploy lethal force against a target will be made not by a named human commander but by an AI system operating faster than human reaction time can track. Defining what "appropriate human judgment" means in that context is not a detail — it is the central doctrinal question of the autonomous weapons era.
Replicator in 2026: What Comes Next
The trajectory for Replicator in 2026 and beyond is reasonably clear from public evidence. Replicator 2 procurement will accelerate, with C-sUAS systems from Anduril and potentially other vendors entering acquisition. The CCA program will proceed toward flight testing and early operational capability. The ACT coordination software competition will narrow to a lead architecture. The FY2026 budget's dedicated autonomy funding will enable continued expansion across services.
The deeper question — whether Replicator represents a genuine strategic solution to the autonomous weapons gap with China — remains open. The program has demonstrated that the United States can mobilize a capable industrial ecosystem, issue contracts at meaningful speed, and field systems at a scale that was not possible through the traditional acquisition process. What it has not yet demonstrated is that the fielded systems represent the kind of combat-credible, producible-at-industrial-scale, tactically integrated autonomous capability that would actually deter or defeat a mass autonomous attack in the Taiwan Strait or elsewhere.
That demonstration will come — or not — in the next 24 months, as Replicator 2 systems move from procurement to training exercises, as CCA prototypes begin flight testing, and as the distributed swarm coordination software is tested in conditions that approximate real contested environments. Until then, Replicator remains a program with significant promise, significant momentum, and a significant accountability deficit that Congress, the public, and America's adversaries are watching closely.
- Replicator 1 declared "thousands fielded" by August 2025 milestone — specifics classified, independent verification impossible
- Replicator 2 first acquisition announced January 11, 2026 — C-sUAS designated as primary mission, directly drawing Ukraine war lessons
- FY2026 DoD AI/autonomy budget: $13.4B standalone line item — first time Congress can meaningfully track this spending
- Anduril holds a potential 10-year, $642.2M Marine Corps counter-drone contract (March 2025); Shield AI's Hivemind was reportedly chosen to fly the YFQ-44A CCA
- CCA (Collaborative Combat Aircraft) program: only Anduril and GA-ASI advanced to prototypes; Shield AI and RTX handle AI pilot autonomy
- ACT swarm coordination software — the hardest problem in Replicator — now in prototype competition between Anduril, L3Harris, Swarm Aero
- Doctrine for human-autonomous lethal decision-making at swarm scale remains the unresolved critical challenge