Artist's rendering of a drone swarm. Photo: US Army
Artist’s rendering of a drone swarm. Photo: US Army

The Pentagon’s plan to field thousands of AI-powered drones under its Replicator initiative is facing setbacks, with technical flaws, costly systems, and bureaucratic hurdles slowing progress.

Launched in 2023 by then–Deputy Defense Secretary Kathleen Hicks, the Replicator program aimed to fast-track the deployment of “small, smart, cheap” unmanned systems for air, land, and sea domains by August 2025 in preparation for a potential conflict in the Indo-Pacific.

Two years in, despite $1 billion in funding and political backing, sources told the Wall Street Journal that the effort has stalled. Some blame the military for pushing unready systems, while others pointed out that setbacks are typical of ambitious programs.

Oversight of Replicator shifted in August from the Defense Innovation Unit (DIU) to US Special Operations Command’s Defense Autonomous Warfare Group (DAWG), led by Lt. Gen. Frank Donovan, to accelerate progress.

Testing Challenges

Recent demos revealed flaws that delayed the launch of systems.

A BlackSea Technologies unmanned boat suffered a rudder failure and floated away, while an Anduril Industries aerial drone showed issues with the launch tubes. Software used in the systems also misidentified objects.

Anduril founder Palmer Luckey defended the effort, noting Replicator had “very, very good things.”

“Could it have been done better? Could it have been more clear about what exactly they were doing? Yes, of course. But big picture, I don’t think it was that bad,” Luckey said.

Bureaucratic Hurdles

People familiar with the matter said uniformed officers with limited technical expertise influenced bulk purchases, leading to the selection of platforms that still required extensive work to operate autonomously.

Additionally, they highlighted that the Replicator’s former overseer, the DIU, bought unfinished concepts and older technology without rigorous testing.

Questionable Acquisitions

The agency purchased hundreds of BlackSea’s Global Autonomous Reconnaissance Craft (GARC) for the Replicator, but they were mainly built for littoral environments rather than complex and long-range Pacific missions.

In addition, the military reportedly ordered frequent software changes and other additions that led to inflated costs and further delayed progress.

AeroVironment’s Switchblade 600 loitering munitions — costing around $100,000 per unit — were also bought in bulk, despite reports of underperformance in Ukraine and vulnerability to electronic warfare. 

While the company said it has upgraded the drones to neutralize major air defense systems, the US Army allegedly did not acquire the enhanced models over concerns of further delays.

Integration challenges also surfaced during a 2024 Pacific drill, where drones from different vendors linked through Anduril’s software were said to have struggled at times to coordinate and complete tasks once out of the operator’s sight.

Urgency in the Indo-Pacific

Despite the setbacks, some officials argue that Replicator still helped acquire and advance new uncrewed systems and autonomous tech in a shorter timespan. 

With China’s potential invasion of Taiwan by 2027, the Pentagon views DAWG’s sped-up task as critical to fielding autonomous drones that can cross vast distances, operate without GPS or radio links, overwhelm defenses at scale, and strike without heavy costs in lives or hardware.

You May Also Like

South Korea Bets on MakinaRocks to Drive Military AI

MakinaRocks will adapt its Runway AI platform for military use, providing autonomous model management capabilities.

UK Military Conducts ‘Largest Ever’ Artificial Intelligence Trial

The UK military has completed its largest artificial intelligence (AI) experiment to…

US Air Force Eyes Smarter Target Tracking With $99M AI Project

US Air Force’s $99M ATA-AI project aims to use AI and machine learning to track stealth aircraft, hypersonic weapons, and drone swarms with high-speed, multi-source precision.

Striveworks Scales US Army NGC2 AI Work Across Two Divisions

The service is scaling AI-enabled command-and-control capabilities to improve battlefield decision-making and intelligence fusion.