The US Navy’s drive to field swarms of autonomous surface vessels has suffered a series of mishaps, including two test accidents off the California coast last month that left one drone boat crippled and another vaulting over its bow. A separate trial weeks earlier capsized a support craft and tossed its captain overboard. Investigators traced the incidents to software failures and human error during interactions between on-board systems and external autonomy code. Following the crashes, the Pentagon’s Defense Innovation Unit indefinitely halted a roughly $20 million contract with L3Harris Technologies, one of the suppliers of the control software. The turbulence compounds earlier turmoil inside the Navy’s procurement ranks: Rear Admiral Kevin Smith, who ran the Program Executive Office for Unmanned and Small Combatants, was fired in June, and the unit itself is now under review for possible restructuring. The setbacks come despite rising political and financial backing for unmanned fleets intended to complicate any Chinese move against Taiwan. The 2023 Replicator initiative earmarks $1 billion to buy thousands of aerial and maritime drones, while last month’s “Big Beautiful Bill” signed by President Donald Trump added almost $5 billion for maritime autonomy systems. BlackSea Technologies has already booked at least $160 million in Navy orders and is producing dozens of Global Autonomous Reconnaissance Craft each month; rival Saronic, valued at roughly $4 billion, has secured about $20 million in prototype agreements. Analysts say the Navy still faces a steep learning curve in integrating commercial-grade software with military requirements and in rewiring an acquisition culture built around large, crewed ships. Until the service can prove the reliability and cost-effectiveness of autonomous swarms, the promise of a rapid, low-cost countermeasure to China’s growing naval power will remain out of reach.