Persistent Force: DARPA's Containerised Swarm Programme and the Commands Being Built to Deploy It.

RFI DARPA-SN-26-33, published by DARPA's Tactical Technology Office in late April 2026 with a 15 May response deadline, is the agency's most explicit public articulation of the next-generation small UAS architecture. The programme seeks concepts for Group 1-3 unmanned aircraft (platforms weighing under 55 pounds) organised into autonomous constellations of up to 500 aircraft. Those aircraft would be housed in and operated from standardised containers compatible with existing military logistics infrastructure: Conex containers, 463L pallets, Tricon modules, and ISU containers. The intent is a system that can be moved anywhere military logistics can reach, emplaced, and activated without fixed infrastructure or large ground support crews.

The autonomy requirement defines the technical challenge. DARPA is examining what it characterises as Autonomy Level 4 operations: the human operator defines the mission objective, and the system handles execution (launch sequencing, airspace deconfliction, multi-agent task allocation, recovery, post-flight diagnostics, recharge or refuel cycles, and relaunch) without step-by-step human intervention. In a contested electromagnetic environment, the drones must sustain operations through GPS jamming and electronic warfare, requiring GPS-denied operating modes, low-probability-of-intercept data links, spectrum-agile communications, and onboard decision logic capable of maintaining mission continuity during degraded connectivity. Multi-agent autonomy features specified in the RFI include formation management, collision deconfliction, path optimisation, dynamic task allocation, and edge-based computing.

The conceptual shift embedded in the RFI is from the drone as a discrete asset to the containerised constellation as a unit of force. A conventional UAS programme procures aircraft. DARPA-SN-26-33 envisions procuring a self-regenerating weapons network that can be dispersed, concealed in unassuming logistics containers, activated on command, and sustained in theatre without the logistical signature that traditional aviation support requires. The strategic logic maps directly onto lessons from the Russia-Ukraine conflict, where dispersed, containerised FPV and strike drone operations have complicated adversary targeting while concealing launch infrastructure.

DARPA's technical programme is being matched by an institutional response. On 22 April 2026, U.S. Southern Command announced the establishment of the SOUTHCOM Autonomous Warfare Command (SAWC), a new entity under Gen. Francis Donovan that will deploy aerial, surface, and underwater drones to counter cartel threats and respond to crises across the Caribbean, Central America, and South America. SOUTHCOM characterised the SAWC primarily as a counter-narcotics and crisis-response tool, but its organisational significance extends further: it is the first geographic combatant command-level autonomous warfare unit in US military history, and its design will serve as the template for equivalent structures across other commands.

On 29 April 2026, Secretary of Defense Pete Hegseth told the House Armed Services Committee that the Pentagon would 'shortly announce a sub-unified command of autonomous warfare.' The announcement was brief and the detail sparse: the Pentagon declined to clarify whether SAWC is the entity Hegseth referenced or whether a separate joint command is planned. The distinction is consequential. A sub-unified command sits below a combatant command but above a joint task force, carrying cross-service authority over budget, personnel, and operations, a fundamentally different instrument from a combatant command-subordinate unit such as SAWC. If the sub-unified command is distinct from SAWC, it represents a new tier of autonomous warfare institutional architecture with authority to coordinate Defence Autonomous Warfare Group spending across all services.

The counter-UAS side of the institutional picture reveals the same pattern of compressed timelines. Joint Interagency Task Force 401, the cross-service counter-UAS body, is seeking $580.3 million in research, development, test, and evaluation funds in the FY27 budget, up from $6.5 million in FY26, a roughly 89-fold increase in a single budget cycle. The asymmetric funding levels between DAWG ($54.6 billion) and JIATF 401 ($580.3 million) indicate that the institutional balance currently prioritises offensive autonomous force generation over defensive capability, though both lines are growing from near-zero baselines and at rates that have no precedent in modern defence budgeting.

The feasibility of what DARPA-SN-26-33 envisions depends on whether the autonomy software stack can deliver reliable Autonomy Level 4 performance at constellation scale in a contested environment. The most direct recent test of that capability was the 26 February 2026 flight of Anduril's YFQ-44A Collaborative Combat Aircraft, in which the aircraft completed mission objectives under Shield AI's Hivemind autonomy stack before switching mid-flight to Anduril's Lattice system and repeating the same objectives before returning to base. The test, conducted over the Mojave Desert, demonstrated that mission-autonomy software can be interchanged mid-sortie on a production-grade airframe, a capability the Air Force has identified as central to the CCA programme's resilience against software single points of failure. The YFQ-44A entered serial production at Anduril's Arsenal-1 facility in Ohio in March 2026.

Shield AI's March 2026 Series G financing provides the capital market's assessment of autonomy stack maturity. The company raised $2 billion at a $12.7 billion post-money valuation (a 140 percent increase from its prior financing) in a round led by Advent International and co-led by JPMorgan Chase's Security and Resiliency Initiative, with an additional $500 million in preferred equity from Blackstone. Shield AI is projecting annual revenue exceeding $540 million in 2026, representing approximately 80 percent year-on-year growth. The simultaneous acquisition of Aechelon Technology, a simulation and synthetic-environment company funded from the Series G proceeds, addresses a key gap in high-autonomy development: the ability to generate synthetic training data for contested-environment and GPS-denied scenarios at scale without live-flight hours.

The doctrine gap, however, is the most underappreciated risk in the autonomous warfare buildout. A 500-drone constellation operating at Autonomy Level 4 is, in the absence of clear rules of engagement, targeting authorities, and cross-service communication protocols, a strategic liability as much as an asset. Institutional experience in Ukraine has demonstrated that drone operations without integrated command doctrine generate as many tactical liabilities as opportunities: fratricide, unintended escalation, and loss of situational awareness are recurring consequences of autonomous systems deployed ahead of the doctrine designed to govern them.