U.S. Army Awards $20B Anduril to Deploy Lattice AI Open Architecture for Battlefield Integration.

U.S. Army awarded Anduril Industries a firm-fixed-price enterprise contract with a cumulative ceiling of $20 billion to consolidate current and future commercial solutions, including the company’s proprietary, open-architecture, AI-enabled Lattice suite, integrated hardware, data, computer infrastructure, and technical support into a single mission-ready capability.

Structured as a 10-year enterprise agreement, the contract consolidates more than 120 separate procurement pathways into a single framework for software, hardware, data infrastructure, and support. Centered on Anduril’s Lattice platform, it enables scalable integration of sensors, AI-driven command-and-control, and selected effectors, with early use focused on counter-UAS operations and joint interoperability.

Read also: U.S. Army picks Anduril Lattice for Integrated Battle Command System Maneuver counter-drone role.

The Army's contract with Anduril uses Lattice as an open architecture backbone that connects sensors, computing, AI-enabled command-and-control, and effectors into a single mission-ready capability (Picture source: Anduril).

The Army said the new framework replaces more than 120 separate procurement actions previously used for Anduril products, with work locations and funding determined by individual orders and ordering authority extending through a five-year base period plus a five-year optional ordering period to March 2036. Army officials also described the agreement as a foundational command-and-control enabler for counter-UAS interoperability, supported by existing data integration with hundreds of Joint and Army systems.

Anduril’s Lattice Command and Control integrates and manages thousands of sensors, effectors, networks, and mission systems. Lattice Mesh is a decentralized network connecting Anduril and third-party systems. Lattice serves as a digital battlespace layer—combining data, correlating tracks, pushing alerts, and providing a common operational picture for faster, informed decisions.

That is also where “open architecture” matters. In U.S. defense acquisition, open architecture does not mean open-source code; it means modular interfaces, accessible data layers, and a system design that allows new components to be integrated or swapped without rebuilding the entire stack. The Department of Defense’s Software Modernization Strategy explicitly calls for modular open systems approaches and more agile management of commercial off-the-shelf software, while the Army’s 2024 software reform agenda argues that software now underpins mission-critical combat and business systems and must be developed through iterative, flexible, data-centric methods rather than rigid, hardware-era procurement logic. The Army needs that approach because the threat evolves faster than traditional acquisition cycles, and a closed system locks modernization speed to a single vendor’s timeline.

The contract language is therefore notable because it explicitly includes not only software but also integrated hardware, data, and computer infrastructure. That points toward Anduril’s deployable compute and command-and-control family, not just Lattice as an interface. Menace-T is described by Anduril as a human-portable C2 solution that combines Lattice with rugged, modular compute for tactical operations, while Menace-X is positioned as an expeditionary command, control, communications, and computing capability built to function in denied, disrupted, intermittent, and limited communications environments. For the Army, that matters because AI-enabled battle management is only as useful as the edge compute and communications resilience supporting it at the brigade, battalion, and company levels.

Viewed through that lens, Anduril’s portfolio begins to show what the Army may actually be buying access to. On the reconnaissance side, Ghost-X is especially relevant because Anduril said it was selected for the U.S. Army’s company-level sUAS directed requirement; the aircraft is described as a tandem-rotor military reconnaissance drone with about 80 minutes of cruise endurance and a range of roughly 15.5 miles, or 25 kilometers. On the sensor side, Spark is marketed as a dual-band hemispheric radar for detecting both fast, low vehicle threats and slow, high air threats while on the move, and Spyglass is an algorithmically enhanced Ku-band phased-array radar delivering high-quality target tracking for integrated weapons systems. Put together, these systems represent the sensing layer that a Lattice-centered architecture could fuse for maneuver units and counter-UAS formations.

The possible armaments and hard-kill effectors matter just as much, because an open architecture only becomes operationally decisive when it links sensing to action. Bolt-M is one of the clearest examples: Anduril says the base Bolt autonomous air vehicle offers more than 20 kilometers of range and more than 45 minutes of endurance for ISR, while the munition variant, Bolt-M, can carry up to three pounds of munition payload against static or moving ground targets. That gives small units a software-connected precision strike option that fits the Army’s demand for distributed lethality. In air defense, Anvil is described as an autonomous kinetic interceptor for low-altitude drone defense, while Anvil-M is the munition variant intended to increase effectiveness against larger and faster threats and to defeat Group 1 and Group 2 unmanned aircraft. At a higher-performance tier, Roadrunner is presented as a reusable VTOL autonomous air vehicle with twin turbojet engines and modular payload capacity, while Roadrunner-M is designed to address threats that cut across legacy air-defense echelons. These are not confirmed line items under the enterprise contract, but they are precisely the type of effectors that make the contract strategically interesting: Lattice can potentially sit above a family of shooters rather than a single proprietary weapon.

If Ghost-X feeds reconnaissance data into Lattice, Spark and Spyglass contribute tracking quality, and Anvil, Roadrunner, or Bolt-M provide response options, the real military advantage is a compressed sensor-to-decision-to-effector chain. That reading is reinforced by the Army's language describing the deal as a common framework for counter-UAS interoperability and by Anduril’s disclosure that JIATF-401 selected Lattice as its enterprise tactical C2 platform for counter-UAS under what the company says is the first $87 million task order issued beneath the new agreement. The early signal, in other words, points toward battle management and air-defense integration as the leading edge of this enterprise vehicle.

Just as important, the contract should not be misread as an immediate $20 billion purchase or a blanket award of every future Army digital program to Anduril. The Army states that the figure is a maximum potential value rather than an obligated amount, that funding and work locations will be set by each order, and that the enterprise agreement will not replace competition for future programs. What it does do is create pre-negotiated pricing, reduce pass-through charges, shorten contracting timelines, and give the government a standard way to procure a growing software-hardware ecosystem without renegotiating the relationship every time a new sensor, compute package, or support module is needed.

For the U.S. Army, that is the deeper meaning of the Anduril award. The service is moving away from treating software as a support function attached to hardware and toward treating software, data, and edge computing as the connective tissue of combat power. The question is no longer whether the Army needs open architecture; its own software strategy already answers that. The real question is whether it can field such architectures fast enough, securely enough, and broadly enough to out-adapt drone saturation, electronic disruption, and machine-speed targeting in the next conflict. This Anduril enterprise contract is one of the clearest signs yet that the Army now sees race as software-defined.