There is a number that has been reverberating around NATO defence establishments since late June 2026, and it is not a budget figure or a vehicle count. It is a ratio: 72 to 1.
Speaking at the Royal United Services Institute (RUSI) Land Warfare Conference in London on June 23, General Sir Roly Walker, Chief of the General Staff and the professional head of the British Army, told the assembled senior military officers, policymakers, and industry leaders that the British Army’s new AI-enabled planning system had done something he had not expected. He had stood at the same conference a year earlier and told the audience that the system would help the Army sense twice as far, decide twice as quick, and strike twice as deep. He was wrong, and he said so.
“A corps planning cycle that once took 72 hours can now take one,” Walker said from the podium. “What they’re going to do with the other 71 hours, I do not know.”
That sentence, delivered with the dry understatement of a British general who has clearly thought carefully about what he is about to say publicly, carried implications that extended well beyond the RUSI conference room. It is the first public, named, verified claim from a serving chief of a major NATO army that AI has fundamentally compressed the basic rhythm of large-scale military planning, not in a laboratory, not in a simulation, but in a system that has been exercised across national borders and is now deployed at corps headquarters level.
What ASGARD Is, and What It Was Supposed to Be
Project ASGARD, the acronym stands for AI-enabled targeting, sensing, and battle management, was described by Walker as a “digital targeting web”: a connected software architecture designed to gather and process battlefield data, help commanders identify targets, coordinate attacks, and make faster decisions across a complex operational environment.
When the British Army first unveiled the project publicly in mid-2025 and deployed a prototype to Estonia for Exercise Hedgehog, the stated ambition was meaningful but measured: sense twice as far, decide twice as fast, strike twice as deep. Those were the metrics Walker publicly committed to. They were, by any historical standard, ambitious. Corps-level planning, the large-scale coordination of tens of thousands of troops, multiple weapons systems, intelligence assets, logistics chains, and allied forces, is the part of military operations that has always moved at the speed of bureaucracy rather than the speed of battle. Three days to complete a planning cycle is fast by historical standards.
ASGARD has apparently made that look slow.
The system is now deployed at corps headquarters level within the Allied Rapid Reaction Corps (ARRC), NATO’s premier deployable corps headquarters, headed by the British Army and staffed by personnel from 21 NATO member states. According to Walker’s speech, as confirmed by the official British Army transcript published on army.mod.uk on June 23, ASGARD is not just shortening planning cycles. A corps that previously prosecuted 24 targets in a day, because that was the pace at which the planning and coordination work could be completed, can now engage 10 times that number. “In fact, they’re limited only by the munitions that are available to fire into the sky,” Walker said.
That is approximately 240 targets per day, typically within 200 kilometres of the frontline, driven through a planning and targeting cycle that now takes an hour instead of three days. And the system, Walker told the conference, “is literally a digital juggernaut that is evolving every eight to twelve weeks.”
The Technology Stack: How ASGARD Actually Works
To appreciate what Walker’s numbers mean, it helps to understand what ASGARD is actually doing under the hood, because it is not a single product from a single vendor. It is an integrated architecture of existing and new systems, connected through a common data layer, built over roughly two years with contributions from 27 industry partners.
The backbone of the network is Anduril UK’s Lattice, a mesh command and control software platform designed specifically to connect sensors, operators, and weapons across a distributed, resilient network. Lattice’s architecture is software-defined, meaning it can integrate with both legacy military systems and newer platforms without requiring those systems to be rebuilt from scratch. On ASGARD, Lattice handles the transit of information from sensor to shooter, the movement of targeting data from wherever it is collected to wherever a decision is made and a weapon is assigned.
Feeding data into that network are sensors running Helsing’s Altra software stack, an edge AI targeting system that runs on the sensors themselves, rather than requiring data to be sent back to a central server for processing. Altra analyses full-motion video from surveillance drones, identifies and classifies targets using computer vision, and generates what the military calls a Cursor-on-Target (CoT) message, a standardised, open-source geospatial data format that can be shared between any system on the network regardless of manufacturer or software environment. Think of it as a universal translation layer: Altra speaks whatever language the target information is in and converts it into a format every other system on the network can read and act on.
The CoT data passes through Android Tactical Awareness Kit (ATAK), a geospatial mapping application running on Android smartphones carried by soldiers, into Lattice, and from there into Systematic’s SitaWare command and control suite, which provides the shared operational picture that commanders use to maintain situational awareness and make targeting decisions. From SitaWare, targets flow into PRIISM, a digital version of the Joint Air-Ground Integration Centre (JAGIC) developed by Research Innovations, which applies additional processing including legal review of the target, collateral damage estimates, and weapon-to-target matching. The firing data then passes to CGI’s FC BISA (Fire Control Battlefield Information System Application) for artillery integration, or to Helsing’s Altra Strike targeting element, which can cue the Helsing HX-2 AI-enabled strike drone to execute the attack autonomously.
Communications across the network are maintained using BlackTree Technologies’ TrellisWare radios, which use a mobile ad hoc networking waveform designed to maintain connectivity in contested electromagnetic environments where conventional radio links might be disrupted. At the divisional level and above, Anduril’s Menace-T edge compute and communications module relays data from the tactical edge to headquarters in seconds, even in environments where conventional satellite uplink is unreliable. During Exercise Griffin Lightning, a NATO deployment that preceded Arrcade Strike, Menace-T allowed commanders to rapidly task battlefield assets including one-way attack drones while accelerating the engagement timeline in near-real-time.
The overall result of all these components working together is exactly what Walker described: not a faster version of the old planning process, but a fundamentally different process, in which a large portion of the analytical and coordination work that previously consumed days of staff officer time is handled automatically by the network, surfacing only the decisions that genuinely require human judgment for a human commander to make.
The Exercise Under London: Ten Terabytes a Day from Charing Cross
Walker’s most vivid illustration of where ASGARD currently stands operationally came from an exercise that took place the month before his RUSI speech, and the location could hardly have been more deliberately chosen for its symbolic weight.
Exercise Arrcade Strike, conducted in May 2026, deployed the ARRC’s headquarters not to a military base or a field position, but to the disused platforms beneath Charing Cross Underground Station, a tube stop on the Northern and Bakerloo lines, a stone’s throw from the Ministry of Defence’s main building in Whitehall and directly beneath Trafalgar Square. The choice was deliberate: operating underground reduces the headquarters’ electromagnetic and physical signature, making it harder to detect and therefore harder to strike. In an era when adversary long-range precision fires and loitering munitions have made static headquarters extremely vulnerable, the ability to command from an urban underground location is a genuine operational advantage.
From those platforms, staffed by ARRC personnel from 21 nations, the headquarters was processing 10 terabytes of data per day, which the British Army described as the equivalent of “nearly three months of non-stop high-definition Netflix.” It was simultaneously managing the exercise scenario, which involved real British troops deployed in Estonia as part of NATO’s enhanced Forward Presence battlegroup, the standing defensive deployment on the Alliance’s eastern flank.
That combination, a concealed headquarters processing sensor, communications, and targeting data at enormous volume in real time, managing units operating hundreds of miles away, is what ASGARD in its current form actually looks like in practice. And it is not a concept or a prototype anymore. Walker said explicitly that this system now sits at the heart of UK Forward Land Forces in Estonia: “Today, the UK possesses a similar Recce-Strike system to the one used by Ukraine to maul Russian forces in the Donbas,” a senior ARRC commander noted during the exercise.
Ukraine, Delta, Maven, and the Global AI Planning Race
ASGARD does not exist in isolation. It is part of a broader and accelerating global competition to field AI-enabled command and targeting architectures, one that we have been tracking extensively at Future Military Technologies.
Ukraine’s Delta platform has been the most battle-tested system of this type in the world. Developed under wartime conditions since 2022, Delta integrates intelligence feeds from drones, satellites, open-source imagery, and human reports into a shared digital operational picture accessible to Ukrainian commanders at all echelons. It has been credited with enabling the rapid target development and engagement cycles that gave Ukrainian forces their notable efficiency advantage over Russian forces in several phases of the conflict, particularly the destruction of Russian armoured columns in the opening weeks of the war and the sustained targeting campaigns in the Donbas. ASGARD was explicitly designed to achieve a comparable capability for the British Army, and Walker has confirmed that the UK has been deploying 50 operational-level electronic warfare systems sourced directly from Ukraine’s experience, while sending thousands of drones to its own units.
Palantir’s Maven Smart System serves an analogous function for the U.S. military, drawing on the original Project Maven AI initiative, launched by the Pentagon in 2017 to accelerate analysis of drone surveillance footage, and expanding it into a broader AI-enabled targeting and intelligence fusion platform. As we detailed in our coverage of Project Convergence Capstone 5, the U.S. Army is pursuing the same fundamental objective as ASGARD through its CJADC2 (Combined Joint All-Domain Command and Control) architecture, a common operating picture that fuses sensor data from all domains and all partner nations into a single, AI-curated battlefield view from which commanders can task any available sensor or weapon. The US Army’s Mission Command System / Common Operational Picture programme, and the edge data mesh demonstrated at Capstone 5, are American iterations of the same underlying concept.
What Walker’s June 23 statement did was provide the first public, quantified proof-of-concept result for this class of system from a named commander at a named event. Numbers like “10 times as many targets” and “72 hours to one” are the kind of operational benchmarks that defence ministries use when writing capability requirements. They change the conversation from “should we invest in AI battle management?” to “why has our corps headquarters not yet achieved comparable results?”
The procurement implications have not been lost on the industry. Anduril, whose Lattice software is the connectivity backbone of ASGARD, raised $5 billion in May 2026 at a valuation of approximately $61 billion, roughly double its valuation from a year earlier. The U.S. Army awarded Anduril an enterprise contract worth up to $20 billion over ten years in March 2026, consolidating more than 120 existing procurement actions. Helsing, the European AI defence company whose Altra stack provides the targeting intelligence that feeds ASGARD, was in the process of raising $1.2 billion at an $18 billion valuation as of mid-2026, a figure that reflects investor confidence in the commercial trajectory of AI-enabled military systems.
The British Army’s Lethality Roadmap: 20/40/40
ASGARD’s development did not happen in isolation from wider British Army thinking. Walker set the direction at the 2024 RUSI Land Warfare Conference when he committed to doubling British Army lethality by 2027 and tripling it by 2030, a target he reaffirmed at the June 2025 conference and is now reporting progress against.
The force structure concept underpinning that ambition is known as the 20/40/40 model: 20 percent of the force built around traditional platforms, tanks, armoured vehicles, artillery, 40 percent composed of expendable autonomous systems such as loitering munitions and one-way attack drones, and 40 percent consisting of reusable AI-enabled assets such as ISR drones, autonomous logistics vehicles, and sensor platforms that can be recovered and reused between missions. ASGARD is the digital layer that connects and orchestrates all three categories.
The rationale behind this force mix maps directly to the operational lessons of Ukraine. A force that is 100 percent composed of expensive, crewed platforms is both economically unsustainable in high-tempo combat and operationally fragile against mass drone and precision missile attack. A force that blends crewed and uncrewed systems, with autonomous expendable munitions available at scale and the AI architecture to task them faster than a human planning process can manage, can absorb losses in the expendable tier while maintaining its overall combat effectiveness. The FPV drone revolution we covered in our deep-dive on the Ukrainian battlefield validates the expendable tier; ASGARD is the brain that makes the whole architecture function at corps scale.
The integration of the DART 250, a British-manufactured jet-powered loitering munition with a range of 250 kilometres, a speed exceeding 400 km/h, GPS-jam resistance, and a seeker that can home in on jamming sources, into the ASGARD network during the Estonia exercise points toward the operational architecture Walker is building. A corps that can detect a target, develop the intelligence, complete legal review, match it to an available weapon, and deliver a strike in under an hour, using an expendable one-way munition tasked through an AI-curated kill chain, is a qualitatively different instrument from the three-day planning cycle it replaced.
The Human Question: What Do You Do With 71 Hours?
Walker’s question, “what they’re going to do with the other 71 hours, I do not know”, was delivered as a wry admission that ASGARD has exceeded its own design goals. But it is also one of the most important questions in the current debate about AI and military decision-making, and it deserves a serious answer rather than a knowing laugh.
The ASGARD system, as it currently operates, keeps humans in the loop at the point of targeting decisions. The AI does not decide to strike a target; it identifies, classifies, geo-locates, and proposes targets, passing them through a sequence of validation steps, including automated legal review and collateral damage estimation within PRIISM, before presenting them to a human commander who approves or declines the engagement. The system handles the volume and speed of the analytical work. The human retains the authority over the lethal decision.
That is the architecture as it exists. The pressure to evolve it will be real. If a corps can prosecute 240 targets a day rather than 24, but is still constrained by the rate at which human commanders can process engagement approvals, each one requiring a review of the intelligence, the legal assessment, and the available weapons, then the 71 reclaimed hours may fill up faster than Walker’s joke implies. As we noted in our coverage of USSOCOM’s AI investment strategy, the distinction between humans “in the loop” (approving each engagement) and humans “on the loop” (monitoring and able to override, but not individually approving) is the central policy question of AI-enabled military targeting, and it is not hypothetical. It is already being navigated in practice by every military investing in these systems.
The ethical and legal community has not been slow to notice. Campaign groups including Drone Wars UK and the UK Stop Killer Robots coalition have raised concerns about the pace at which ASGARD’s AI-curated targeting is being integrated into operational British Army practice, arguing that the speed of the system, and the volume of targets it can generate, risk compressing human deliberation time in ways that the formal “human in the loop” framing does not fully address. These are not fringe concerns; they sit at the heart of how international humanitarian law should apply to AI-enabled targeting systems, and they will shape the policy framework within which ASGARD and its successors operate.
Walker addressed the scope of the system’s current development directly in his official speech, noting that ASGARD “is not just a digitised sensor-to-shooter system” but is instead functioning as “the foundation of an agentic AI headquarters, where humans increasingly get to say yes or no”, a formulation that acknowledges the direction of travel without eliding the questions it raises.
What the Rest of NATO Is Watching
For allied militaries watching Walker’s RUSI address, the immediate takeaway is competitive. A British corps operating with ASGARD can now plan and execute targeting cycles at a pace that a conventionally equipped corps cannot match. In a NATO coalition operation, that creates asymmetry within the alliance, and a strong incentive for other members to achieve comparable capability.
The tactical communications infrastructure that ASGARD depends on points toward the investment direction. The Nokia and Motorola next-generation tactical communications network being built for UK forces, which we covered earlier this year, provides the secure, high-bandwidth link between headquarters and tactical units that ASGARD’s data volumes require. Germany is pursuing its own AI-enabled command and networking investment, as we reported in our coverage of Nokia and blackned’s deployable tactical network for the Bundeswehr. France, Poland, and the Baltic states are all at various stages of similar programmes. What ASGARD does is provide the first publicly documented performance benchmark for this class of capability, a real number, from a real exercise, reported by a named general at a named conference, that every defence ministry in the alliance can now measure itself against.
Walker closed the loop explicitly in his RUSI speech. Preparing against Russia means building a British corps “capable of doing what a Ukrainian corps can do today.” And in the next year, he said, “I expect to see much greater numbers of our remote and autonomous systems forward on our eastern flank, ready to strike and act within 30 minutes.”
That 30-minute readiness window is only achievable if the planning and targeting infrastructure has already done its work. ASGARD, processing 10 terabytes a day from a London tube station while managing troops in Estonia, is the system designed to make it possible.
AI planning revolution
Walker’s “tank versus horse” analogy, offered at the RUSI conference as his framing for the significance of the AI planning revolution, was precise in a way that the first people to see tanks might not have fully appreciated. The horse was not simply a slower form of transport, it was the entire conceptual infrastructure of cavalry warfare, logistics, and operational tempo. The tank did not just replace the horse; it made obsolete the doctrine built around the horse’s limitations.
ASGARD’s 72-to-1 ratio is a similar kind of disruption. It does not just make the old planning process faster; it makes a planning process built around three-day cycles conceptually obsolete. A corps that can re-plan in an hour can respond to a fast-moving battlefield in ways that a corps requiring three days simply cannot. It can adapt to breakthroughs and reverses before they become decisive. It can strike targets of opportunity that would have expired before the old planning cycle completed. And it can do all of this from a concealed headquarters under a tube station, running on a network built from commercial software components that updates every eight to twelve weeks.
For the adversary planning to fight that corps, the challenge is not just the new weapons or the new drones. It is the speed of the mind behind them.
