The next competitive advantage in military autonomy lies in bringing platforms, software, communications and support into service together. That changes how defence companies earn revenue, how customers keep sovereignty and how industrial capacity becomes usable military power.
Andy Demir · 16-minute read · Updated 12 October 2026
Executive Summary
In the first week of October, the US Army moved the data layer of its Next-Generation Command and Control programme from division-scale prototyping towards corps-scale fielding, with a $162.8 million base period inside a ceiling of up to $1.8 billion, while NATO reported autonomy experiments at REPMUS that connected nearly 300 capabilities under human control. Both show where commercial value in military autonomy is won or lost: in the gap between a successful trial, a framework ceiling and a supported capability in service.
My July argument that autonomy is an architecture problem holds. The sharper conclusion is that competitive power sits with whoever controls the deployment cycle: integration, acceptance, approved change and continued availability. Over the next twelve months, I expect mission-specific integration to advance faster than any universal allied architecture; the evidence to watch for is funded fielding, reusable acceptance evidence, and support contracts that specify who pays for change. For the board of a supplier whose equipment must plug into someone else's architecture, the decision worth taking this autumn is to run one live opportunity through a deployment-stage review linking engineering, contract terms and cash, and to stage capital against acceptance milestones rather than framework ceilings.
In This Essay
- October brings a more concrete test of the architecture thesis
- Growing budgets still have to become usable capability
- The deployment cycle determines where value can accumulate
- Europe's opportunity runs through national integration
- Military aviation connects Australia, Europe and the United States
- Recurring revenue depends on the right to change the system
- The strongest counterargument deserves a place in the investment case
- The next twelve months will test conversion into service
- A portfolio discussion around deployment and cash
- Questions for the C-Suite and the Board
- References
October brings a more concrete test of the architecture thesis
On 5 October 2026, Anduril announced that the US Army had awarded it a five-year contract to field the common data layer of Next-Generation Command and Control (NGC2), beginning with I Corps, the corps oriented to the Pacific. The base period is worth $162.8 million, and options could take the total to $1.8 billion. The company described the step as a move from division-scale prototyping to corps-scale fielding, and it added a sentence every supplier in this market should read twice: no single company, it said, should control every element of the Army's command-and-control stack. These are company-reported terms, and the distance between the base period and the ceiling matters (Lawrence, 2026).
Two weeks earlier, NATO's Allied Command Transformation reported on Task Force X-Arctic at REPMUS 2026, the uncrewed maritime exercise off Tróia, Portugal, that ended on 25 September. The task force connected nearly 300 capabilities from nations, industry, and academia to a shared data layer, and its mission engineering layer automatically tasked and re-tasked them based on the commander's intent, with the assets under ultimate human control. NATO's 8 October summary placed this ahead of a trial in Norway at the end of the year. Together, the two announcements show a shift from demonstrating individual vehicles to testing how mixed assets from different owners operate as a single force (NATO Allied Command Transformation, 2026; NATO, 2026b).
When I wrote about this subject in July, my central argument was that autonomous warfare had become a problem of system architecture. I hold to that, and I would now state the commercial conclusion more precisely. An architecture creates value when equipment can enter service, accept changes, and remain usable under the conditions the customer actually faces. The industrial competition therefore runs through the whole deployment cycle, including the less visible work of acceptance, support and configuration control, and the company that makes that cycle repeatable will usually keep more of the value than the company with the best demonstration.
Autonomy also needs a careful definition, because an uncrewed vehicle can still depend on continuous remote control. Navigation autonomy, automated mission planning, coordinated tasking, and authority over the use of force are separate functions, each with its own approval path. In the Arctic experiment, a human set the intent and reviewed the system's recommendations throughout, which serves as a useful reminder that greater automation does not equate to unrestricted decision authority (NATO Allied Command Transformation, 2026).
Growing budgets still have to become usable capability
SIPRI's April 2026 release put world military expenditure in 2025 at $2,887 billion, up 2.9% in real terms, with Europe, including Russia and Ukraine, at $864 billion and Asia and Oceania at $681 billion. SIPRI also noted that spending approved by the US Congress for 2026 has risen above $1 trillion. These are totals for all military expenditure, covering personnel, operations and construction as well as equipment and research, so they describe the size of the fiscal tide rather than an addressable market for autonomous systems (SIPRI, 2026).
NATO's policy language has also moved. The Hague commitment of June 2025 is a target for 2035: at least 3.5% of GDP for core defence and up to 1.5% for broader defence- and security-related spending, with the trajectory to be reviewed in 2029. At Ankara in July 2026, Allies named uncrewed systems among their capability priorities, committed to an interoperable transatlantic warfighting cloud and to adopting powerful AI models, and announced more than $50 billion in new procurements. Neither declaration assigns a share of that money to autonomy (NATO, 2025, 2026a).
For a commercial director, the chain that matters runs from policy to an appropriated requirement, then an order, delivery, customer acceptance and payment, and each link has its own clock. Headline budgets and framework ceilings therefore belong in market intelligence, and funded orders belong in the delivery and cash forecast.
The Army's March enterprise contract with Anduril shows the distinction unusually clearly. The Army's own announcement gives a total estimated value of up to $20 billion and states that this is the maximum potential value, not an obligated amount. The ten-year contract has a five-year base period and a five-year optional ordering period, and it consolidates more than 120 separate procurement actions into one framework with pre-negotiated terms. It shortens the route to buying hardware, software and support. The Army also stated that the contract will not substitute for competition on future programmes, so it neither creates $20 billion of revenue nor closes the market (US Army, 2026).
The deployment cycle determines where value can accumulate
The deployment cycle is a more useful industrial map than a fixed division between valuable software and commoditised hardware. A manufacturer supplies a platform or component. An integrator connects it to communications, mission data and command systems. Testing establishes the accepted configuration, and the customer authorises its use. Once fielded, the capability needs replenishment, maintenance, training and controlled change. Each of these steps creates a different commercial position, even where one company supplies several of them.
NGC2 shows how physical these steps remain. After Project Convergence Capstone 6 in the Mojave Desert in July, senior Army leaders called the test a success while acknowledging that hardware had performed poorly in the heat, and DefenseScoop reported that a new set of drones in the exercise was not yet connected to NGC2. In September, III Armoured Corps began receiving baseline NGC2 capabilities through a consolidated fielding process. The software sets the direction, but environmental qualification, power, cooling and the integration of each new drone decide when a unit can actually rely on it (Lawrence, 2026).
Value accumulates where a supplier makes those transitions repeatable. A standard integration package shortens the work for the next payload. Reusable test evidence lowers the cost of the next approved configuration. A local support organisation improves availability and response time. The benefit is specific: fewer engineering hours per deployment, earlier acceptance, lower inventory exposure or a larger support scope. None of it can be read from the presence of AI in a product description.
A Dutch example makes the same point at a smaller scale. Reuters reported on 8 October that a late-September test at a former airbase in Katwijk aan Zee saw Intelic's Nexus software combine radars, sensors and cameras from other companies, use a reconnaissance drone with image recognition to verify the target, and launch an interceptor drone after a human made the final decision. Intelic holds a €30 million contract with the Dutch Ministry of Defence, and its system is being rolled out in Ukraine alongside Ukraine's own Delta battle-management system. It remains a demonstration, not evidence of general operational performance (Deutsch & Biesemans, 2026).
There is a second-order consequence for smaller suppliers. Access to a repeatable integration route can widen their addressable market because each sale requires less bespoke connection work, whereas dependence on a proprietary interface weakens their bargaining position. In an August commentary for RUSI, Linus Terhorst argues that separating autonomy software from hardware lowers barriers to entry and supports industrial surge, and he describes how the UK has tried to limit vendor lock-in by retaining intellectual property and re-competing contracts. My extension is that integration terms also decide the cost and timing of revenue conversion for every company that joins the architecture. A supplier that needs six months of bespoke work per customer has a different cash profile from one that plugs in against a published interface in six weeks, even if both sell the same sensor (Terhorst, 2026).
Europe's opportunity runs through national integration
Europe's opportunity is not a European copy of an American software prime. Defence customers retain national responsibility for operations, support, and the level of dependence they will accept, so a common product may require different communications, weapons, evidence, and support arrangements from one customer to the next. The attractive position is a common technical core that meets those national requirements without requiring a rebuild of the whole system.
Boeing's German MQ-28 team is a concrete example. At ILA Berlin on 10 June 2026, Boeing brought Diehl Defence and Rohde & Schwarz alongside Rheinmetall. Diehl's stated role covers weapons integration, technical data, product support, and training; Rohde & Schwarz is to integrate mission and communications systems with Bundeswehr command and weapon systems and to support operations, maintenance, and logistics; Rheinmetall takes responsibility for integration into Bundeswehr systems. Boeing Defence Australia stated its aim to provide the MQ-28 to the Luftwaffe by 2029, and Rohde & Schwarz described its participation as part of a letter of intent. This is an industrial offer for a potential programme, not a German procurement award (Boeing, 2026).
This division of work shows why sovereignty can create a service and integration market. A national partner's value lies in qualified interfaces, customer-specific technical data, and ongoing support, which can outlast the initial delivery by decades. The cost is engineering capacity and lost scale, so contracts that separate reusable development from one-customer work are easier to manage and to price.
Helsing's financing shows the capital now available to European entrants. On 13 July 2026 the company announced a $1.8 billion Series E at an $18 billion valuation and said it remains predominantly European-owned. That is evidence of financing capacity and investor expectations. A funding announcement does not reveal programme profitability, cash conversion or the return on a particular production investment (Helsing, 2026c).
European procurement speed is often blamed on regulation, with the AI Act as the usual example. The text is narrower than that. Article 2(3) excludes AI systems where and to the extent that they are placed on the market, put into service, or used exclusively for military, defence, or national-security purposes, and the consolidated text as at 27 July 2026, after the Digital Omnibus amendments, retains that exclusion. A dual-use product sold to civil customers as well is not covered for its civil use, and other obligations still apply. For industrial planning, the intended use and the customer's acceptance requirements provide more insight into timelines than the AI Act does (European Commission, 2026).
Military aviation connects Australia, Europe and the United States
Military aviation makes the gap between a capable aircraft and an accepted force configuration visible. In December 2025, Boeing and the Royal Australian Air Force reported an engagement in which an MQ-28, an E-7A Wedgetail and an F/A-18F Super Hornet launched from separate locations. An E-7A operator took custody of the MQ-28 for safety and engagement oversight, the Super Hornet tracked the target and shared the data, and the MQ-28 received authorisation from the E-7A before destroying a fighter-class target drone with an AIM-120 missile. Boeing said the integration was completed in under eight months. The significance lies in the demonstrated combination of assets, oversight, and weapons integration, subject to the limitations of a company-reported test (Boeing, 2025).
The Australian programme's move towards German industry shows a path from domestic development to international adaptation. For a supplier, the question goes beyond whether a platform can be exported. It is who performs the adaptation, who owns the resulting evidence, and whether a second customer can reuse the work. That is also why I would read Asia-Pacific through specific programmes like this one, and their operating requirements, rather than through a regional share of a loosely defined global autonomy market.
The aviation portfolio is broadening quickly. Helsing unveiled the CA-1 Electronic Attack at ILA in June 2026 as an autonomous electronic-warfare aircraft that shares its airframe, propulsion, autonomy software and ground control with the kinetic CA-1 variant. Helsing plans first flights of the attack version in early 2027 and gives initial operating capability dates of 2029 and 2031 for the two variants. Anduril and Archer Aviation unveiled Thunder at Farnborough on 20 July as a Group 5 autonomous attack rotorcraft; full-scale surrogate aircraft have flown, and Thunder's own first flight is planned for 2027. Both are development announcements. Helsing's shared airframe is itself a deployment-cycle choice: one qualification base, several payloads (Helsing, 2026b; Archer Aviation, 2026).
Space is entering the same architecture through the supply of information. In May, Helsing and OHB announced a joint venture, KIRK, to lead a consortium with HENSOLDT and Kongsberg to develop a space-based tactical surveillance, reconnaissance, and targeting system. It is a development initiative, not an operational constellation. The industrial point is that platform autonomy can depend on data produced elsewhere, thereby bringing ground infrastructure, space-qualified sensors, and information rights into the commercial relationship (Helsing, 2026a).
Recurring revenue depends on the right to change the system
In July I illustrated lifecycle economics with a universal split between hardware and operating costs. I would not repeat it. A fleet of expendable drones and a fleet of reusable combat aircraft have different replacement, maintenance, and software requirements, and no single percentage applies to both. Any comparison needs a defined mission, fleet policy, operating period and treatment of losses.
A stronger assessment starts with the contract. Who funds a software revision driven by a new threat? Does the customer receive rights to modify the system, or only a licence to use the current version? Is validation included in the service price? Can a local partner install an approved update? The answers determine whether continuing work generates recurring revenue, consumes an existing warranty provision, or requires a new order. The word software, on its own, says little about the resulting margin.
Anduril's May financing announcement is a useful reference point. The company raised $5 billion at a $61 billion valuation, and its chief executive wrote that revenue doubled in 2025 to $2.2 billion. These are disclosures by a private company, not audited segment results, and they do not support a margin comparison with listed primes. They do show how much capital and commercial scale now sit behind a software-led defence model. In addition to Anduril's own October statement that no single company should control the whole command-and-control stack, they also suggest that even the strongest new entrants expect customers to insist on change rights and portability (Bort, 2026; Lawrence, 2026).
Recurring support pays when it combines a real customer need with reusable engineering, and becomes expensive when every customer requires a separate software branch, security process and acceptance package. That relationship between reuse and sovereignty is where industrial strategy meets the income statement.
The strongest counterargument deserves a place in the investment case
The strongest alternative view is that production and replacement capacity will capture more value than orchestration. A customer facing high consumption may favour equipment that is cheap, available and simple to operate, and extra integration can add cost and dependencies that do little for a particular mission. That view supports the distinction between expendable mass and higher-end platforms, and it weakens any claim that all hardware suppliers are heading towards commodity economics. The Mojave heat problems point the same way: a capability can be held back by physical equipment long after the software is ready.
Established primes also bring real strengths: platform responsibility, production infrastructure, customer relationships and support capacity. Flight control, propulsion, sensing and qualification remain necessary even as mission software gains influence, and the Boeing examples show those capabilities being combined with autonomy rather than displaced by it. The outcome remains programme-specific.
The same discipline applies to contested communications. NATO's experiments address complex connectivity under human oversight; they do not establish that every network will hold under every adversarial condition. Public evidence does not support a single figure for how far autonomy can scale before communications become the limit. A better industrial question is which demonstrated capability depends on an available link, what happens when that link is interrupted, and how the accepted configuration handles that case.
I therefore keep the July expectation of uneven, partly fragmented adoption, and I would now treat fragmentation as a commercial design condition. Companies can create value by lowering the cost of connecting and supporting different systems, while customers keep legitimate reasons to hold national boundaries. Full convergence is not needed for those improvements to pay.
The next twelve months will test conversion into service
Between October 2026 and October 2027, I expect integration around particular missions to advance faster than any universal, cross-allied autonomy architecture. I put this as more likely than not, roughly two chances in three, and it follows from the fragmented-adoption view I set out in July. The recent examples all centre on defined requirements: an Army data layer for a corps, maritime awareness in the High North, a national offer for a collaborative combat aircraft and counter-drone protection. Each gives a customer a bounded capability it can evaluate, fund and accept.
Three developments would confirm the view. Funded orders would follow the framework and development announcements, starting with option exercises on the NGC2 contract. Acceptance evidence would become reusable when a second platform or national partner joins an existing architecture. And support contracts would specify who pays for updates, testing, and availability after delivery. They would say more than another valuation or prototype unveiling.
The calendar gives clear checkpoints. NATO's Norway trial is due at the end of 2026, and Task Force X-Arctic plans a small-scale exercise in the first half of 2027 and a full multi-domain demonstration in the High North by the summer of 2027. Helsing's CA-1 attack variant is due to fly in early 2027, and Thunder's first flight is planned for the same year. None of these is evidence of delivery today (NATO Allied Command Transformation, 2026; NATO, 2026b; Helsing, 2026b; Archer Aviation, 2026).
The view would weaken if interoperability work stays confined to demonstrations, if operational fielding is repeatedly deferred, or if customers spend most new money on isolated platforms without continuing integration support. It would also need revision if one common architecture gains broad operational adoption sooner than I expect. If the central case holds, the suppliers that gain are those with reusable interfaces, qualification evidence, and local support capability; the ones that lose ground are those whose offer depends on bespoke integration for every customer, regardless of the quality of their hardware.
A portfolio discussion around deployment and cash
My addressee here is the board of a European tier-one or tier-two defence supplier whose sensors, effectors, or platforms must plug into an architecture controlled by someone else. Were I part of that discussion this autumn, the option I would put to the chair is simple enough to minute: run one live autonomy opportunity through a deployment-stage review, and link the next capital commitment to an acceptance milestone rather than to a framework ceiling. It sits alongside the existing pipeline review.
The review distinguishes five stages: a successful demonstration, eligibility under a purchasing framework, a funded order, an accepted configuration and a supported capability. Each has value, but each supports a different planning assumption, and making the stages visible stops optional contract value from driving inventory and hiring ahead of customer commitment. In the first 90 days, one accountable programme owner, with commercial, engineering and finance colleagues, would map one opportunity: which evidence already exists, what acceptance requires, which national adaptations are still unfunded, and what the support and update arrangement should say. By six months, that understanding should be in the customer conversation early enough to shape the contract, including who pays for threat-driven changes and who may install an approved update. By twelve months, the board would know whether the approach reduced rework and accelerated acceptance, and could decide whether to extend it across the portfolio.
The economic case should be measured, not promised. I would not put a percentage on it before the pilot, because the evidence to support one is the company's own. The indicators are clear enough: engineering hours per additional deployment, months from award to acceptance, the share of backlog that is funded rather than framework value, inventory committed ahead of orders, and support revenue as a share of the installed base. If the first two move in the right direction on one programme, the case for a wider roll-out makes itself.
Operations would add one question: can the promised configuration be produced and maintained repeatedly? If it depends on a scarce sensor or a single specialist engineer, a second qualified source or documented integration knowledge may do more to build delivery confidence than extra assembly capacity.
Why this autumn? The Army is moving NGC2 from prototype to corps fielding, NATO's Arctic programme is moving into its second phase in the first half of 2027, and the German MQ-28 team is working toward a 2029 date. Integration terms for companies joining these architectures are being set now, and they will be much harder to change once the first configuration is accepted. Waiting does not close the market, but it can leave a supplier joining on terms written for someone else.
The trade-off is the cost of building reusable capability before repeat demand is secure. I would favour staged investment, with capacity following funded commitments wherever possible, because committing the whole capital plan to a framework ceiling creates its own exposure. I would advise differently for a supplier whose business is mainly expendable mass sold on price and availability, where production scale matters more than integration, or for a supplier where the customer insists on owning the architecture and its intellectual property outright. For most suppliers between those cases, the middle path is an explicit line from the next acceptance milestone to the next industrial commitment. Autonomous capability becomes commercially durable when the organisation can keep delivering an approved system as its mission changes, and that is a discipline a board can start building on a single programme.
Questions for the C-Suite and the Board
Which opportunities in our remit have a funded route from demonstration to acceptance? For the chief commercial officer, separating those stages shows where engineering time and production commitments rest on evidence of conversion.
Where could reusable integration evidence shorten the next customer deployment? A programme manager can identify work that improves delivery confidence while reducing repeated engineering cost.
Which software changes are included in the price, and which require a new order? For the CFO, the answer clarifies support scope, margin exposure and the real basis for recurring revenue.
What national support capability would make our existing offer more useful to the customer? Integration, training or maintenance partnerships may widen market access without owning every function.
Which dependency most influences availability after initial delivery? A supplier, component, approval or specialist skill may be a better investment target than additional assembly capacity.
What rights would preserve customer choice while protecting our reusable intellectual property? This opens a practical discussion with the customer about portability, national adaptations and long-term partnership economics.
What evidence would justify the next capital commitment? Linking expenditure to an order, an acceptance milestone, or a repeat requirement keeps the growth plan responsive to real demand.
References
Archer Aviation. (2026, July 20). Anduril and Archer unveil jointly-developed autonomous VTOL platform for commercial and defense applications [Press release]. Archer Aviation. https://archer.com/news/anduril-and-archer-unveil-jointly-developed-autonomous-vtol-platform-for-commercial-and-defense-applications
Boeing. (2025, December 9). Boeing, RAAF achieve CCA missile fire from MQ-28 Ghost Bat [Press release]. Boeing. https://investors.boeing.com/investors/news/press-release-details/2025/Boeing-RAAF-Achieve-CCA-Missile-Fire-from-MQ-28-Ghost-Bat/default.aspx
Boeing. (2026, June 10). Boeing erweitert deutsches MQ-28 Ghost Bat-Industrieteam [Boeing expands German MQ-28 Ghost Bat industry team] [Press release]. Boeing Deutschland. https://www.boeing.de/presse/2026/boeing-erweitert-deutsches-mq-28-ghost-bat-industrieteam
Bort, J. (2026, May 13). Anduril raises $5B, doubles valuation to $61B. TechCrunch. https://techcrunch.com/2026/05/13/anduril-raises-5b-doubles-valuation-to-61b/
Deutsch, A., & Biesemans, B. (2026, October 8). Dutch firm uses AI to help drone systems 'talk' on Ukraine's battlefield. Reuters. https://www.reuters.com/business/autos-transportation/dutch-firm-uses-ai-help-drone-systems-talk-ukraines-battlefield-2026-10-08/
European Commission. (2026). Article 2: Scope. Regulation (EU) 2024/1689 (AI Act), consolidated text as at 27 July 2026. AI Act Service Desk. Retrieved October 12, 2026. https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-2
Helsing. (2026a, May 19). Helsing and OHB establish joint venture "KIRK" for tactical space-based reconnaissance systems [Press release]. Helsing. https://helsing.ai/newsroom/helsing-and-ohb-establish-joint-venture-kirk-for-tactical-space-based-reconnaissance-systems
Helsing. (2026b, June 10). Helsing introduces the CA-1 Electronic Attack, an autonomous combat aircraft for electronic warfare [Press release]. Helsing. https://helsing.ai/newsroom/helsing-introduces-the-ca-1-electronic-attack-an-autonomous-combat-aircraft-for-electronic-warfare
Helsing. (2026c, July 13). Helsing raises US$1.8bn in Series E [Press release]. Helsing. https://helsing.ai/newsroom/helsing-raises-1-8bn-in-series-e
Lawrence, D. F. (2026, October 6). Army awards Anduril contract worth up to $1.8B over five years to expand NGC2. DefenseScoop. https://defensescoop.com/2026/10/06/army-awards-anduril-1-8b-contract-expand-ngc2/
NATO. (2025, June 25). The Hague Summit Declaration. North Atlantic Treaty Organisation. https://www.nato.int/en/about-us/official-texts-and-resources/official-texts/2025/06/25/the-hague-summit-declaration
NATO. (2026a, July 8). The Ankara Summit Declaration. North Atlantic Treaty Organisation. https://www.nato.int/en/about-us/official-texts-and-resources/official-texts/2026/07/08/the-ankara-summit-declaration
NATO. (2026b, October 8). REPMUS 2026 in Portugal brings together NATO Allies to test uncrewed maritime systems. North Atlantic Treaty Organisation. https://www.nato.int/en/news-and-events/articles/news/2026/10/08/repmus-2026-in-portugal-brings-together-nato-allies-to-test-uncrewed-maritime-systems
NATO Allied Command Transformation. (2026, September 24). Task Force X-Arctic takes its next step at REPMUS 2026. NATO Allied Command Transformation. https://www.act.nato.int/article/tfx-arctic-repmus/
Stockholm International Peace Research Institute [SIPRI]. (2026, April 27). Global military spending rise continues as European and Asian expenditures surge [Press release]. SIPRI. https://www.sipri.org/media/press-release/2026/global-military-spending-rise-continues-european-and-asian-expenditures-surge
Terhorst, L. (2026, August 4). Swarming software's defence industrial implications. Royal United Services Institute. https://www.rusi.org/explore-our-research/publications/commentary/swarming-softwares-defence-industrial-implications
US Army. (2026, March 13). U.S. Army awards enterprise contract for IT commercial solutions. U.S. Army. https://www.army.mil/article/291074
About Andy Demir
Andy Demir is the editor of SwissTechnic and writes on strategy, international growth and capital in aerospace, defence, space and advanced technology. Drawing on a career in commercial leadership, the analysis examines how technology, sovereignty and industrial capability shape competitive power and long-term value.
