Which business models work in military robotics?

Has military robotics finally moved from demos to real procurement?

Military robotics is clearly becoming a real procurement market now, with several categories already moving from dozens of prototypes to thousands or even tens of thousands of systems.

Ukraine gives us the cleanest evidence because battlefield demand has forced procurement to move unusually fast. Its Defence Procurement Agency said contracts for drones reached UAH 333.6 billion in the first half of 2026, twice the amount contracted during the same period a year earlier. FPV drones represented the largest share.

Ground robotics is following the same path. By July, Ukraine had already contracted more than 22,000 unmanned ground vehicles for 2026, almost twice as many as during all of 2025. The useful detail is what soldiers are actually doing with them: most purchases are logistics robots, while Ukrainian units completed more than 16,600 robotic logistics and evacuation missions in June alone. That was 122% more than in January.

The U.S. market is moving toward volume too. Skydio received a U.S. Army order worth more than $52 million for over 2,500 X10D drones, which the company described as the Army's largest single-vendor tactical small-drone order. Anduril had meanwhile accumulated more than 120 separate U.S. government procurement actions before the Army grouped its commercial products into one enterprise contract.

Some categories of military robotics are therefore already behaving like industrial procurement markets. The harder question now is which companies can turn those orders into durable businesses.

What does a military robotics business model actually need to prove?

A military robotics business model works when operational success keeps generating funded orders, follow-on purchases or service revenue without forcing the company to reinvent the product for every customer.

That distinction removes a lot of noise from the sector. A $1 billion contract ceiling can produce very little revenue if customers never issue meaningful orders under it. A $50 million production order can tell us much more when thousands of units are actually being delivered.

AeroVironment gives us a useful benchmark because its financial statements show what mature defense demand looks like. The company finished fiscal 2026 with roughly $1.2 billion of funded backlog and $2.7 billion of annual bookings. Management expected about 85% of that funded backlog to convert into revenue during the following fiscal year. Total revenue had reached $1.98 billion, compared with $821 million a year earlier, although acquisitions explain part of that jump.

Funded backlog means customers have committed money. Bookings show new business arriving. Production revenue tells us systems are leaving factories. Reorders tell us customers still want them after using them.

When we assess the main models today, the gap between the strongest and weakest ones is already fairly clear.

Business model	Evidence of repeat demand	Recurring potential	Our view today
Mass-produced attritable robots	Very strong	High through replacement	Proven
Reusable hardware + autonomy software	Strong	High	One of the best models
Contractor-operated robotic services	Strong in ISR	Very high	Proven in selected missions
Maintenance and sustainment	Strong	High	Proven
Open platform + payload ecosystem	Growing	High	Strong
Sovereign licensing / local production	Growing quickly	High	Increasingly important
Specialist autonomy or subsystem supplier	Mixed but improving	High	Attractive when deeply integrated
Standalone autonomy software	Still limited	Potentially very high	Promising, less proven
One-off robot sales	Strong demand can exist	Low	Viable but exposed
Prototype engineering	Easy to monetize initially	Low	Weak as an end state

If you want more recent data on this point, please see our latest military robotics market report.

Is simply selling military robots still a good business?

Yes, selling military robots still works very well today when the product is standardized, genuinely useful and easy enough to manufacture in volume.

The idea that hardware is already becoming secondary to software gets ahead of the procurement evidence. Skydio's 2,500-plus X10D order is fundamentally a hardware purchase. Ukraine is buying drones and ground robots in enormous quantities. Helsing announced another 6,000 HX-2 strike drones for Ukraine after an earlier 4,000-unit order.

AeroVironment also shows that physical autonomous systems can support a substantial public business. Its Uncrewed Aircraft Systems division generated about $364 million in fiscal 2026 revenue, while Precision Strike and Defense Systems generated about $848 million. Across the whole company, product revenue remained a major part of the business even after the BlueHalo acquisition expanded services.

What has changed is the standard that hardware companies must meet. A robot that performs brilliantly in demonstrations but takes months to build has a weak position in a market increasingly obsessed with replacement speed, supply chains and cost per mission.

Can cheap expendable drones really make good money?

Yes, expendable military drones can support very large businesses, but the economics depend heavily on volume because individual airframes are becoming easier to replace and harder to differentiate.

Ukraine makes that visible. Drone contracting reached UAH 333.6 billion in the first half of 2026, double the level from a year earlier, while FPV systems accounted for the largest share. The UK has also moved toward six-figure annual drone deliveries to Ukraine after supplying only around 10,000 in 2024.

That scale creates natural repeat demand. A reusable surveillance aircraft may remain in a fleet for years; an FPV strike drone can generate another procurement need almost immediately after being used.

The uncomfortable part for manufacturers is that consumption does not guarantee pricing power. Ukraine now has hundreds of domestic drone suppliers, and its DOT-Chain Defence marketplace allows military units to choose equipment directly. The government says roughly 95% of drones bought through the Defence Procurement Agency are domestically produced.

Helsing is trying to protect itself from that pressure with the HX-2. The company combines a mass-manufactured strike drone with onboard autonomy, electronic-warfare resilience and integration into its wider Altra software environment.

Expendable drone businesses look strongest when the company controls something harder to commoditize than the basic airframe: autonomy, targeting performance, protected components, manufacturing throughput, proprietary mission data or the software coordinating many drones at once.

Is hardware plus autonomy software the best military robotics business model?

For reusable military robots, hardware plus autonomy software currently looks like the strongest all-round model.

Shield AI gives us one of the clearest examples. The Indian Army selected V-BAT aircraft alongside licenses for Hivemind autonomy software and the Hivemind software-development kit. Shield AI therefore gets to monetize the aircraft, the autonomy layer and future sovereign development around the same deployment.

The same logic appears at Anduril on a much larger architecture. The company's Army enterprise vehicle covers commercial hardware, software, data infrastructure and support. Separately, the Army chose Anduril to lead the common data-layer baseline for Next Generation Command and Control, where Lattice connects data, applications and systems from several vendors.

That combination creates several chances to make money from one customer relationship. The initial vehicle can be followed by autonomy upgrades, integrations, new payloads, training, support and fleet expansion.

Shield AI's latest operating update reinforces the point. The company said fiscal 2026 orders and revenue had both doubled year over year, while Hivemind was piloting 30 classes of vehicles. The interesting part is the breadth: Hivemind is gradually moving beyond one aircraft and into a reusable autonomy layer.

If you want more recent data on this point, please see our latest military robotics market report.

Can military autonomy software make money without selling the robot?

Yes, but standalone military autonomy software still has much less procurement evidence behind it than software bundled with hardware or integration work.

The opportunity is obvious. An autonomy stack that works across many vehicles can theoretically scale much faster than a manufacturer building each platform itself. Shield AI is already moving in that direction. Taiwan's National Chung-Shan Institute signed a contract to integrate Hivemind into locally developed unmanned systems, while ST Engineering is evaluating Hivemind across selected platforms.

Those deals matter because Shield AI does not need to manufacture every vehicle using its software. Its Hivemind SDK can become part of another country's or manufacturer's robotics stack.

Anduril's role in the U.S. Army's NGC2 architecture points in the same direction. Lattice increasingly connects applications, sensors, data and systems built by multiple companies. The Army moved from prototype work into an operational common data baseline after field validations with major divisions.

Still, pure software procurement remains relatively uncommon. Even Shield AI's strongest commercial examples frequently combine software licenses with V-BAT aircraft, integration or engineering. Defense customers usually buy an operational outcome, and somebody still has to make the software work on the actual machine.

Does robotics-as-a-service actually work for military customers?

Yes, robotics-as-a-service already works in defense, especially for persistent surveillance missions where militaries care more about coverage and intelligence than owning every vehicle.

Shield AI has used this model with the U.S. Coast Guard, winning an ISR-as-a-service contract vehicle with a ceiling close to $198 million around V-BAT operations. The company operates the aircraft instead of simply shipping them to the customer.

Maritime robotics provides even stronger evidence. Saildrone has spent years operating unmanned vessels for U.S. forces and government agencies while retaining responsibility for the vehicles and fleet operations. A U.S. Navy extension announced this year added $16.3 million for operations in Southern Command after the deployed fleet had grown to roughly 20 vessels.

Saronic is now entering the same market. A Navy delivery order for contractor-owned, contractor-operated Corsair unmanned surface vessels carried a potential value of about $57.8 million, with roughly $29 million initially obligated.

This model can produce unusually attractive revenue because one robot can be monetized repeatedly across months of operations. Capital intensity is the trade-off: the company has to finance or own the robots and keep them working.

Why can military robot maintenance become a business of its own?

Military robot maintenance can produce years of revenue after the original sale because reusable fleets need repairs, spares, training, software changes and engineering support throughout their operating lives.

QinetiQ's U.S. ground-robot business shows how long that tail can become. Contracts around TALON and Common Robotic System fleets have included depot repairs, spare parts, maintenance, training, conversions, engineering upgrades and field support. Those activities continue long after the original robot leaves the factory.

Software-defined fleets add even more work. Army procurement documents around Anduril's Lattice-Fortress environment have included sustainment, engineering and integration support because deployed systems need to evolve as networks, sensors and threats change.

NATO is now explicitly thinking about procurement in similar terms. Its new counter-UAS framework contracts mention the possibility of shared through-life support alongside the initial equipment, with common command-and-control infrastructure connecting interchangeable sensors and effectors.

A large installed base can therefore keep generating revenue even when new-unit production slows.

Are faster military marketplaces changing which robotics companies win?

Yes, faster defense marketplaces are making it easier for good military robots to reach users while exposing weak products much faster.

Ukraine's DOT-Chain Defence marketplace is the most advanced example we found. Frontline units receive budgets and choose equipment themselves, while the procurement agency handles contracting, payment and logistics. By mid-2026 the marketplace covered almost 800 products from more than 200 Ukrainian manufacturers, and average delivery time for in-stock products had fallen to nine days.

That is a major change from the classic defense model where a central procurement organization could spend years selecting one standardized system.

NATO is moving in the same general direction. Its procurement agency established five framework contracts for counter-drone systems this year, giving allies access to prequalified solutions built around a common command-and-control backbone. NATO Engine is also trying to connect defense demand with unused industrial capacity across allied countries; by July it reported more than 840 participating companies and over 100 active opportunities.

Smaller robotics companies can enter the market faster, while customers can also switch suppliers more easily when battlefield data shows that another product performs better.

Can international sales become as important as selling to the home military?

Yes, international military robotics revenue can become a major part of the business, and local production or software licensing is making those exports easier to win.

Shield AI shows where the market is heading. India is buying V-BAT aircraft together with Hivemind software licenses and an SDK that supports sovereign development. Taiwan has separately contracted Hivemind integration onto locally developed unmanned systems. These deals let Shield AI participate in national robotics programs without insisting that every platform come from its own factory.

That model fits what governments increasingly want. Countries like India, Poland, Taiwan and Ukraine care about supply security and national control over critical defense technology. A vendor willing to license software, support local assembly or share production can access customers that might reject a completely foreign black-box system.

AeroVironment shows that international demand can already become meaningful at scale. Its fiscal 2026 filings reported roughly $556 million of international revenue against nearly $2 billion of total company revenue. Foreign customers therefore represented close to 28% of the business.

Do open military robots create better businesses than closed systems?

Usually yes, because open military robots can keep generating revenue through new payloads, software and subsystems, while specialist suppliers can also sell into several competing platforms.

The maritime market makes this easy to see. Saronic designed Corsair around an open architecture that can accommodate different payloads and autonomy systems. Saildrone and Lockheed Martin have taken another route: Saildrone provides the unmanned vessel while Lockheed integrates existing defense payloads and combat systems.

The partnership recently demonstrated how quickly that approach can change a robot's role. Within months, a Saildrone Surveyor had been integrated into military exercises involving electronic-warfare capabilities and remotely launched weapons. The vessel had effectively moved from an intelligence platform toward a wider combat-system role without requiring a completely new hull.

Ground robotics is developing similarly. Milrem has repeatedly integrated third-party weapon stations and other payloads onto THeMIS. Rheinmetall's newer autonomous logistics programs combine vehicle technology, autonomy and human-machine interfaces supplied by different companies.

That also creates room for specialist suppliers. Rheinmetall can act as prime contractor while Harbinger contributes vehicle technology, Forterra supplies autonomy and Primordial Labs provides the human-machine interface. Shield AI's newer Hivemind contracts point in the same direction: the autonomy stack can generate revenue even when another company builds the physical vehicle.

The prime contractor captures more of the contract and owns the main customer relationship. A specialist can potentially appear across many competing programs with less integration burden.

The best specialist position is to own technology that several primes need and cannot quickly recreate.

If you want more recent data on this point, please see our latest military robotics market report.

Which military robotics markets support the best business models?

The best business model changes sharply between expendable drones, reusable aircraft, ground robots and unmanned vessels because each type of machine has a different operating life and replacement cycle.

FPV and one-way strike drones behave increasingly like munitions. Their commercial engine is replenishment. Reusable tactical UAVs create more room for software, maintenance and payload upgrades. Ground logistics robots produce fleet sales plus a large sustainment tail because mechanical systems operating in difficult terrain need constant maintenance.

Maritime systems are especially well suited to service revenue. A contractor-owned autonomous vessel can remain deployed for months and sell surveillance output continuously. Cross-platform autonomy sits at the other extreme because the same software can theoretically generate revenue across many different physical systems.

Counter-drone systems create another distinctive model. Threats change so quickly that operators continually need new sensors, electronic-warfare techniques and interceptors. The command-and-control layer can remain while surrounding components are replaced.

Military robotics category	Model that fits best	Why
FPV drones	High-volume product sales	Rapid battlefield consumption creates constant replacement
One-way strike drones	Production + autonomy	Scale matters, while software preserves differentiation
Tactical reusable UAVs	Hardware + software + support	Aircraft can be upgraded throughout their service life
Ground logistics robots	Fleet sales + sustainment	Heavy use creates maintenance and replacement demand
EOD robots	Hardware + long-term support	Fleets often remain deployed for many years
Unmanned surface vessels	Service model or platform sales	Persistent missions create high asset utilization
Large autonomous aircraft	Production + autonomy integration	High unit values support sophisticated software and support
Counter-UAS systems	Integration + recurring upgrades	Sensors and effectors change with the threat
Cross-platform autonomy	Licensing + engineering	One software stack can reach many vehicles
Command-and-control software	Enterprise software + integration	Value increases as more systems connect to the network

Do military robotics companies really need recurring revenue?

No. Military robotics companies can build very good businesses without SaaS-style recurring revenue, as long as something keeps bringing the customer back.

Attritable drones provide the simplest example. Each system may generate revenue only once, yet battlefield consumption creates repeated orders. That can be economically equivalent to recurring revenue when procurement volumes remain high.

Reusable platforms create recurrence differently. Shield AI can sell autonomy licenses after the aircraft purchase. QinetiQ can repair and support robots already in service. Saildrone can bill for continued operations. An open platform can generate new payload and integration revenue years after delivery.

AeroVironment's financial performance is useful here because it shows that defense hardware does not need to look like enterprise SaaS to become financially meaningful. The company produced nearly $2 billion of fiscal 2026 revenue, a $1.2 billion funded backlog and $286 million of adjusted EBITDA across a portfolio still heavily exposed to physical defense products. Its GAAP result remained negative because of acquisition-related and other expenses, so we should not confuse adjusted profitability with clean bottom-line profitability.

The useful question is whether deployment naturally leads to another purchase: replacement units, software, mission hours, spare parts, training, payloads or fleet expansion.

What is the biggest business-model trap in military robotics?

The biggest trap in military robotics is treating large contract announcements as proof that customers are actually buying large quantities.

Defense contracting makes this particularly dangerous. Anduril's Army enterprise contract carries a ceiling as high as $20 billion, but that figure represents the maximum value available under the vehicle rather than $20 billion of guaranteed orders. Its real significance is that more than 120 previous procurement actions can now flow through a simpler framework.

Multi-award contracts create the same problem. A robotics company may announce that it has been selected for a $500 million or $1 billion contract vehicle while receiving only a small fraction of that amount later.

Production orders deserve more weight. Skydio's $52 million-plus award tells us the Army is purchasing more than 2,500 actual X10D drones. Ukraine's 22,000-plus UGV contracts show real units entering procurement. AeroVironment's funded backlog tells us money has already been committed.

Prototype awards sit lower down. They can finance useful engineering and occasionally create an excellent route into production, but plenty never get there.

We would therefore read military robotics announcements in a strict order: funded production and active service contracts first, funded backlog and follow-on orders next, framework ceilings after that, prototypes below them, and memoranda or demonstrations last.

If you want more recent data on this point, please see our latest military robotics market report.

Will military robots become cheap hardware while software captures all the value?

Only the simplest military robots are likely to become true commodity hardware; advanced platforms will keep substantial value in the physical machine.

Small drones already show serious commoditization pressure. Ukraine has a dense domestic supplier base, military units can compare products directly through marketplaces, and procurement increasingly reflects frontline performance. A basic airframe therefore has fewer ways to defend a large premium.

Software gains importance in that environment because autonomy, mission planning and coordinated control can improve many otherwise similar machines.

Physics still sets hard limits. A maritime vessel's range, payload, reliability and sea-state performance cannot be fixed with a software update. The same applies to aircraft endurance, acoustic signature, propulsion, thermal management and electronic-warfare resilience. Those features can determine whether the mission works at all.

Simple expendable systems will face increasingly brutal price and manufacturing competition. Sophisticated reusable platforms can maintain stronger margins when their physical performance remains difficult to reproduce.

Which business models actually work in military robotics?

The military robotics business models working best today are mass-produced attritable systems, reusable hardware combined with autonomy and support, and contractor-operated robotic services.

Mass-produced expendable robots already have the strongest proof of volume. Ukraine's procurement shows what happens when autonomous systems become consumables: orders can reach tens or hundreds of thousands because battlefield use continually creates replacement demand. The risk is straightforward price compression, which makes manufacturing efficiency and software differentiation increasingly important.

Reusable hardware plus autonomy looks stronger over a full customer lifetime. Shield AI can sell V-BAT, Hivemind licenses and engineering around the same deployment. Anduril can sell physical systems while Lattice extends into command, data and integration. These companies have several ways to make money after the first robot ships.

Contractor-operated fleets are producing some of the clearest recurring revenue. Saildrone, Shield AI and now Saronic have all shown that governments will pay for autonomous ISR or maritime operations without necessarily owning and operating every vehicle themselves.

Sustainment is less glamorous and extremely real. Mature fleets need repairs, spares, training and engineering for years. Open platforms add another revenue layer because payloads and mission systems can be changed without replacing the entire robot.

Standalone autonomy software still deserves a lower ranking today. Shield AI and Anduril are proving that autonomy can spread across third-party platforms, but much of the meaningful procurement still comes with integration, hardware or broader system responsibility attached.

Prototype contracting comes last. It can finance development and open doors, yet companies that stay dependent on prototype awards never really escape project work.

The clearest pattern across all of these models is repeatability. The strongest military robotics companies have built a reason for the customer to spend again after the first deployment: another expendable vehicle, another fleet expansion, another software license, another mission month, another payload or another maintenance cycle.

Rank	Business model	Why it works now	Main weakness	Verdict
1	Hardware + autonomy + sustainment	Several revenue layers around one fleet	Hard to execute well	Strongest overall
2	Mass-produced attritable robots	Battlefield consumption creates repeated orders	Price pressure	Extremely strong
3	Contractor-operated robotic services	One fleet generates repeated mission revenue	Capital intensive	Strong in ISR and maritime
4	Open platform + payload ecosystem	Deployed robots keep absorbing upgrades	Needs broad adoption	Strong
5	Sustainment and support	Installed fleets generate years of spending	Depends on platform longevity	Proven
6	Sovereign licensing and local production	Opens large foreign programs	IP and technology-transfer risk	Growing quickly
7	Specialist subsystem supplier	Can appear across several platforms	Less control over the customer	Attractive
8	Standalone autonomy software	Excellent potential operating leverage	Still relatively little standalone procurement	Promising
9	One-time robot sales	Straightforward and proven	Weak lifetime customer value	Works
10	Prototype contracting	Customers help finance development	Easy to get stuck before production	Poor endpoint

If you want more recent data on this point, please see our latest military robotics market report.