Which military robotics startups are moving from demos to deployment?

What should count as a military robot actually being deployed?

A military robot is genuinely deployed when an operational unit keeps using it for real missions, rather than testing it once during an exercise or technology trial.

Defense companies stretch the word “deployed” surprisingly far. A drone flown at Project Convergence, a ground robot tested by soldiers for a week or an autonomous boat participating in a naval exercise can all appear in company announcements as deployment milestones. Those events tell us something about technical maturity, but they do not show that a military relies on the system.

We use a higher bar. Combat use qualifies. Repeated operational missions qualify. Formal fielding to active units qualifies. So does follow-on procurement when the first systems are already being used by those units. The strongest cases combine several of these.

The U.S. Army offers a useful comparison. Overland AI, Forterra and Scout AI have all put autonomous vehicles in soldiers’ hands through serious Army programs, but much of that work has explicitly been described as experimentation or prototype evaluation. Skydio’s X10D has gone further: systems were sent directly to operational brigades under urgent requirements, followed by broader procurement authority.

Ukraine makes the distinction even clearer. When a brigade repeatedly sends the same unmanned ground vehicle on ammunition runs and casualty evacuations, repairs damaged systems and asks for hundreds more, we are well beyond a demo.

For this article, the practical test is simple: would the military unit notice if these robots disappeared tomorrow?

Why are military robots finally leaving test ranges, and are Western militaries actually buying them faster?

Military robots are reaching real units faster because Ukraine has compressed development cycles while the U.S. and European militaries are opening shorter routes from testing to procurement.

Ukraine has had the biggest effect. Before the war, a European robotics startup could spend years moving from prototype to government trial and then wait several more years for a procurement program. These days, equipment can reach a Ukrainian unit, encounter jamming, mud, damaged communications and real enemy adaptation, then return to the manufacturer with useful feedback within days or weeks.

ARX Robotics is a good example. The company has discussed how early ground robots exposed weaknesses once they reached battlefield conditions. It kept redesigning them, and GEREON has since grown into what ARX calls the largest Western-built UGV fleet in Ukraine. At the end of September, ARX and DEUTZ delivered the first GEREON vehicles from a new industrial production line only 12 weeks after starting the manufacturing ramp.

Quantum Systems followed a similar path with Vector reconnaissance drones. The company moved from supplying Ukraine to building local repair, training, development and manufacturing capacity there. Vector then won orders elsewhere in Europe and with the U.S. Army.

The procurement system is also becoming less rigid. The U.S. Army has sent commercial drones directly to operational formations, used urgent requirements to bypass slower processes and created marketplaces where units can compare systems before buying them. Skydio X10D benefited from that approach. Ghost X moved through Replicator and Army transformation initiatives. Neros Archer reached an Army program only a few years after the company was founded.

The Navy has pushed a similar idea into autonomous ships. Several commercially developed vessels were evaluated under its MUSV Marketplace, after which successful candidates moved quickly toward production. Saronic’s Marauder went from Navy testing to a production award within months.

Ukraine has taken operator-driven procurement even further by giving frontline units more influence over what gets purchased.

The old acquisition system is still alive, and prototype contracts still fail to turn into recurring orders surprisingly often. But a capable robotics startup now has a much more believable route from prototype to operational unit than it did five years ago.

Which military robotics startups have most clearly moved from demos to deployment?

Quantum Systems, Shield AI, ARX Robotics, Milrem Robotics, Neros, Helsing and Forterra currently have some of the strongest evidence of real military deployment, while Skydio, Anduril, Saildrone and Saronic are moving through different stages of institutional fielding.

Quantum Systems has supplied hundreds of Vector reconnaissance systems to Ukraine and then won additional military customers outside the war. Shield AI’s V-BAT has operated in Ukraine and is now being incorporated into European naval forces. ARX has moved from small batches of GEREON ground robots into a fleet measured in hundreds. Milrem has years of Ukrainian THeMIS experience and a much larger follow-on fleet underway.

The younger strike-drone companies are scaling even faster. Neros moved from supplying FPV systems for Ukraine to a 6,000-drone coalition contract and then a U.S. order covering 14,000 drones and payloads. Helsing announced 6,000 HX-2 strike drones for Ukraine and now says thousands have reached Ukrainian forces. Independent reporting has also shown HX-2 systems being used by frontline crews.

Forterra disclosed that it manufactured and sent 105 Lancer autonomous ground vehicles to Ukraine in roughly six months. It also has a more institutional U.S. path through autonomous tactical vehicles and the Marine Corps ROGUE-Fires program.

Skydio and Anduril are particularly interesting because their strongest public evidence comes from U.S. Army fielding rather than a Ukraine-scale combat fleet. Saildrone has accumulated years of persistent maritime operations. Saronic has recently crossed another important line: after Navy testing of Marauder, it won a production award for the 180-foot autonomous vessel.

Startup	Main system	Strongest evidence today	Where we place it
Quantum Systems	Vector	Hundreds supplied to Ukraine plus repeat allied procurement	Clearly deployed
Shield AI	V-BAT	Ukraine missions and operational European naval adoption	Clearly deployed
ARX Robotics	GEREON	Hundreds deployed or contracted for Ukraine	Clearly deployed
Milrem Robotics	THeMIS	Years of Ukrainian use plus 150+ additional vehicles planned	Clearly deployed
Neros	Archer	Ukraine scale-up followed by a 14,000-system U.S. order	Scaling rapidly
Helsing	HX-2	Thousands sent to Ukraine with frontline use documented	Clearly deployed
Forterra	Lancer / AutoDrive	105 vehicles sent to Ukraine plus U.S. production programs	Clearly deployed
Skydio	X10D	Formal U.S. Army fielding and follow-on procurement	Institutional fielding
Anduril	Ghost X	Fielding with U.S. Army formations	Institutional fielding
Saildrone	Explorer / Voyager	Years of operational naval missions	Clearly deployed
Saronic	Corsair / Marauder	Corsair operations and new Marauder Navy production award	Entering larger-scale deployment

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

Is Quantum Systems already a real military supplier rather than a drone startup?

Quantum Systems has clearly crossed that line: Vector is now a fielded military reconnaissance system with combat experience, repeat procurement and customers beyond Ukraine.

Vector began operating in Ukraine years ago. Quantum then built repair and support capacity inside the country, later adding local manufacturing and continuing to supply larger quantities.

More than 500 Vector systems had already been publicly delivered to Ukraine before the company’s newer expansion. Ukrainian forces have continued selecting the system, including through procurement mechanisms where frontline units get more influence over what they receive.

The adoption has travelled. Spain ordered 91 Vector systems containing 182 aircraft. Romania selected the platform. The U.S. Army later awarded Quantum a $15.3 million contract for Vector AI under its company-level UAS effort.

Quantum is now stretching beyond reconnaissance. The company has moved into interceptor drones through Ukrainian partnerships, with an initial contract covering 15,000 systems and further German-backed procurement expected to reach much larger quantities. Another joint venture received an initial order for 2,000 TerMIT ground robots.

Quantum Systems now looks much more like a military robotics supplier with a proven reconnaissance franchise than a commercial-drone company trying to enter defense.

Has Shield AI’s V-BAT really become an operational military drone?

Shield AI’s V-BAT has already become an operational military aircraft, with deployment evidence across Ukraine and several European armed forces.

Ukraine gives V-BAT its hardest operating environment. Shield AI says the aircraft has conducted hundreds of targeting missions there, including flights under heavy electronic warfare where GPS and communications can disappear.

The stronger evidence for durable adoption comes from what happened afterward. Greece has used V-BAT and expanded its fleet. The Royal Netherlands Navy tested the system at sea, declared it operational and is acquiring 12 aircraft while preparing eight ships to support V-BAT operations. Poland has also ordered the platform for maritime missions.

Those customers tell us different things. Ukrainian use tests whether V-BAT survives a hostile battlefield. Greece shows that another military found enough value to buy more. Dutch adoption goes further by integrating the aircraft into naval operations and ship infrastructure.

V-BAT is also entering recurring ISR-service contracts instead of relying only on aircraft sales.

The remaining question is scale: Shield AI has proved that V-BAT can get deployed; we still do not know whether that installed base will grow into hundreds of aircraft.

Are Skydio and Anduril actually fielded by the U.S. Army now?

Skydio and Anduril have both moved beyond ordinary Army demos, although their publicly visible deployment record still looks more institutional than combat-heavy.

Skydio has the cleaner case. The Army sent X10D drones directly to the 101st Airborne Division under an urgent operational request and said the systems were delivered within a month. The service later described X10D as entering initial fielding and opened the way for additional procurement through its Short-Range Reconnaissance program.

Anduril’s Ghost X is following a similar path. U.S. Army and National Guard formations have incorporated the drone into reconnaissance units, and Ghost X has appeared repeatedly in Transformation in Contact activity where units are reorganizing around new drones and sensing systems. The Pentagon also selected Ghost X under its Replicator effort.

The newer Army drone marketplace makes this transition easier. Units can train on aircraft such as Ghost X, compare them with alternatives and then purchase what fits their mission.

Neither company currently has the same publicly visible combat pattern as Quantum Systems or some Ukraine-focused manufacturers. Still, calling Ghost X or X10D “demo drones” today would undersell what has happened. Both are being absorbed into real Army formations.

Are Neros and Helsing scaling military drones faster than larger robotics companies?

Neros and Helsing are moving at a completely different scale because small strike drones can jump from battlefield validation to orders of several thousand units almost immediately.

Neros first supplied Archer FPV drones into the Ukraine ecosystem. It then secured a 6,000-drone order through the international Drone Coalition. During that ramp, Defense News reported that Neros was producing roughly 1,500 Archer drones per month from its Los Angeles facility.

The U.S. Army subsequently selected Archer for its Purpose-Built Attritable Systems effort. More recently, Neros announced a U.S. order worth more than $100 million covering 14,000 drones and payloads following Pentagon testing.

Helsing has taken a similar route in Europe. It announced an initial 4,000-drone HF-1 program for Ukraine and followed it with 6,000 HX-2 strike drones. The company now says thousands of HX-2s are deployed there. Deutsche Welle has independently reported from Ukrainian crews operating HX-2 systems on the frontline.

These numbers dwarf most larger military robot fleets. ARX is measured in hundreds of UGVs. Milrem’s major Ukrainian expansion exceeds 150 THeMIS vehicles. Forterra sent 105 Lancers. The Netherlands is acquiring 12 V-BATs.

System	Publicly visible scale	Type
Quantum-linked interceptor program	15,000 initial systems, with further expansion planned	Interceptor drone
Neros latest U.S. order	14,000 drones and payloads	FPV / strike drone
Helsing HX-2	6,000-unit announced order	Strike drone
Neros Ukraine coalition order	6,000	FPV drone
Quantum/Tencore TerMIT	2,000 initial systems	Ground robot
Quantum Vector in Ukraine	500+ previously delivered	ISR drone
ARX GEREON	Hundreds deployed or under contract	Ground robot
Milrem THeMIS expansion	150+ additional systems	Ground robot
Forterra Lancer	105 delivered to Ukraine	Autonomous ground vehicle
Dutch V-BAT fleet	12 aircraft	VTOL ISR drone

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

Are ground robots finally becoming normal battlefield equipment?

Ground robots are finally becoming useful battlefield equipment at meaningful scale, particularly for logistics, casualty evacuation, reconnaissance and other missions that keep soldiers away from exposed routes.

ARX Robotics is the clearest recent example. Its GEREON fleet in Ukraine has expanded several times, with hundreds of systems now deployed or contracted. The vehicles are being used for jobs such as resupply, casualty evacuation and support in dangerous areas.

The industrial side has caught up quickly. At the end of September, ARX and engine manufacturer DEUTZ delivered the first GEREON systems from DEUTZ’s Ulm plant only 12 weeks after announcing the production ramp. Those vehicles are heading to Ukraine to reinforce an existing fleet.

ARX has also formed a joint venture with Ukrainian manufacturer Roboneers to produce the Rys Pro UGV. The partners say they want to reach thousands of units in the first year and eventually tens of thousands annually. Those remain targets, so we should not count them as deployment.

Milrem Robotics gives us a longer track record. THeMIS vehicles have operated in Ukraine since 2022 across logistics, engineering and other support roles. An initial fleet of around 15 is now being followed by a program for more than 150 additional systems.

Forterra adds an American case. The company says it produced and delivered 105 Lancer autonomous vehicles to Ukraine in roughly six months. It is also supplying autonomy for larger U.S. military vehicles, including the Marine Corps’ ROGUE-Fires program.

Ground robots carrying ammunition or evacuating casualties are already becoming much less exotic.

Which ground-robotics startups are still mostly being tested rather than routinely used?

Overland AI remains one of the clearest examples of a technically serious military robotics startup that is still much closer to large-scale Army experimentation than routine fleet deployment.

Overland has done far more than a trade-show demo. U.S. soldiers have operated ULTRA autonomous vehicles during demanding Army exercises. The company works closely with Army research organizations and has won agreements under programs designed to push autonomous mobility into real formations.

Yet the Army itself still describes much of this work as prototype evaluation. Soldiers test the vehicles, generate data and help the service decide what should eventually be bought.

That puts Overland at a different stage from ARX in Ukraine, where hundreds of ground robots are already tied to recurring operational missions.

Forterra shows why company-level labels can be misleading. Some of its Army autonomy work remains experimental, while its Lancer vehicles have already gone to Ukraine in triple-digit numbers and its ROGUE-Fires work has entered production.

Overland is currently close to crossing the gap. What we still need to see is a sizeable operational fleet that stays with units after the experiments end.

Is Saronic already moving autonomous warships into the U.S. Navy?

Saronic has just made one of the biggest jumps in military robotics: the U.S. Navy has moved its 180-foot Marauder from testing into production.

Marauder went through an eight-day Navy evaluation involving more than 140 scenarios and continuous operations. Tests included autonomous maneuvering around other vessels as well as passive tracking without relying on active radar.

The Navy then awarded Saronic a production agreement under its Medium Unmanned Surface Vessel Marketplace. Navy information reported alongside the award indicates Saronic, HII and Galliano Marine Service are each expected to produce ten MUSVs, with an average procurement cost of roughly $40 million per vessel across the program.

Saronic was already building additional Marauder hulls before the production decision. The company says the first vessel has accumulated hundreds of hours of on-water testing and that more hulls are moving through production.

Its smaller Corsair gives Saronic some earlier operational experience, but Marauder is a much harder test. A 180-foot unmanned ship has to survive long missions, maritime traffic, maintenance cycles and Navy integration.

For now, Saronic sits just below companies with mature operational fleets. The Navy has committed to production; recurring fleet missions are the next proof point.

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

Is Saildrone further along than newer autonomous warship startups?

Saildrone is further along operationally than almost any newer maritime robotics startup because its vessels have already spent years doing persistent naval surveillance missions.

The U.S. Navy began operating Saildrone Explorers extensively in the Middle East several years ago. Fifth Fleet later reported more than 25,000 cumulative hours of unmanned operations across its regional experimentation, and individual Saildrones have stayed at sea for months at a time.

One episode captured how operational the system had become: Iranian forces attempted to seize a U.S.-operated Saildrone in the Arabian Gulf.

Saildrone’s larger Voyager has since supported additional Navy missions, including persistent intelligence, surveillance and reconnaissance work.

Saronic may eventually build a much more consequential autonomous combat fleet because Marauder can carry larger payloads and operate as a proper naval vessel. Today, however, Saildrone has the deeper record of sustained operations.

Are deployed military robots actually autonomous yet?

Most military robots currently being deployed use useful chunks of autonomy, but humans still control or authorize the most important parts of many missions.

Ground robots make this especially clear. GEREON and THeMIS can use autonomous navigation and other software assistance, yet some of their most valuable battlefield jobs remain straightforward: carry supplies, evacuate wounded soldiers, transport equipment and keep people off dangerous roads.

Strike drones are moving faster toward onboard autonomy. Neros has added terminal guidance that can continue the final attack after an operator designates a target. Helsing says HX-2 uses onboard software to navigate and reacquire designated targets when communications or satellite navigation deteriorate.

Shield AI pushes further with Hivemind, its autonomous-pilot software. V-BAT can increasingly handle mission functions without continuous human control, particularly when communications become unreliable.

Forterra and Overland are doing similar work on the ground, where autonomous driving through rough terrain without dependable GPS is genuinely difficult.

Curiously, higher autonomy does not guarantee faster deployment. Fairly simple teleoperated UGVs are already working in large numbers in Ukraine while some technically ambitious autonomous U.S. vehicles remain in Army evaluation programs.

Militaries currently care more about whether the robot completes its mission than whether the company can call it fully autonomous.

Is electronic warfare becoming the real test for military drones?

Electronic warfare has become one of the biggest separators between drones that look good during demonstrations and drones that remain useful in an actual war.

Ukraine has forced manufacturers to assume that GPS may disappear, command links may be jammed and the enemy may start adapting to a new drone shortly after it appears.

Shield AI points to V-BAT missions in Ukraine where the aircraft has operated through GPS and communications disruption. Helsing designed HX-2 around a similar problem, using onboard computing for navigation and terminal targeting when external links become unreliable. Ukrainian crews have now used the system in frontline conditions.

Neros is moving in the same direction with Archer. Its newer autonomous terminal-guidance functions reduce the aircraft’s dependence on an uninterrupted control link during the final attack.

Quantum Systems has spent years modifying Vector around Ukrainian electronic-warfare conditions. Its expansion into interceptor drones also reflects how quickly the environment has changed: cheap drones increasingly need to detect or destroy other cheap drones.

More decision-making is therefore moving onto the aircraft because manufacturers can no longer assume the network will always survive.

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

Why are reconnaissance and logistics robots getting deployed faster than robotic tanks?

Reconnaissance and logistics robots are getting deployed faster because commanders can use them immediately without asking the machine to solve every hard battlefield problem at once.

Quantum Systems’ Vector finds targets and provides surveillance. Shield AI’s V-BAT performs longer-range ISR. Skydio X10D and Anduril Ghost X help small units see farther. Saildrone watches large stretches of ocean.

On the ground, ARX’s GEREON and Milrem’s THeMIS are heavily used for logistics, casualty evacuation and support work. Those missions matter enormously when artillery, mines and drones make a short supply run dangerous for soldiers.

Small strike drones sit somewhere in between. Neros Archer and Helsing HX-2 are weapons, but their job remains narrow. They fly toward a designated target and attack it. They do not need to behave like an autonomous armored crew capable of interpreting every situation on the battlefield for hours.

Commanders are therefore adopting robots one dangerous human task at a time. That is proving much easier than fielding an autonomous tank.

Which startups can turn Ukraine deployment into repeat NATO orders and real production?

Quantum Systems, Shield AI, Neros and increasingly ARX show the clearest path from Ukrainian battlefield validation into broader Western procurement, while their manufacturing strategies suggest they can support more than one-off deployments.

Quantum Systems has already passed that test better than most. Vector went from Ukrainian combat use to contracts with Spain, Romania and the U.S. Army. The company now manufactures and repairs systems in Ukraine, builds aircraft elsewhere in Europe and increasingly works through joint ventures.

Shield AI has followed another version of the same route. V-BAT’s Ukrainian experience now sits alongside operational adoption by Greece and the Netherlands and procurement from Poland.

Neros may be the most striking younger example. Archer was shaped heavily by what Ukrainian operators needed from FPV drones. The company then won a U.S. Army selection and, more recently, a 14,000-system U.S. order. During an earlier production ramp, Neros was already building roughly 1,500 Archer drones per month.

ARX is pushing Ukrainian lessons back into Western armies while expanding its industrial base. GEREON has participated in major allied exercises, DEUTZ is now manufacturing the vehicle in Germany, and ARX has partnered with Roboneers to expand Ukrainian and European production.

Helsing has taken a more vertically integrated approach. It built dedicated production capacity around high-volume autonomous weapons and then announced 6,000 additional HX-2 systems for Ukraine.

Forterra’s quantities are smaller, but producing more than 100 autonomous tactical vehicles in roughly half a year is mechanically far harder than assembling FPV drones.

Saronic faces the biggest manufacturing challenge of the group. Marauder is 180 feet long. The company already has multiple hulls moving through production and now has a Navy production award.

Ukraine accepts rapid iteration and frequent equipment losses because it is fighting an active war. NATO militaries usually demand more documentation, longer support commitments, cybersecurity and integration with existing systems. Companies that can satisfy both environments are the ones most likely to turn wartime deployment into durable defense businesses.

Company	Current scaling evidence	What we can reasonably conclude
Neros	Roughly 1,500 Archer drones/month during earlier ramp; 14,000-system U.S. order	High-volume drone production is real
Helsing	Dedicated factory plus 6,000 HX-2 order	Production built around wartime quantities
Quantum Systems	Ukrainian manufacturing plus multi-company joint ventures	Distributed production is becoming a core advantage
ARX Robotics	DEUTZ line producing GEREON after a 12-week ramp	Heavy UGV output can scale faster than traditional programs
Quantum/Tencore	2,000-unit initial TerMIT program	Four-digit Western UGV production is becoming plausible
Forterra	105 Lancers delivered in roughly six months	Meaningful tactical-vehicle production already demonstrated
Milrem Robotics	150+ additional THeMIS vehicles planned for Ukraine	Mature UGV moving into a much larger fleet
Saronic	Multiple Marauder hulls under construction plus Navy production	Autonomous shipbuilding has crossed into serial production

Which military robotics startups are really moving from demos to deployment?

The strongest deployment cases today are Quantum Systems, Shield AI, ARX Robotics, Milrem Robotics, Neros, Helsing and Forterra, while Skydio, Anduril, Saildrone and Saronic show that the transition is spreading into mainstream U.S. and allied forces.

Quantum Systems has built the deepest bridge from Ukrainian combat experience into broader allied procurement. Shield AI has turned V-BAT into an operational aircraft across several militaries. ARX has taken Western UGV deployment from dozens of systems toward hundreds. Milrem has the longest-running Western UGV combat record in the group.

Neros and Helsing show where the numbers can go next. Their strike-drone orders are already measured in thousands and, increasingly, tens of thousands. Forterra proves that heavier autonomous ground vehicles can move beyond one-off prototypes too.

Skydio and Anduril deserve a slightly different label. Their U.S. Army adoption is real, but we have less public evidence of long-duration combat use than we do for systems shaped by Ukraine. Saildrone has the opposite profile: years of genuine operations, but in a narrower persistent-surveillance role.

Saronic is the company whose position has changed most recently. Marauder has now passed Navy testing and entered production, so its next test is whether those ships become a reliable operational fleet.

Across all of these companies, the systems advancing fastest solve a clear military problem today: finding targets, moving supplies, evacuating wounded soldiers, patrolling water or delivering a cheap strike.

When soldiers use the first robots, ask for more and keep using the replacements, the demo phase is over.

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