One operator, two hundred machines. That is the ratio Elbit Systems put on the table on September 9, when its FUSE unit gathered the company's autonomous drones and ground vehicles under a single software roof and argued that the hard part was never the hardware.
The claim runs against a decade of defence-robotics marketing. Airframes and ground vehicles have become a crowded trade. Making hundreds of them behave as one formation, in a place where satellite navigation has been switched off, with a human still approving targets, has not.
The unit reports more than 1,000,000 operational flight hours and over 100,000 autonomous ground hours across platforms already delivered.
The investment case is a margin shift, from the platform to the software layer that commands it.
The coordination problem Elbit is selling
FUSE sits inside Elbit's C4I & Cyber division and folded at least four related activities into one business. At its centre sits Dominion-X, an AI-powered operating system that coordinates aerial and ground systems in real time, shares mission data across a common network, and keeps running when GNSS (global navigation satellite systems) is jammed or unavailable.
One2Many is the operating concept. A small crew supervises a distributed formation rather than one operator per airframe. Elbit says the architecture supports manned-unmanned teaming, with autonomous assets working alongside crewed platforms inside existing command structures instead of a parallel chain.
The maturity argument leans on hours, not slides. Elbit reports more than 1,000,000 operational flight hours and over 100,000 autonomous ground operation hours across the portfolio, with thousands of drones and hundreds of ground platforms already delivered. Those are the company's figures, released at launch, and they are why FUSE describes Dominion-X as fielded rather than demonstrated.
Autonomous flight-hours logged by the portfolio
Elbit reports the figure at the FUSE launch, alongside more than 100,000 autonomous ground operation hours. · Elbit Systems, 2026
A portfolio dressed as a business unit
The launch grouped a wide set of systems. Unmanned Tactical Force covers recon-strike formations. Pirani-X handles seek-and-strike swarm missions for manoeuvre forces. Armored Wingman pairs ground and airborne assets with crewed vehicles. A group 1 and 2 uncrewed aircraft family, spanning X-Intra, THOR and Magni-X, covers reconnaissance, communications relay and resupply. Ground Autonomy Kits add autonomous off-road driving and teleoperation to vehicles Elbit does not build.
That breadth matters for procurement. Armies rarely buy a swarm. They buy a programme, then extend it. By wrapping drones, ground vehicles and retrofit kits into one line, FUSE offers buyers a single integration point and gives Elbit repeat revenue as those formations grow.
Scale is the other half of the pitch. Elbit reported $2,287.1 million in revenue for the three months to June 30, 2026, and an order backlog of $32.0 billion. A unit that converts an existing backlog into software-attached upgrades can grow without waiting for new platform wins.
Elbit order backlog as of June 30, 2026
The parent enters the autonomy push with a record book of orders and quarterly revenue of $2.29 billion. · Elbit Systems, 2026
Military forces are entering a new operational era in which combat effectiveness will increasingly depend not only on autonomous platforms themselves, but on the ability to integrate and coordinate large numbers of autonomous systems within existing maneuver forces in real time.— Eyal Dahan, CEO of FUSE
Dahan's framing is the tell. The unit is selling the connective tissue between drones, vehicles and the people who command them, not another airframe.
What One2Many changes in practice
Where the software margin sits
Defence primes have spent years buying autonomy startups to avoid being disintermediated by software. Anduril markets Lattice as the layer above hardware. Shield AI built Hivemind to fly aircraft it did not manufacture. Helsing sells AI into European platforms. Elbit's move is the incumbent's version of the same bet, made with a manufacturing base and a delivery record its challengers lack.
As we wrote in September, the sharpest bet in this market belonged to a company with no hardware at all, an AI brain that commands other firms' machines. Elbit is betting the opposite: that the incumbent with the factories, the integrators and the backlog captures the orchestration layer first.
The economics favour whoever controls the interface. Software attached to a delivered platform carries gross margins no airframe can match, and every new platform that joins the network raises switching costs for the customer.
What has to hold for the thesis to pay
Three things. Procurement has to keep funding multi-system programmes rather than point purchases. Dominion-X has to stay interoperable as allied countries field their own autonomy stacks. And the human-in-the-loop requirement, which Elbit says it enforces for targeting, has to remain a design constraint rather than a brake on the tempo it is selling.
Elbit earned the right to make the argument. FUSE runs with 160 staff, about two thirds of them engineers at its main site, against the parent's 21,000 employees. That is the point. It is a bet that a focused software organisation can pull a sprawling hardware portfolio onto one operating system before a rival does.
Will orchestration software become the defence-robotics profit pool?
Probability: 55%. The hardware is converging, so differentiation migrates to coordination, and accounting tends to follow the margin.
✅ Arguments for
Software revenue is recurring, higher-margin and sticky once a fleet standardises on one operating system.
Elbit's backlog and manufacturing scale let it bundle software into contracts it already expects to win.
Confirmation criteria: a prime discloses software or autonomy revenue separately, or a multi-nation programme names an orchestration layer as a deliverable.
❌ Arguments against
Export rules and interoperability mandates can slow adoption of any one proprietary stack.
Primes have bundled software into hardware contracts for decades without ever breaking it out.
Disconfirmation criteria: competing stacks fragment procurement and buyers keep paying per platform rather than per fleet.
Development scenarios
🟢 Optimistic scenario (25%)
Implications: the incumbent primes re-rate as software businesses with a hardware tail.
🟡 Base-case scenario (50%)
Implications: the strategic value is real, the margin uplift is gradual, and disclosure stays fuzzy.
🔴 Pessimistic scenario (25%)
Implications: autonomy remains a feature of hardware rather than a business of its own.
Whether Elbit breaks out autonomy revenue or contract awards tied to Dominion-X.
Whether allied forces standardise on one autonomy stack or keep buying point platforms.
Whether rival primes match Elbit by reporting autonomy software separately.
Whether the human-in-the-loop rule constrains tempo or settles into a stable design norm.