The most interesting thing about an autonomous drone is not that it flies itself. It is that it stays.
A docked drone on an infrastructure site is a permanent aviation asset. It sits in a weatherproof station, charges itself, launches on an alarm or a schedule, flies a route no human has planned that morning, and waits. It has a certificate, a flight envelope, an insurance policy, and a maintenance cycle. And for most of any given day, it does nothing at all.
That idle capacity is what nobody in this industry talks about - and where the money is.
For a decade, the honest description of a commercial drone was "one job, expensively." You mobilized a crew and an aircraft, performed a task, invoiced, and drove home; every flight carried the full weight of getting there.
Autonomy is usually sold as a labor story — no pilot, no crew, no travel. True, and the least interesting part. The real shift: once an aircraft is permanently stationed, certified, and paid for by one budget line, every additional mission is nearly free.
Suddenly the question is no longer "can a drone do this?" It is "what else can this drone do while it's standing there?"
That question is operational and commercial, not technological. And it is being asked wherever these stations are going up — power grids, water systems, rail corridors, ports, industrial estates, and solar fields, from Israel to Iberia to Texas to the Gulf.
What that looks like in service
Over eight working days last July, one of our autonomous drones flew 69 sorties above an 88-megawatt solar field in southern Israel — not a technology demonstration but an operational deployment of a capability in service today. Its job was security: perimeter surveillance, fence-line monitoring, launch-on-alarm response. It passed a full security drill — perimeter breach, fence contact, suspicious object — with a response time no patrol vehicle can match.
It also carried a thermal camera that fed our solar-analysis platform.
By the end of those eight days, it had identified 154 defects. Forty-two were disconnected strings — rows of panels producing nothing while the site's monitoring system reported normal operation. Twenty-seven were repaired before we left, returning roughly $95,000 a year of electricity to the grid.
The 42 dead strings were costing about $147,000 a year. And that was the good news: thermal imagery of the same site 13 months earlier showed 168 disconnected strings — close to $590,000 of electricity financed, insured, and never sold.
The site manager watched the first day's map come up on a laptop in a portacabin. He had walked those rows every week for a year. He asked how long they had been dead, then stopped asking questions and started writing down string numbers.
Two operational problems. One aircraft. One cost base — because the flight was happening anyway.
Why the second mission was never economic before
The site was not badly run. It runs the way nearly every solar field in the world runs: SCADA reporting at the inverter level, an annual or semi-annual inspection, and a contractor doing his best with the information he has. The losses came to 0.47% of output — a rounding error on a spreadsheet and a fortune in the bank.
The scale of that rounding error is documented, and it is not an Israeli problem. The world passed 2,974 gigawatts of installed solar capacity at the end of 2025. Peer-reviewed research covering 3.3 million modules at more than 2,000 installations found that 36.5% carry a detectable thermal defect.
Industry data spanning 373 gigawatts of operating assets found equipment-driven power loss averaging 5.08% in 2025 — against a five-year average of 3.5% — with revenue exposure reaching $5,070 per megawatt per year. Losses are widening. When those analysts last published an industry-wide number, they put it at $4.6 billion a year; the global fleet has since more than doubled and the loss rate has risen.
The industry knows how to find these faults: aerial thermography under IEC TS 62446-3 is mature work. It happens once a year because inspection has always been an event: a crew arrives, flies for two days, invoices, leaves.
The report describes the site as it was on a Tuesday in March; everything that breaks in April is found the following March. The same dataset shows the cost: sites inspected quarterly outperform annual ones by 36%, and sites with stationed drones average 3% loss rather than 5%.
Continuous inspection has never closed as a business case: mobilization is the dominant cost, and you cannot mobilize monthly. But when the aircraft is already stationed — bought, installed, certified, and flying on the security budget — mobilization has already been paid for by someone else, for another reason entirely. The arithmetic inverts.
Which raises the obvious question: If the marginal mission is nearly free, what is anyone paying for?
Not the flight. The flight is the cheap part. What costs money — and what almost nobody has — is the layer turning pixels into a repair order: a module-level model of the asset, defect classification against it, and a calculation converting each fault into kilowatt-hours and money at that site's tariff. That is where the value sits, and why this has not happened already. The industry spent a decade buying aircraft when it should have been buying analysis.
What still has to be fixed
I would rather be candid than promotional, so let me name what did not work.
The drone found the faults. Closing them was harder. The contractor had not been trained on the analytics platform, so findings were relayed by phone and email instead of dispatched as tasks. The inspection system was not integrated with the site's SCADA, so an alarm could not automatically generate a flight. Both are solvable within weeks, and both are the difference between finding half a million dollars of loss and recovering it. Detection is not value; repair is.
There is a limit worth stating, because too much of this sector oversells it. The best published defect-classification accuracy in this field is 93%. Autonomous acquisition and machine-first triage are real today; a human reviewing exceptions remains in the loop, and anyone telling an asset owner otherwise is selling something.
The wider point
Israel has built a genuine drone industry — hundreds of operators, a civil aviation authority faster on beyond-visual-line-of-sight operations than most of Europe, and thousands of certified pilots - more than 2,000 of whom we trained ourselves. But nothing in the economics above is Israeli. The constraint, in every market I have worked in, is no longer technology or airspace. It is an imagination about what a stationed aircraft is for.
Security today. Inspection on the same flight. And wherever a stationed asset and a control layer already exist, the test that killed logistics, environmental monitoring, and emergency response as standalone drone businesses — one job, expensively — starts to come out differently.
At 0.47% of output on a well-run site, none of this is a headline problem. It is a quiet one. And quiet problems, compounded across three terawatts, are the expensive kind.
Itamar Kohali, Col. (Res.), is founder and CEO of Propeller Drones Ltd., Israel's largest drone services provider.