In 2012, a worker at a Pennsylvania power plant died doing a rope inspection 100 feet up. The boiler had been shutting down 30% of the year. Jake Loosararian visited the plant the next year and came away with one question: why does a human do this by hand, one square foot at a time?
The same robots, sensors and AI platform that check Navy hulls also read power plants, refineries and pipelines. Gecko says its data suggests US thermal plants could add 15 to 20 gigawatts (GW) of capacity without building anything new.
Gecko is opening a manufacturing facility in Sewickley, Pennsylvania to build components for the defense industrial base. It is the clearest sign yet that inspection robotics is becoming a procurement category, not a novelty.
The context is a Navy that cannot get its ships out of drydock fast enough. Two of every five vessels sit in port for maintenance. The goal set by the Chief of Naval Operations is 80% fleet readiness by 2027. The contract, awarded in March 2026 through the General Services Administration, starts with 18 ships in the Pacific Fleet.
TIMELINE: Gecko Robotics
─────────────────────────────────────────────────────────────
2013 ──── 2016 ──── 2022 ──── 2026-03 ──── 2026-07
🔬 🧪 🏭 📜 🔥
Founded YC Navy $71M Navy Sewickley
in a batch surface IDIQ manufacturing
power intake fleet contract facility
plant work
Source: Gecko Robotics, Defense One, TBPN Digest (2026)
The contract: 18 ships, one ceiling, a new procurement category
The deal is an indefinite-delivery, indefinite-quantity contract (IDIQ) with a $71 million ceiling and an initial award of $54 million over five years. Every military service can buy through it. It starts on destroyers, amphibious warships and littoral combat ships.
Gecko Robotics Navy contract
Five-year IDIQ, $54M initial award, $71M ceiling. 50x faster repair identification. A single flight-deck evaluation eliminated over 3 months of maintenance delay. · Gecko Robotics, March 2026
The robots do the inspection work humans used to do on ropes and scaffolding. Magnetic wheeled units scale hulls to map corrosion. A flatter model called Komodo rolls across flight decks like a paint roller, scanning non-skid surfaces and flagging which areas need work. The data feeds a laptop map, not a paper report.
What matters to the Navy is the timing. Robots and sensors inspect a vessel before it ever reaches drydock, so the yard knows what it is fixing before the ship arrives. Gecko's co-founder and president Troy Demmer described it as understanding "the material health of this asset" before the dock empties.
We're leveraging autonomy and AI to deploy technologies prior to a ship or a submarine coming into a dry dock facility so that we have an understanding of, what's the material health of this asset? How do we make targeted, efficient repairs so that we can get that asset back in the fight?— Troy Demmer, Co-founder and President, Gecko Robotics
The moat is the data platform, not the robot
The wall-climbing robot is the visible part. The durable part is Cantilever, the software platform that stores 13 years of structural health data across half a million assets.
Every ultrasonic reading, every corrosion map, every weld scan goes into the same system. That dataset is what lets its AI predict failures instead of reporting them. The company calls it a single source of truth for the physical world. The Navy's Justin Fanelli frames it as cracking the cost equation.
Where value hasn't improved, that's where opportunity lives. Cracking the cost equation is just as important as cracking the physics equation.— Justin Fanelli, Chief Technology Officer, Department of the Navy
The origin story explains the data obsession. Bootstrapped for three years, with Loosararian sleeping on a friend's floor in Pittsburgh, the company went through Y Combinator in 2016. Rejecting advice to build the product in a lab and launch autonomously, he stayed close to the independent power producers who actually feel asset failure in their margins.
Two markets, one dataset: energy pays the bills, defense scales the model
The company's own framing is straightforward: energy is the foundation, defense is the growth vector. A refinery or power plant facing unplanned downtime loses an estimated $30 million per day. Independent power producers lack the cost-pass-through of regulated utilities, so an asset failure is a direct hit to their income.
Same sensors. Same AI. Same Cantilever platform. Energy customers pay for reliability, defense customers pay for readiness. Each contract improves the model the other side uses.
Most defense robotics companies build for one buyer. It built for the hardest civilian market first, plants where a shutdown costs millions a day, and the Navy contract then redeployed that capability. The reverse order is far more common and far less durable. For investors, that is the pattern to study.
The 15-20 GW claim: more power without new plants
Loosararian's interviews point to a bigger number than the Navy contract. Better data on existing US thermal plants, he claims, could add 15-20 GW of capacity without building a single new facility.
What would 15-20 GW actually mean?
Why it matters: capacity added through utilization costs a fraction of new build, with no permitting, no interconnection queue, no multi-year construction.
What could go wrong
The ceiling is not the award. An IDIQ with a $71M ceiling is an upper bound; the initial order is $54M, and contract value only materializes if the Navy keeps tasking Gecko. The 15-20 GW figure is self-reported and not yet independently audited. And Gecko is a company that sells its own data. The same platform that finds defects also proves the platform finds defects.
Whether the Pacific Fleet task orders actually grow past the initial 18 ships, and whether any independent operator publishes a verified capacity gain from Gecko's plant optimization data.
There is also the broader shift it sits inside: robotic inspection is moving from pilot to procurement. As we wrote in August, State Grid's order of 8,500 inspection robots showed the scale the category can reach on the energy side. The Navy deal is the defense mirror of that same curve.
Signals worth tracking
Task-order growth on the Navy IDIQ beyond the initial 18 ships
Independent verification of the 15-20 GW thermal capacity claim
Sewickley facility output, whether it becomes a shared defense-industrial hub
Whether other primes license Cantilever-class data platforms instead of building their own
Gecko is 13 years in, bootstrapped, profitable by its own account, and now holding one of the Navy's flagship inspection contracts. The robot that climbs a boiler wall in Pennsylvania and the robot that scales a destroyer hull in the Pacific are the same machine. That is the entire thesis in one image.
