The San Francisco startup had sensors that cut building power bills by a third. It walked away to build swarms of autonomous submarine drones instead. The reason: submariners kept describing a gap nobody was filling.
Most founders pivot to escape a problem. AICE pivoted toward one. The company had built something that worked - plug-and-play sensors that snapped onto a building's equipment in hours and used AI to shave energy bills by roughly a third. Customers were happy. Then the two founders left it on the table and went to build autonomous submarine drones. The decision sounds reckless until you hear what changed their minds.
AICE (YC P26) now describes itself in one blunt line: swarms of autonomous subsea drones for the monitoring and protection of critical infrastructure. Translated, that means small underwater robots, working together, watching the parts of the world almost nobody watches - the cables, pipelines, ports, and naval assets sitting on the ocean floor. The company is based in San Francisco. The drones are designed and built in France.
The story starts on land. AICE's first product was an energy play - sensors that installed at equipment level with no power interruption and no heavy capital outlay, then started finding waste the next day. The pitch when the company launched on Y Combinator was tidy: cut a commercial building's energy bill by 20 to 30 percent. That is a real business.
What moved them off it was conversation. The founders started talking to submariners and defense people, and the same complaint kept surfacing: the tools for watching what happens underwater are expensive, rigid, and scarce. There is a lot of critical infrastructure down there and very little cheap, flexible hardware guarding it. AICE decided that gap was bigger than the one inside an office building.
Pivots usually carry a whiff of failure - the founder retreats from a market that rejected them. This one runs the other way. The energy product had shipped, installed, and delivered savings. Leaving it was a choice about opportunity, not survival, and that distinction matters. It is easy to be stubborn about an idea. It is harder, and usually smarter, to be stubborn about the team and flexible about what that team builds. AICE kept the founders and swapped the mission.
AI energy sensors that cut commercial building power bills by up to ~30%.
Modular subsea drone swarms for maritime defense and infrastructure monitoring.
The insight is boring and huge at once: the hardest infrastructure to protect is the infrastructure you can't see.The AICE thesis, in plain terms
The core idea is modularity. Instead of building one perfect, expensive underwater vehicle that does a single job, AICE builds a modular unmanned underwater vehicle - a UUV - whose configuration can be reshaped for the task. Speed, range, sonar, payload: swappable. Then you deploy them in coordinated swarms rather than sending down one costly machine and hoping it is in the right place.
The drones are designed to work in the band where most of the interesting infrastructure sits, roughly 100 to 600 meters down. Deep enough to matter, shallow enough to be a practical patrol zone. Their three jobs are simple to state and hard to do: surveillance, detection, protection.
Modularity is not just an engineering preference here; it is the business case. A rigid, single-purpose vehicle forces the buyer to purchase a new machine every time the mission changes. A body that accepts different sonar rigs, sensors, and payloads lets one platform stretch across jobs that would otherwise need several. That is how you drive down the cost per deployment, and cost per deployment is the number that decides whether a navy can afford to watch the seabed continuously or only occasionally.
Where the work is: cables, pipelines and coastal assets cluster in the mid-depth band AICE targets, not the abyss.
Gregoire Chomette, co-founder and CEO, has an unusual resume for a defense hardware company: he worked at NASA on asteroid impact risk and holds a master's in aeronautics and astronautics from MIT, class of 2021. His old job was estimating the odds of a rare, catastrophic event from space. His new job is estimating threats from below the waterline. The through-line is the same - model the low-probability, high-consequence event before it happens.
His co-founder, Remi Bouteiller, comes from competition robotics as a former European Robotic Champion. Between them the company pairs space-systems rigor with hands-on robotics that has been tested under a clock. It is a small team, which in this market is less a weakness than a filter: underwater defense is slow, capital-heavy, and unglamorous, and most well-funded startups won't touch it.
AICE was stubborn about the founders, not the idea. That is why the pivot worked.On why they could walk away from traction
The undersea domain has quietly become contested. The physical backbone of the internet runs along the seabed, and the pipelines and ports that move energy and goods sit in the same water. Most of the world's data traffic crosses the ocean floor at some point, riding cables that are, physically, not much thicker than a garden hose. When one is cut - by an anchor, by accident, or by intent - the disruption ripples across continents. Yet the sheer length of that infrastructure makes it almost impossible to guard with the tools navies use today.
The incumbents building underwater vehicles - large defense primes and specialist robotics firms - tend to ship expensive, single-purpose systems. AICE's wedge is to attack the thing that industry treats as a fixed cost: the price and rigidity of the hardware itself. Companies like Anduril, Saab, Kongsberg and Teledyne operate in and around this space, and the lineage of consumer-grade underwater robotics reaches back to projects like OpenROV. AICE's answer is not to out-engineer a single flagship vehicle but to change the unit of competition from the vehicle to the fleet.
| Approach | Typical incumbent UUV | AICE |
|---|---|---|
| Configuration | Single-purpose, fixed | Modular, swappable |
| Deployment | One costly vehicle | Coordinated swarm |
| Cost posture | High, bespoke | Lower, repeatable |
| Adaptability | Slow to reconfigure | Reconfigure per mission |
The bet is that many good-enough drones, coordinated, beat one expensive genius machine. That is a strategy as much as a spec sheet. It also explains the manufacturing choice: building in France gives the company a defense-adjacent industrial base while it operates from San Francisco and the Y Combinator network.
There is a second effect to the swarm approach that is easy to miss. A single vehicle is a single point of failure - lose it and the mission ends. A swarm degrades gracefully; lose one drone and the rest keep watching. For infrastructure that has to be monitored around the clock, coverage that survives the loss of any one unit is worth more than the peak performance of a lone machine. That is the kind of trade-off you only appreciate after talking to the people who actually run these operations, which is exactly where AICE's roadmap came from.
AICE is seed-stage - a $130,000 round closed in early 2026 through Y Combinator's P26 batch - and the defense line is still young. The intended buyers are navies, defense agencies, and the operators of undersea critical infrastructure. The revenue model that comes with that territory is government and defense procurement: long sales cycles, high stakes, and a premium on hardware that actually survives the environment it is dropped into.
None of that is fast. But the company's most interesting quality may be its willingness to do the slow thing. AICE had the option of a comfortable energy-software business and chose a market that is harder to enter, harder to sell into, and harder to build for. If there is a moat here, part of it is simply that most people won't attempt this.
The expertise question is where the founders' backgrounds stop being a curiosity and start being a fit. Building an autonomous system that operates unattended at depth is a problem of sensing, navigation, and decision-making under uncertainty - close cousins of the aerospace and robotics work both founders already know. Chomette's asteroid-risk work at NASA was, at its core, an exercise in reasoning about a hostile environment you cannot directly observe. Bouteiller's competition robotics is the applied, deadline-driven counterpart. Underwater autonomy sits squarely between the two.
A swarm of good-enough beats a single expensive genius. That is the whole product philosophy in one sentence.On modular vs. bespoke
The name, oddly, is a fossil of the old plan. The company kept AICE and the domain aicepower.com from its energy days - a small reminder that the business you launch is not always the business you become. For a company now pointed at the ocean floor, "power" turned out to mean something other than kilowatt-hours.