The infrastructure layer behind robotaxi economics.
Ride-hail removed the driver. It did not remove the depot. Here is what an autonomous fleet needs on the ground, in the order it needs it, and why that layer decides whether the vehicle earns.
The first generation of ride-hail worked because the infrastructure was already there. Drivers supplied the car, the parking spot, the fuel, the car wash, and the mechanic. The network supplied software and demand. It was capital-light for the network because the physical burden sat, invisibly, with thousands of individuals. An autonomous fleet inverts that. Every physical need a person used to absorb now has to be provided deliberately by a facility and a team.
Commercial autonomous ride-hail operators run dedicated depots for charging, cleaning, sensor calibration, inspection, and maintenance, staffed around the clock. This is visible in public reporting, permit filings, and the operators' own descriptions of their operations.
Autonomous vehicles in commercial service return to base between service blocks rather than remaining on the road continuously; a meaningful share of each day is spent charging, cleaning, or waiting.
Vehicle recovery, incident response, and remote assistance are standing functions at every operator, not exceptions.
The real per-vehicle cost of the depot layer, at scale, for a vehicle that is privately owned rather than operator-owned.
How much of that layer a manufacturer or network will provide to private owners, and how much will be the owner's responsibility.
The productive-hours ceiling for a purpose-built robotaxi once charging, turnover, and service are accounted for, which sets the revenue ceiling.
Whether shared, multi-owner depots will be treated by networks and insurers the same way as single-operator facilities.
Think of autonomous fleet infrastructure as a stack. Each layer depends on the one below it, and the economics of the vehicle are decided by how well the stack runs.
Land and building. A place where vehicles arrive, park, are worked on, and depart at all hours without conflict with neighbors, traffic, or zoning. This eliminates most candidate sites before anything else is considered.
Power. Enough electrical capacity to charge a meaningful share of the fleet at once, at a cost per kilowatt-hour the vehicle's revenue can support. The most underestimated layer, and the subject of Intel #03.
Charging sequence. Chargers are easy to buy and hard to run well. Vehicles should be topped up before peak, not during it, and never all at once in a way that trips the site's power ceiling.
Turnover. Cleaning, inspection, and the small resets between service blocks. On a shared vehicle this is the difference between a five-star ride and a complaint, and it has to be fast.
Service coordination. Tires, brakes, glass, sensors, body work, manufacturer service. Most happens off site. The operator's job is to batch it so one trip resolves several needs and the vehicle is out for hours, not days.
Recovery. A stopped vehicle is a live incident until someone reaches it. Response time depends on how close the depot is and how many vehicles one team covers.
Monitoring. Status of every vehicle, charger, bay, and open task. Without it the other layers run blind. With it, the operator can manage the only number an owner cares about: productive vehicle hours.
Administration. Insurance, registration, network compliance, incident documentation, owner reporting. Mandatory and easy to fall behind on at fleet scale.
Fleet infrastructure has to exist at neighborhood scale for private autonomous ownership to work at all.
The large operators are building this stack for their own fleets. What they are not building is a version a private individual with three vehicles, or a family office with forty, can plug into. That customer is the gap in the market, and that is the model SWARM is building: SWARM Hubs as the physical layer, SWARM Operations as the team, SWARM OS as the visibility layer. Owners bring the vehicles. The infrastructure is already there when they arrive.
SWARM Intel is published by SWARM, an independent autonomous fleet infrastructure company based in Austin, Texas. SWARM is not affiliated with any vehicle manufacturer or network operator. Confirmed statements reflect public reporting at the time of writing and should be verified against primary sources; nothing here is investment advice.