Dan Keene spent years building a battery-swapping network for shared scooters. The operators who survived that market, he writes, were the ones who could charge, repair and reposition tens of thousands of vehicles a day. On Saturday he applied the same lesson to robotaxis in The Robot Report. His company, Aseon Labs, builds robotic stations that charge, clean and inspect autonomous cars, so he sells the fix for the problem he describes. His arithmetic holds up anyway, and it gives any founder with hardware in the field a way to count the work a demo leaves out.
What happens between rides
Keene starts from Waymo's own numbers: more than 220 million autonomous miles, around half a million paid rides a week, 14 cities. Uber operates in more than 15,000. When a Waymo car finishes a ride with a spilled latte on the back seat, it can drive up to 15 miles, empty, to a central depot to be cleaned, charged and inspected. He puts each of those resets at roughly two hours. Of the 86 million Waymo miles reported in California through 2025, 54% carried a passenger, according to state data he cites. Depot runs are one source of those empty miles, and he is clear they aren't the only one.
The driving is autonomous. The servicing, in Keene's telling, is manual, and it happens in a building the operator had to find, lease and wire for power. That building is where the cost of each new city piles up. He cites California's regulator targeting six months just to energize a standard charging site, 1.9 years for a new circuit, 2.8 years for a substation upgrade and 8.9 years for a new substation. Those clocks start over in every city, and Phoenix, Zurich and Riyadh each need a different depot.
Count the touches per unit of work
The same count works for anything that does repeated work in the field. Pick the unit your customer pays for: a ride, a procedure, a pallet moved, a sample run. Then list every time a person or a machine has to handle the product before it's ready for the next unit. Charging, cleaning, inspection, calibration, restocking consumables, clearing a fault, moving it back into position. For each touch, write down who does it, where it happens and how long it takes. That list is your service loop.
A demo shows none of this. It runs one unit, on a charged machine an engineer prepared the night before. Keene observes that every robotics demo ends with someone off camera carrying in a charging cable. That person belongs in the financial model, with a wage, a drive time and a building to work from.
Surgical robotics runs the same loop, and hospitals count it. After every case the instruments go to sterile processing and the room gets turned over for the next patient. A robot that adds, say, twenty minutes of turnover to each case can cost a busy room its last case of the day, and the hospital's value analysis committee will run that math whether the company ran it first or not.
The second number is time per site
The touch count sets your margin. Time to stand up servicing at a new site sets how fast you can grow. Keene's answer is a standard servicing unit that arrives on a flatbed and goes live within a day. Whether his version works is open, and he says himself that his company is early. The principle under it holds for any fleet: if each new site needs custom construction, permits and a utility upgrade, growth moves at the speed of the slowest permit office, and more capital won't speed it up.
Waymo has already handed part of its loop to other companies. Zeekr, a carmaker, builds its vans, and four outside operators (Moove, Avomo, Avis and Flexdrive) run its depots. Deciding which touches you own, which you contract out and which you design out of the product is a product decision. A sensor that keeps itself clean removes a touch. A battery that swaps in minutes shortens one. The cheapest place to make those calls is the design review, while the service loop is still a drawing and a change costs an engineering revision. After launch the same change costs a lease.
Dave's take
When I review a hardware roadmap, I ask to see the service loop drawn next to the product architecture, with a name and a time on every touch. The economics of a machine in the field have to work on the first site, and the loop is most of that math. A team that can't draw that loop is still selling the demo.
From Dave’s video library
Dave on why a working demo isn't demand, and why a paying buyer is the clock that decides when a build is ready.
I’m here to help you scale.
Work With DavePrefer a smaller first step? Book a $500 one-hour working session →
Dave Saunders is the founder of Base Reality Group and a Fractional CPO for product companies. He was a founder and operator at Galen Robotics, where the surgical-robotics platform earned FDA De Novo authorization in 2023, and he managed a 35-patent portfolio licensed from Johns Hopkins. He wrote Founders Who Finish and publishes The Build. More about Dave →