Nevada authorized roughly 8,000 robotaxis. The permit was the easy part.
On August 20 the Nevada Transportation Authority unanimously approved three driverless operations in Las Vegas. Tesla for up to 5,000 vehicles, Waymo for 1,000, and Aviari Services running on Uber for another 1,000, all inside twelve months (TechCrunch). Tesla had been sitting at a ten-vehicle interim cap in late July. Ten to five thousand in about three weeks.
Zoox is already carrying paying passengers. They started August 10 after a federal exemption for a vehicle built with no steering wheel and no pedals, allowing up to 2,500 a year for two years (Reuters).
This is my home market. I spent eight years at Tesla running sales, service, delivery and retail operations, including the Southwest, and I built the enterprise fleet relationships here with Hertz, Kaptyn and Las Vegas Metro Police. So I read these approvals a little differently than most of the coverage does.
The permit is the starting gun, not the finish line
Regulatory approval is a number on paper. It says you are allowed to put vehicles on the road. It says nothing about whether you can keep them there.
What decides that is unglamorous. Depot throughput. Charging capacity and how long a car sits on a charger versus how long it needs to. Cleaning cycle times, which nobody models correctly until they have seen what a Saturday night on the Strip does to an interior. How fast a vehicle gets back to earning after a curb strike. Whether you can hire and retain technicians in a market where every casino, airline and construction firm is competing for the same mechanical talent.
I moved more than 3,500 delivery vehicles a quarter against fixed windows, and ran service on 150 or more cars a month across in-shop, mobile and virtual channels. The thing that taught me is that capacity is almost never a software problem. It is bays, chargers, people and a site plan. You can have a perfect dispatch algorithm and still be capped by the fact that you have eleven charging stalls and a queue of thirty cars at 2am.
Utilization is an operations number
The metrics that determine whether any of this makes money are utilization and revenue per vehicle hour. Both are set by dispatch density and dwell time, and dwell time is an operations problem wearing a technology costume.
Every minute a vehicle spends not moving is a minute of depreciation with no revenue against it. That covers charging, cleaning, inspection, repositioning and waiting. A fleet that can cut average dwell by fifteen minutes per vehicle per day does not need fifteen percent more cars to serve fifteen percent more demand. That is the entire margin conversation, and it happens in the depot, not in the model.
Las Vegas is a specific place
Anyone scaling here inherits three local constraints that do not show up in a national rollout plan.
Airport access. Harry Reid is one of the busiest origin and destination pairs in the country for ground transportation, and the rules for where a vehicle can stage, pick up and drop off are their own operating discipline.
Strip curb space. There is very little of it, it is contested, and it is controlled by properties that have their own commercial interests in who picks up at their door.
Event traffic. This city runs on conventions, fights, races and shows. Demand is not a smooth curve, it is a series of spikes, and the spikes are the whole business. CES alone reshapes the road network for a week.
The depot math does not care
Tesla, Waymo, Zoox and Uber are pursuing this with genuinely different vehicles, different sensor approaches and different economics. Vertically integrated, partnership driven, purpose built.
None of that changes the depot math. All four still have to charge, clean, inspect, repair and reposition physical vehicles in a hot desert city with contested curb space and spiky demand. The company that wins here will be the one that treats the operation as the product, not as the thing that happens after the product ships.
Three technologies, three business models, same operating constraints.