LAS VEGAS, 22 AUG 2026 — The Nevada Transportation Authority unanimously approved three permits on 20 August allowing Tesla, Uber and Waymo to run commercial robotaxi services in Clark County, which contains Las Vegas. The permits run for twelve months and authorise up to 5,000 vehicles for Tesla, 1,000 for Waymo and 1,000 for Uber.
At the same proceeding, Tesla's Cybercab chief engineer Eric Early said he did not think the company would be in a position to deploy 5,000 vehicles by this time next year. The ceiling and the expectation were stated in the same room, on the same day, and only one of them will be quoted.
What was actually granted
Uber does not build autonomous vehicles and will operate through partnerships with Motional, the Hyundai subsidiary, and Zoox, which holds its own permit for 100 vehicles.
A permit ceiling is an authorisation to deploy up to a number. It is not a commitment, a forecast or a plan, and nothing obliges an operator to reach it. Regulators grant headroom because returning for an amendment is administratively expensive, so the ceiling is generally set well above what anyone expects to build.
The number in the headline is not the number to watch
The headline figure of eight thousand vehicles requires some care.
First, the three named per-operator ceilings total seven thousand, with Zoox's separate hundred bringing the identifiable total to seven thousand one hundred. Reporting also refers to a total of up to eight thousand across the county. We could not reconcile the difference from the available material, and it may reflect permits or allocations not itemised in the coverage.
Second, and more importantly, the ceiling is the least informative number available. The operator with the largest allocation used the same hearing to say it does not expect to reach it. That is an unusually direct statement against interest, and it deserves more weight than the permit total precisely because a company has little incentive to talk down its own capability in front of a regulator that has just been generous.
The real number to track will be vehicles in revenue service. None of the operators have published this figure, which will likely be a fraction of the permitted ceiling for the first several months.
Why the ceiling was set so high anyway
The gap between the authorised ceiling and the expected deployment is not regulatory theatre; the two numbers answer different questions.
The regulator is deciding whether an operator is fit to run a passenger service and, if so, what upper bound keeps the deployment within what the county can absorb and oversee. That bound is a safety and traffic-management judgement made once. The operator, meanwhile, is solving a supply-chain problem: how fast it can build, deliver, commission, insure and support its vehicles.
A regulator that set the ceiling at the operator's realistic twelve-month capability would guarantee a second hearing the moment that capability improved. Setting it high and letting operations bind is the ordinary and sensible outcome, not a failure of scrutiny.
Three operators, three different propositions
Grouping these permits together obscures how unlike the three deployments are.
Waymo operates a mature commercial service in several American cities with a vehicle platform and remote-support model that has been running for years. A thousand vehicles in a new metropolitan area is an expansion of a proven operation.
Uber is not deploying its own technology at all. It is supplying demand, dispatch and payments to fleets that Motional and Zoox operate, which makes its allocation a commitment by its partners rather than by Uber.
Tesla is the outlier in both directions. It has the largest ceiling, the least commercial driverless operating history of the three, and an approach that relies on camera-based autonomy rather than the sensor stacks its competitors use. Its earlier Nevada activity was constrained to a small geofenced deployment on the Strip with a low speed limit and no airport pickups.
The relationship between that earlier constrained arrangement and this passenger-carrier permit is not described in the available reporting, and the two are likely different instruments granted for different purposes. This is not a case of a regulator reversing itself in three days. The simpler explanation is that two separate approvals are doing two separate jobs.
Why this matters outside the United States
Clark County is an unusually favourable proving ground, and that is the point for anyone assessing what these permits predict elsewhere.
Las Vegas offers a grid street layout, minimal rain and no snow, a large population of visitors without cars, and dense trip demand concentrated between a small number of destinations. Almost every environmental variable that makes autonomy hard has been removed.
Cities across ASEAN present close to the inverse. Dense mixed traffic with motorcycles operating outside lane discipline, monsoon rainfall that degrades both cameras and lidar, informal kerbside behaviour and inconsistent road marking are the normal case rather than an edge case. Performance in Clark County will say very little about performance in Jakarta, Ho Chi Minh City or Bangkok, and it should not be cited as though it does.
What does transfer is the regulatory model. A transport authority granting a fixed-term, capped, per-operator passenger-carrier permit is a structure any regulator can copy, and it is considerably more measured than either a blanket authorisation or an indefinite prohibition.
What remains unconfirmed
The available reporting does not describe the permits' operational conditions. It is unclear whether safety operators are required, what geofence applies, or what restrictions on speed, weather, or airport access are in effect. Incident-reporting obligations are also unknown. No fare or pricing information has been published.
The relationship between these permits and Tesla's earlier and much smaller Nevada authorisation is not established, nor is the reconciliation between the itemised per-operator ceilings and the reported county total. No launch dates have been announced by any of the three operators, and it is not stated whether the twelve-month terms are renewable.
What to watch for
The most important figure to watch for is the number of vehicles actually in service. If the operators do not publish it, the permitted ceiling will become the public figure by default for a year — which suits everyone except the public.
For Tesla specifically, we have a rare public benchmark from their own engineer. Tracking their actual fleet size against his modest expectation will be a cleaner test than relying on company forecasts.
Finally, the value of this entire regulatory model depends on incident reporting. A capped, fixed-term permit is only a meaningful control if the authority publishes what happened. Whether Nevada does so will show other regulators if this approach is worth copying.
View author profile → · Editorial policy
