Nevada Set Tesla's Robotaxi Ceiling at 5,000. It's Still Running 10.

The Nevada Transportation Authority approved permit ceilings for Tesla, Waymo, and Uber robotaxis on August 20, 2026, TechCrunch reported: up to 5,000 vehicles for Tesla, 1,000 each for Waymo and Uber, over the next 12 months. That is the number making headlines. The number that actually describes what is driving around Las Vegas today is 10: a separate NTA order, issued three weeks earlier, still caps Tesla's live robotaxi fleet at 10 vehicles on a fixed Strip corridor.

What the ceiling actually covers

Clark County, the county that includes Las Vegas, is where all three companies now hold approval. Tesla Robotaxi LLC's ceiling is 5,000 vehicles. Waymo and Uber are each capped at 1,000. Uber's slice covers robotaxis it runs through its self-driving partners, Motional and Zoox, since Uber does not build its own autonomous driving system.

Nevada's order does not explain why Tesla's ceiling is five times larger than the other two. The simplest documented reason: the numbers track what each company asked for. Tesla's own filing in June requested a ceiling as high as 5,000 vehicles, while Waymo and Uber each applied for 1,000. That is an application detail, not a regulatory judgment about which company is closer to ready.

The 10-vehicle order underneath it

Three weeks before the ceiling approval, on July 27, the NTA issued a separate interim order that is the actual rulebook for Tesla's robotaxis right now. Axios reported the terms: 10 vehicles, a top speed of 45 mph, confined to a fixed corridor on the Las Vegas Strip. Electrek's follow-up adds two more restrictions: no pickups at Harry Reid International Airport, and a requirement the order calls "appropriate human supervision."

That is the fleet Tesla is actually running while it holds a permit for 500 times as many vehicles. The ceiling is a future maximum. The interim order is the present-tense reality, and nothing in the August 20 approval replaces it.

The 5,000 has always been a ceiling for us. I don't think we'll be in a position by this time next year to deploy 5,000 vehicles.

Early's own estimate lands closer to 2,500 vehicles a year from now, and he ties any real scaling to FSD v15, the next major version of Tesla's self-driving software, which is not expected before late 2026 or early 2027. The company holding the largest ceiling is also the company telling reporters not to expect it filled.

There's also an open safety investigation

Tesla's robotaxis run the same camera-only Full Self-Driving stack that is currently under active federal review. NHTSA opened Engineering Analysis EA26002 in March 2026, covering roughly 3.2 million vehicles and tied to nine visibility-related crashes, including one fatality. An engineering analysis is the last procedural step before NHTSA can seek a recall.

Why a build studio cares

This is the exact gap our audits are built to catch, just running at civic scale instead of inside one client's product. A permit ceiling is a claim. Ten vehicles on a fixed corridor at 45 mph is the behavior. Our Surface Audit's first line item is checking UI claims against actual behavior, because the two routinely diverge, and the divergence stays invisible unless someone actually checks what is running instead of reading what the announcement says is running.

The tell here was ordinary, not exotic. TechCrunch's own story includes Tesla's engineer downplaying the ceiling, but not the interim order that explains why: the 10-vehicle cap, the route and speed limits, the open NHTSA investigation. Surfacing those took separate reporting from two other outlets. A vendor's claims work the same way: a hedge in a sales pitch is a hint, not a disclosure. Only opening the actual system tells you what the hedge was covering for.

Next step: read TechCrunch's coverage of the permit approval next to Axios and Electrek's reporting on the actual operating restrictions. If a vendor's claims about what their software does need the same side-by-side check, write to us at hello@gattyworks.com.