Tesla Wants Cybercabs in Twelve States. It Has 562 Reported Autonomous Miles in California.

We were northbound on the 5 near the 133 exit, our toddler in the back, when we came up on a Tesla Cybercab.

No steering wheel. One person sitting in what would be the driver's seat if the car had anything for a driver to hold. Both of us saw it, and you can't mistake it for anything else.

Then we passed it. It was holding the speed limit, which tells you what we were doing.

Nothing holds the speed limit on that stretch of the 5. And whatever was operating that car, it wasn't the person sitting in it, because there was nothing in front of them to operate.

I'd assumed those were all in Austin. So I went home and spent an hour figuring out what a Cybercab was doing on an Orange County freeway, and whether I could buy a few and run them myself.

The second answer is no. The first one is more interesting.

Image courtesy of Tesla, Inc.

What California has licensed Tesla to do

The DMV publishes its autonomous vehicle permit holders in three lists. Tesla Robotaxi LLC appears on exactly one of them.

It holds a permit for testing with a driver, current as of the August 12 update. It is absent from the driverless testing list, last updated April 3, and absent from the deployment list. Six companies hold driverless testing permits in this state and Tesla isn't among them.

The DMV describes a testing-with-a-driver permit as requiring a human in the driver's seat who can take control of the vehicle at any time. That permit lets a manufacturer test on any public road in California, so Irvine is not the problem.

The steering wheel is. Whoever was sitting in the left seat of that car couldn't have taken control of anything.

There are innocent explanations and they belong here. The driverless list is five months old, so a permit granted since April wouldn't appear on it. Tesla could be moving the vehicle under some arrangement the published lists don't capture.

I put these questions to the California DMV and to Tesla on September 8. Neither responded by publication. Tesla has had no US communications department since it dissolved the function in October 2020.

Five hundred sixty-two

Tesla reported zero miles of autonomous testing in California in 2025. It has reported none since 2019. Its total on the state's record going back to 2016 is 562 miles.

Autonomous vehicle permit holders in this state logged more than 9 million test miles in the year ending November 30, 2025, by the DMV's own count. Tesla's 562 covers a decade.

That's the company that told an audience in Austin on September 3 that it wants Cybercabs operating in roughly twelve states before the year is out.

What Tesla actually put up

On September 3, Tesla launched the Cybercab at an invitation-only event in Austin. The same day, the company published a form called "Help Us Build Our Robotaxi Network."

It collects your contact information and asks you to pick one of four boxes. Cybercab fleet purchasing. Mobility hubs and infrastructure. Event collaboration. Other.

That is the entire document.

No price. No eligibility criteria. No terms. No timeline. No operational requirements. No revenue split.

A form that asks whether you'd like to buy something without telling you what it costs or when you could have it is not a sales channel.

I filled it out on September 8 and checked the fleet purchasing box. My LLC is From Where I Ride. It holds a photography business and a writing practice, owns no vehicles, employs no drivers, operates no depot and has no transportation authority of any kind.

Nothing on the form asked about any of that. There is no qualification gate, no fleet size question, no capital question, no operating history question. A company genuinely preparing to place vehicles with third-party operators would want to know which operators. This one wanted my email address.

I submitted it on September 8. Nothing had come back by publication.

The people who already paid for this aren't on it either. Tesla has been selling Full Self-Driving since 2019 partly on the promise that owners could add their own cars to a robotaxi network and earn money while they slept. Some of them paid $15,000. Seven years later they're still outside the fence, and the invitation went to businesses instead.

The first thing wrong with the offer

The number Tesla put in front of those owners in 2019 was $30,000 a year in gross profit per vehicle.

Sit with it. If a car parked on Tesla's own network genuinely threw off thirty thousand dollars a year, Tesla would own every one of them. It has the capital, the factory, the software and the app. There is no version of that arithmetic where the correct move is handing the money to a stranger in Ohio.

Companies keep the assets that earn and sell the ones that don't.

What Tesla holds onto in the arrangement it floated is the software margin, the app, the dispatch priority and the ability to revise the revenue split whenever it likes. What you'd get is the vehicle, the note on the vehicle, and whatever the vehicle is worth in five years. That's a franchise where the franchisor sets your price and competes against you in your own market.

The second thing, which is worse

Six weeks before that form went up, Tesla told its own shareholders it couldn't build the car at volume.

The first-quarter update said Cybercab, Semi and Megapack 3 were on schedule for volume production starting in 2026. The second-quarter update, published July 22, dropped the Cybercab out of that sentence. The reason given was battery pack capacity, specifically 4680 cell output, which Tesla named as the main limiting factor on near-term production volume.

Then September 3 arrived. Tesla launched the car and named twelve states without naming which twelve. No production rate. No volume timeline. No price, which is a habit the industry has been picking up and Scout Motors has been running for two years. And a form asking businesses whether they'd like to buy fleets.

Forty-five. That's how many Cybercabs were authorized for driverless operation in Texas the week it launched.

Nobody opens a sales channel for a product they're short of. They ration it. A company with a hundred interested buyers and forty-five cars does not go hunting for buyer one hundred and one.

Which makes the form a survey rather than a sales channel, and what it's measuring is whether outside capital exists to carry the depreciation before Tesla has to carry it itself.

What you'd actually be underwriting

Say Tesla calls you back. Your accountant would have to price five things before you sign anything.

The residual is the whole ballgame. A used Cybercab has nothing to be resold into. Two seats, no steering wheel, no pedals, and autonomy that only functions inside Tesla's network.

You can't sell it to a rideshare driver. You can't convert it to a delivery van. If Tesla ends the relationship you own 3,113 pounds of aluminum with 418 miles of range and nowhere to take it.

Every fleet operator I've ever talked to prices the exit before the entry. The person selling you the car sets the exit price.

Insurance is the other blank line. Carriers have barely worked out how to write a driverless fleet for Waymo, which self-insures at scale and has years of incident data behind it. A small operator asking for coverage on a vehicle class NHTSA opened an audit into on launch day is going to get a long silence and then a number.

Then California again. The DMV adopted new rules on April 29 that took effect July 1, and under them the autonomous vehicle company is treated as the driver.

Incidents get reported within 72 hours, collisions within 24. Local officials can geofence a vehicle out of an emergency zone in two minutes and require a live emergency line answered in 30 seconds. Repeat violations cost you fleet size, then your permit.

Either Tesla holds that permit and you don't control your own asset, or you hold it and you bought a compliance department along with the cars.

Somebody already ran this experiment

MisterGreen is a Dutch leasing company that bought more than 4,000 Teslas. Part of the case for that fleet was the robotaxi income Tesla had been describing since 2019.

The income never arrived. What arrived instead was Tesla cutting new-car prices, repeatedly, which pushed the residuals on MisterGreen's fleet down at roughly three times the rate of the broader used-car market.

The company went bankrupt in December 2025. About $40 million gone.

Nothing improper happened. Tesla priced its cars to move its cars, which is what a manufacturer does. MisterGreen's mistake was owning an asset whose resale value its supplier could reset at will.

The question underneath all of this

There's a version of this where the fleet arithmetic is the small story.

The average new vehicle sold for $49,758 in June, and Kelley Blue Book's own read is that prices held under fifty thousand dollars this year mostly because buyers keep moving down-market. Subcompact SUV sales jumped more than 23 percent. AAA puts the all-in cost of owning a new vehicle at $11,577 a year, which works out to about 77 cents a mile at 15,000 miles.

Set that against a robotaxi fare and the arithmetic isn't close. Waymo runs somewhere between $1.50 and $2.50 a mile in its markets. UberX sits a little under that. A person driving 15,000 miles a year at those rates would pay two to three times what owning the car costs them, and the gap gets wider if the car they own is paid off.

So the crossover people keep predicting hasn't happened, and the fares that would make it happen don't exist outside a slide.

Look at what's being built instead. Tesla is manufacturing a vehicle with no steering wheel and no pedals, and it has published no retail price and no consumer sale timeline for it. Musk floated a figure under thirty thousand dollars when he showed the concept in 2024 and Tesla has not confirmed it since. Hand a private buyer the keys today and they own a parked appliance.

A car sold once earns once. A car in a fleet earns by the mile for as long as it runs. Every manufacturer in this business can do that subtraction, and it shows up in what they choose to build long before it shows up in what you're allowed to buy. It is the same calculation sitting underneath the data your car already sends home, and underneath the camera that watches you drive it.

The industry has stopped needing you to own a car, which is a different thing from taking one away from you.

Back on the 5

Whatever that car was doing near the 133, Tesla has 562 autonomous miles on California's books and no permit to run a vehicle nobody can steer.

The form is still up.