The same afternoon Tesla’s Cybercab carried its first paying passengers in Austin, a page went live on Tesla’s website titled “Help Us Build Our Robotaxi Network.” It asks for a name, a company and a deployment region, and it offers a checkbox for Cybercab fleet vehicle purchasing. No price, no ownership terms and no revenue split have been published. Tesla had 45 Cybercabs registered in Texas as of Friday morning, among 420 autonomous vehicles in its state fleet.

Federal regulators moved the same day. The National Highway Traffic Safety Administration opened an audit into about 1,000 Cybercabs , examining the process and data Tesla relied on to certify a vehicle that has no steering wheel, no pedals and no mirrors. In one week Tesla told the market it intends to be the platform for the robotaxi business rather than the owner of every car in it, and the regulator asked to see the paperwork.

The First Week In Numbers

Tesla launched at an invitation-only event at ACL Live on Thursday and opened rides to the public on Friday afternoon in parts of Austin. Ashok Elluswamy, Tesla’s AI chief, said the service was “open to the entire public.” Executives described dynamic pricing that rises with demand and promised “a first-class experience at coach price.”

The early fares show what a 45-car fleet looks like against an established one. Around lunchtime Thursday a Robotaxi-branded Model Y from Montopolis to ACL Live was quoted at $19.58 , while Uber, which partners with Waymo in Austin, quoted $12.96 for an electric vehicle on the same route. On Friday a two-and-a-half-mile Cybercab ride down South Congress was priced at $12.15 with a projected wait of 40 to 50 minutes, against $7.96 and a wait under ten minutes on Uber. Those are snapshots, and ride-hailing prices move minute to minute, but the direction was consistent. Tesla’s earlier fixed fares of $4.20 and then $6.90 are gone.

Cybercab production began at Giga Texas in April, and Musk warned that initial output would be very slow before ramping toward the end of the year. Waymo has 988 vehicles registered in Texas and nationally runs about 3,800 cars delivering more than 500,000 paid rides a week . Tesla is starting from a fleet roughly one eightieth that size, at prices above the incumbent, with waits measured in the better part of an hour. Today Tesla’s robotaxi is scarcer and dearer than its rivals in its own home city. The order form is the plan for changing that.

The form is the business-model disclosure. A company buys Cybercabs, runs them on Tesla’s network, and Tesla takes a share of every fare. Musk has put the car’s target price under $30,000 , and closer to $25,000 on earnings calls. The buyer carries the capital cost and the depreciation. Tesla carries the software, the app, the pricing and the dispatch.

That is Uber’s structure, in which drivers own the cars and the network owns the customer, with one addition. Tesla also builds and sells the car, so it collects a manufacturing margin at the front of the arrangement and a platform take for the life of the vehicle. The depreciation risk that comes with owning a fleet of Teslas is exactly what the form moves off Tesla’s books and onto the buyer’s.

There is a precedent for what that risk looks like. MisterGreen, a Dutch leasing company, built a fleet of more than 4,000 Teslas partly on the expectation of robotaxi income and went bankrupt in December after write-downs on the fleet’s value. Electrek’s read on the new form, published Sunday , is that a company would not sell a profitable fleet business to outsiders. The structural answer is that Uber, Airbnb and the app stores were all built by letting someone else own the asset while the network kept the toll.

A car with no steering wheel has one use, and its owner’s return depends entirely on the software Tesla keeps shipping to it.

That dependence cuts in Tesla’s favor as well. A private owner who paid $15,000 for Full Self-Driving could wait indefinitely for the robotaxi income promised at Autonomy Day in 2019. A fleet buyer with a hundred Cybercabs cannot, and neither can Tesla, because the take on every ride is Tesla’s revenue too. The incentives line up in a way they did not when the promise was made to individual owners.

The Audit Is About A Rule Already Moving

NHTSA’s Audit Query AQ26002 examines the process and technical data Tesla relied on when it certified the Cybercab. Specifically, it asks how far Tesla’s certification depended on determining that certain federal standards do not apply to the vehicle. Most of the Federal Motor Vehicle Safety Standards were written around a human driver. They assume a wheel, a brake pedal, side mirrors and controls within reach, and the Cybercab has none of them.

Two roads exist for a vehicle like that. Zoox took the exemption road. On July 30 NHTSA granted it the first commercial exemption from eight standards, including the brake-system rule. It covers up to 2,500 vehicles a year over two years, with the agency keeping oversight as the fleet grows.

Tesla took the self-certification road, declaring the standards inapplicable and putting the cars on the street. The audit is NHTSA deciding whether the second road is open.

The regulator’s own posture suggests it wants it to be. In June NHTSA proposed rewriting the brake standard so that the pedal requirements apply only to vehicles with manual controls, while every vehicle still has to meet the stopping-distance rules. In July it changed its exemption process so vehicles can be built before an exemption is granted, and it is updating its automated-vehicle guidance for the first time since 2017. Its administrator, Jonathan Morrison, framed the Cybercab audit as “balancing innovation with safety oversight,” which he said would let the United States keep its lead in the technology. The audit decides which of two roads Tesla’s volume travels on, and both roads are open.

If fleet buyers show up, Tesla’s robotaxi economics become a manufacturing margin at the point of sale plus a software take on every ride, with someone else’s capital tied up in the cars. Waymo’s economics are its own capital in 3,800 vehicles plus the cost of running them. The Austin fare gap reflects fleet size today, and the fleet-sales channel is Tesla’s mechanism for closing it without funding thousands of cars itself.

The companies that fit this shape are the ones that own a network and let others own the assets on it. Uber does not own the cars on its platform and Airbnb does not own the rooms. Tesla is reaching for that structure with a difference that matters to a manufacturer, which is that it also gets paid for the asset.

The signals from here are concrete. The first named fleet buyer, and the revenue split it accepts, will price the model. The outcome of AQ26002 will show whether Tesla stays on the self-certification road or gets pushed toward Zoox’s exemption cap. And the fare gap against Uber, as the Austin fleet grows past 45 cars, will say more than the launch event did.

Tesla built a car nobody can drive, and in the week it went on sale, the buyer it named was a fleet.