On September 3, Tesla asked businesses whether they want to buy Cybercab fleets. It published no price, no revenue share, and nothing at all about who keeps the vehicles running. We work on that question full time. Below are our estimates for the six things Tesla left blank, the precedent behind each, and what it costs to keep one of these running.
Every one of these is built from somebody else's published figure, never from our own arithmetic alone. None of it is a Tesla number and Tesla may contradict all six. We publish the confidence rating too, because the two that matter most are the two we are least sure of.
Musk's "under $30,000" is a vehicle configuration price, which at Tesla means before destination and order fees. Goldman Sachs estimates a $20,000 to $30,000 build cost at scale.
The discount is the interesting part. The only volume schedule Tesla has ever published is its Sourcewell government contract: 2 percent at 6 to 15 vehicles, 3 percent at 16 to 50, 4 percent at 51 or more. Four percent is the ceiling, and it took a government purchasing cooperative to get it. When Hertz ordered 100,000 Model 3s, Musk said publicly they got no discount at all. Budget above the headline, not at it.
This one decides whether any of it works, and it has the least support of anything on this page.
The 25 to 30 percent figure everyone quotes traces to a single event: Tesla Autonomy Day, April 22, 2019. Tesla has never restated a number since. Meanwhile Uber's effective take rate is now around 42 percent, up from about 32 percent in 2022, and platform take rates drift up rather than down. On the Q2 2026 call Musk said Tesla expects to be "vertically integrated with robotaxi as we are in the rest of our business."
The hardest data point is from the bulls. ARK Invest's published 2029 model, the most aggressive Tesla valuation on Wall Street, gives the fleet partner $0.20 of a $1.213 per-mile fare. That is Tesla keeping roughly 84 percent. If you are modelling a 70 percent owner share, you are more optimistic than ARK.
No autonomous vehicle program anywhere publishes a minimum. Tesla's only published quantity breakpoints are those Sourcewell tiers at 6, 16 and 51 units, and those are discount thresholds rather than entry requirements.
Against that sit the deals that actually got signed. Uber committed to a minimum of 35,000 vehicles with Nuro and Lucid, 2,000 with Pony in Europe, 1,200 with WeRide in the Middle East. Toyota placed 1,000 with Pony. Every disclosed arrangement was struck between two well capitalised corporates at four or five figures of vehicles. If Tesla sets its bar at that scale, the individual buying six to fifty cars does not get in the door at all.
Musk's vertical integration line points to Tesla keeping the work, and while the fleet is small its own depots can absorb it.
Four things point the other way. Tesla already runs an Approved Collision Center program that admits third parties under a published operating standard. It publishes service and diagnostic information specifically for independent repairers. It is defending a certified antitrust class action, allowed to proceed in June 2024, over exactly this question of parts access. And it has acknowledged its own service capacity as a constraint before, on a fleet a fraction of the size a national robotaxi network needs.
Waymo is the counter-model and it is instructive: it outsources 100 percent of fleet maintenance, and picks partners for traditional fleet and EV competence rather than autonomy expertise.
Tesla has published no Cybercab warranty. More tellingly it publishes none for the Semi either, which is the closest commercial vehicle precedent it has.
Its consumer Extended Service Agreement runs $50 a month for a Model 3 and $60 for a Model Y, caps at four extra years or 100,000 miles, carries a $100 deductible per visit, and excludes the high voltage battery, the drive unit, the low voltage battery, glass and every wear item. A robotaxi passes 100,000 miles in under two years. There is no Tesla warranty product in existence that survives this duty cycle.
Four external anchors, none of them Tesla. Waymo derives to roughly $142,000 per vehicle per year, from $355 million across about 2,500 vehicles, at premium fares in dense metros. LongYield models about $79,500 at 20 rides a day, a rate it calls best in class. ARK's 2029 model implies $121,300 at 100,000 miles a year.
And Pony.ai, the only operator anywhere that has published a real per-vehicle figure, reported about $55 per vehicle per day in Shenzhen in March 2026 on 25 orders. That annualises to roughly $20,000 at Chinese fare levels, and it is the floor worth remembering.
Tesla's Austin fares sit well below Waymo's, so expect the lower half of that band at the same utilization.
Revenue share and minimum fleet size are both low confidence, and both are load bearing. If Tesla takes 42 percent like Uber rather than 30, the base case below loses about $8,900 a vehicle a year. If Tesla sets a minimum in the hundreds, the independent owner this entire market assumes does not exist. We would rather show you that on the front page than bury it.
Most cost models circulating right now are built from ordinary EV rideshare assumptions. Every one of these breaks that model, and each is documented in Tesla's own filings and service documentation.
The Cybercab has no onboard AC charger. Level 2 charging is not slow on this vehicle. It is impossible. Every depot port has to be DC.
Installed Level 2: $7,000 to $15,000 per port. Installed DC fast: $50,000 to $250,000 per port. Tesla's own Supercharger for Business configurator prices an eight-stall V4 site at $940,000, roughly $117,500 a stall. The cheap tier of depot charging does not exist here.
Tesla calibrates cameras by driving, not by targets. Clear the calibration, then drive 20 to 25 miles, up to 100, in the middle lane of a multi-lane highway with clear markings and light traffic. On a Model 3 a technician does that by steering the car.
A Cybercab has no steering wheel. Every windshield replacement, camera swap and front-end repair ends in a procedure the vehicle's own design makes hard to perform, and Tesla has published nothing about how it is done on a controls-free car. This is the single biggest open question in Cybercab servicing.
Four electromechanical actuators, one per caliper. No master cylinder and no brake fluid flush, which deletes a familiar service line and adds four components with no published service interval, no published part cost and no third-party diagnostic path.
NHTSA audit query AQ26002, opened September 4, is examining whether Tesla properly self-certified the Cybercab against all applicable federal safety standards. The notice names none specifically, though the brake standard is the obvious candidate for a car with no pedal.
The front and rear tires are different sizes, so rotation is left to right only. Tesla has published no Cybercab-specific interval, and its standard guidance across models is every 6,250 miles. At robotaxi mileage that is roughly ten rotations a vehicle a year, and side-to-side rotation removes the wear-equalizing benefit a normal rotation gives you.
Some good news: at 3,113 lb curb weight and 219 hp through the front wheels, the Cybercab is far lighter and less torquey than the Model Y, so per-mile tire wear should land below the 4.0 to 4.4 cents per mile measured on Model 3 and Model Y.
Tesla's basic new vehicle warranty is four years or 50,000 miles. At robotaxi duty of roughly 60,000 miles a year, you hit the mileage limit in seven to ten months. The years never matter.
The warranty also excludes tires, glass, wiper blades, brake pads, cabin filters, alignment, balancing, and seat, trim and upholstery punctures and tears. Every interior mess is out of pocket by written policy. Tesla has published no Cybercab warranty document at all, and its used vehicle warranty still excludes ride-share, passenger-for-hire and "use by multiple drivers" in those words.
This is the fear everyone brings to the first call, and the evidence does not support it. An Uber Model 3 passed 255,000 miles on its original pack with no more than 12 percent capacity loss and no drive unit replacement. A 2016 Model S passed 280,000 miles of heavy rideshare with about 6 percent loss.
What actually drains a fleet account is tires, control arm bushings, glass, low-voltage batteries, interior damage, brake corrosion from disuse rather than wear, and the four to six month collision repair cycles Tesla owners routinely report. Downtime, not parts, is the expensive line.
Tesla publishes no revenue per vehicle and no utilization data. This model is built from Waymo's regulator-filed utilization figures, Tesla's observed Austin fares, and measured Tesla service costs. Change anything you disagree with. The assumptions are listed under the panel.
| Per vehicle per year | Amount |
|---|---|
| Gross fares | |
| Your share after network take | |
| Miles driven, including deadhead | |
| Vehicle depreciation | |
| Energy | |
| Service, maintenance and cleaning | |
| Insurance and permits | |
| Net before tax |
Move the rides slider and revenue swings by tens of thousands. You do not control that number. Tesla's dispatch does. Switch charging from managed to public and watch what happens to a line you do control. Between a well-run and a badly run fleet there is roughly ten thousand dollars a vehicle a year, and all of it sits on the cost side. That gap is the entire reason this company exists.
Assumptions: 8 miles driven per paid ride including deadhead, derived from Waymo's 4 million weekly miles against 500,000 weekly rides. Vehicle at our estimated $32,000 delivered price over four years, per section 01. Managed depot energy at $0.12/kWh against $0.42 on the public network, at 165 Wh per mile from EPA filings. Insurance $8,000 and permits $500 per vehicle per year, from published taxi and livery bands. Managed service is our rate card below; self-managed assumes the same work bought retail at roughly 35 percent above negotiated fleet rates, and does not price your own time or the extra downtime. Every one of these is an assumption, not a Tesla figure.
Pitlane is our managed service. It is a national network, not a depot company. Certified partner shops, mobile technicians and charging sites do the work under our standard, our rates and our documentation. You get a single relationship instead of forty.
Nobody else in this market does. Published fleet management pricing today covers telematics software only, at $15 to $50 a vehicle a month. The full-service per-vehicle fees charged by Element, Holman, Enterprise and Merchants are negotiated privately and never disclosed. If you are deciding whether to put capital into an asset class, an unpublished price is not a price. Ours is on the page.
No. Tesla has sold zero Cybercabs to third parties. What exists is an interest form, which Tesla's own fine print calls "a request to be considered for future Robotaxi opportunities, not a purchase order." Reported signups reached roughly 73,000, though that number came from a single post on X rather than from Tesla, and a Tesla engineer separately called it tens of thousands. Either way none of them is an order, because there is nothing to order yet.
Fill the form out anyway. It costs nothing and position in a queue is worth something. Just do not budget against it.
Tesla removed the ride-share and commercial-use exclusion from its new vehicle warranty in a 2024 revision, and it is absent from the current document. The used vehicle warranty still carries it in full, including a clause excluding "use by multiple drivers," which would cover essentially any fleet.
Tesla has published no Cybercab warranty at all. Treat coverage as unknown, and note that the mileage limit expires in under a year regardless.
No, and this is the most common error in circulating fleet models. The 45W commercial clean vehicle credit was terminated. The IRS is unambiguous: no credit is available for vehicles acquired after September 30, 2025. Sections 30D and 25E went the same way.
What remains is depreciation. The 2025 tax act permanently restored 100 percent bonus depreciation for qualifying property placed in service after January 19, 2025, and raised the Section 179 limit to $2.5 million. That is a timing benefit, not a subsidy, and whether the Section 280F caps apply depends on gross vehicle weight rating and on the transporting-persons-for-compensation exception. Take that one to a tax advisor, not a website.
There is no published carrier program for third-party robotaxi fleet owners. Underwriting today is bespoke. The usable benchmarks are the taxi band at roughly $9,500 a year and limousine at roughly $10,400, with Uber Black livery quoting $4,800 to $14,400 depending on metro.
On liability: at Level 4 the operator of the automated driving system generally carries responsibility while the system is inside its approved operating conditions. Outside them, or where maintenance was deficient, liability moves back toward the owner and the service vendor. In August 2025 a Florida jury apportioned Tesla 33 percent in an Autopilot case that returned $129 million compensatory and $200 million punitive. Juries apportion. They do not hand it all to the manufacturer.
Roughly 21 states permit driverless commercial operation, with Florida, Georgia, Arizona and Tennessee among the lightest touch. Texas is no longer permissionless: SB 2807 became enforceable May 28, 2026, and the personal-use exemption does not cover a revenue fleet.
California is effectively closed to a controls-free vehicle. Tesla operates there today on a charter-party carrier permit, the same authorization a limousine company holds, and the CPUC has stated plainly that Tesla is not operating an autonomous vehicle service. That workaround depends on a human sitting in the driver's seat, which a Cybercab does not have.
No, and we are careful about that phrase for a reason. Under the Howey test, buying a revenue-producing asset, handing it to a manager and collecting returns generated by the manager's efforts is the fact pattern the Supreme Court addressed in SEC v. Edwards, where payphones sold with a management leaseback were held to be securities. The Court specifically held that a fixed return is no different from a variable one for that purpose.
That is why we sell a service to people who already own vehicles, rather than selling vehicles bundled with management. Anyone offering you the bundle with a projected yield attached should be asked how they have handled that question.
At the enterprise scale, plenty. Every large robotaxi program already has an operator: Transdev, Moove and Avomo, Avis Budget, Element Fleet and Lyft's Flexdrive for Waymo; Hertz and Oro Mobility for Uber; Zoox in house. Those are multi-year contracts with counterparties that own their own vehicles.
At the scale of somebody buying six to fifty cars, effectively nobody. Every incumbent is built for one large customer per city. That is the gap we are built for, and we would rather say so than pretend we invented the category.
We will send back a cost model for your market and fleet size, built from the same sources as this page. No obligation and no deck. If your plan does not work, we will tell you that instead.