Tesla launched the purpose-built Cybercab in Austin on September 3 with dozens of cars on the road. The harder number arrived just before it: 420 vehicles registered to Tesla Robotaxi in Texas, split between 375 Model Ys and 45 Cybercabs, reported from a September 3 pull of the state registry.
That is the first usable starting point for a year-end fleet estimate. It is also narrower than it looks. The 420 is a state registry count, not a tally of cars carrying fares at the same moment. Tesla does not publish inventories by city, and press reporting puts the unsupervised operating fleet at more than 200 vehicles across six cities, below the registered total in Texas alone.
Our city-level starting split is therefore an estimate from a hard state total: 240 vehicles in Austin, 90 in Dallas and 90 in Houston. We add 90 estimated vehicles across Miami, Orlando and Tampa, producing a 510-vehicle national starting point. The Texas total is observed; the allocation and Florida count are ours.
From there, the probability cone above ends December at 2,500 vehicles in the P10 case, 7,550 at P50 and 14,100 at P90. Those are not Tesla targets. They are the output of twelve city curves, each with a different starting fleet, regulatory path and probability of opening or scaling before the year ends.
The cities make the curve
The six live markets carry almost all of the downside case. Austin is the anchor, Dallas and Houston reuse the same Texas authorisation, and Miami, Orlando and Tampa operate under Florida's statewide framework. Even if no new metro contributes meaningfully, those six can add cars without waiting for another city to approve them.
The upper half of the cone comes from markets that can arrive in waves:
- Las Vegas is the strongest new-city candidate. Nevada approved Tesla for up to 5,000 autonomous vehicles during the first twelve months. That is a ceiling, not a deployment forecast; Tesla's own Cybercab chief engineer said 2,500 vehicles in Clark County would satisfy the company.
- Phoenix has a short regulatory path. Tesla appears on Arizona's autonomous-vehicle operator list, and Arizona requires a law-enforcement interaction plan, federal compliance, registration and insurance rather than California's layered testing-and-deployment process.
- Charlotte is evidence of preparation, not permission. Fourteen Cybercabs and fourteen Model Ys with Texas plates were photographed there in July, and Tesla advertised for an operator to oversee ride-hailing data collection. We give it a 100-vehicle midpoint and a 600-vehicle upper case by December, but zero in the downside.
- San Antonio and Jacksonville are replication bets. Their states already host Tesla's driverless service. Neither city has an announced Tesla launch, so both are zero in P10 and only 100 vehicles apiece at P50.
- The Bay Area stays at zero in the midpoint. California permits Tesla to test with a safety driver. It does not yet permit Tesla to run a commercial driverless passenger service, so only the P90 case gives the region a token 100 vehicles before year-end.
The city trajectories below use the same values as the aggregate cone. They are S-curves rather than straight lines because a fleet normally spends weeks near zero, accelerates after a launch clears, then slows as depot, charging and support capacity fill.
Every city estimate, including the zeros
The midpoint is intentionally concentrated: Austin accounts for 2,200 of the 7,550 vehicles, or 29%. The six already-live metros contribute 6,000, nearly four fifths of P50. Las Vegas and Phoenix contribute another 1,300. Charlotte, San Antonio and Jacksonville together contribute only 250, while California contributes nothing.
That composition matters more than the headline total. A 7,550-car midpoint does not require twelve successful launches. It requires the existing six markets to absorb roughly 5,490 additional vehicles in 118 days, then Las Vegas and Phoenix to supply most of the remaining expansion.
Vehicles fleeted at 31 December 2026 — R40 scenarios
| Metro | Start | P10 | P50 | P90 |
|---|---|---|---|---|
| Austin | 240E | 1,000 | 2,200 | 3,200 |
| Dallas | 90E | 350 | 900 | 1,500 |
| Houston | 90E | 350 | 900 | 1,500 |
| Miami | 40E | 300 | 800 | 1,300 |
| Orlando | 30E | 250 | 650 | 1,100 |
| Tampa | 20E | 200 | 550 | 900 |
| Las Vegas | 0 | 50 | 700 | 1,600 |
| Phoenix | 0 | 0 | 600 | 1,300 |
| Charlotte | 0 | 0 | 100 | 600 |
| San Antonio | 0 | 0 | 100 | 600 |
| Jacksonville | 0 | 0 | 50 | 400 |
| SF Bay Area | 0 | 0 | 0 | 100 |
All figures are estimates. Current-city starts are allocated from Texas's observed total and an estimated Florida fleet. Zeros mean the scenario assumes no driverless commercial launch by year-end, not that testing stops.
The P90 case is much harder. It requires 14,100 vehicles by December, including 1,600 in Las Vegas, 1,300 in Phoenix and 1,700 across four markets that are not operating driverless Tesla service today. That is not merely a faster factory. It is several regulatory and operational decisions landing close together, followed immediately by vehicles, depots, charging, cleaning and remote support.
The P10 case is different. Its 2,500 vehicles can be reached almost entirely inside markets that are already live: 1,000 in Austin, 700 across Dallas and Houston, and 750 across the three Florida metros. Las Vegas contributes only 50 and every other candidate market contributes zero. That is why 2,500 is a genuine downside rather than a disguised failure case.
What the range means for Tesla's model
Our Tesla model starts its Robotaxi line at 1,000 vehicles in commercial service, holds that fleet flat through the September quarter and then adds 600 vehicles in the December quarter. Its year-end count is therefore 1,600 — below even the 2,500 P10 case here.
That is not enough evidence to change the model. The model counts vehicles in commercial service; the cone counts vehicles registered or operationally committed because that is the only observable basis available. A registered car can sit at a depot, and the gap between registered and earning is precisely what Tesla does not disclose.
It does show where the valuation sensitivity sits. Using the $70,000 of annual revenue per working vehicle from our fleet-size arithmetic, the three year-end cases imply annualised revenue of roughly $175 million, $529 million and $987 million. Against Tesla's $103.6 billion trailing revenue base used in that analysis, they equal 0.17%, 0.51% and 0.95%.
The top of the cone almost reaches the 14,800 vehicles needed to move Tesla's revenue by one percentage point. The midpoint crosses the roughly 7,000-car threshold for one cent of quarterly EPS under our assumptions. The downside does neither.
This is the useful way to read the chart. The probability cone is not a claim that 7,550 is known more precisely than 7,500. It says the 2026 fleet becomes financially visible only if several city S-curves steepen together. Austin alone cannot do it, and a permit alone cannot do it. Tesla needs both the cars and somewhere lawful to put them.
What to watch
- Whether Texas registrations move beyond 420 during September. The Cybercab launch supplied 45 purpose-built cars but no published weekly deployment rate. The next registry pull will show whether launch inventory was a batch or the beginning of a curve.
- Las Vegas vehicles, not the 5,000-vehicle ceiling. A few hundred cars by December supports the midpoint. A token fleet leaves the new-city wave mostly in Phoenix.
- A Phoenix service date. Arizona's path is open enough to move quickly, but preparation is not a launch. The midpoint assigns Phoenix 600 vehicles and therefore needs a decision soon.
- The registered-to-operating gap. Texas has 420 registered vehicles while reporting puts the six-city unsupervised fleet above 200. If the gap widens, the cone is measuring inventory rather than revenue capacity.
- Any California driverless filing. The Bay Area remains zero at P50 until Tesla enters the state's driverless testing and commercial passenger process. A filing would move the probability before a single car moves.
Tesla's Texas registrations are press-reported from direct queries of the state system; Tesla has not published a city inventory. The six live metros and production status come from Tesla's Robotaxi page and Q2 update. The launch and more-than-200 operating vehicles are reported by AP. Nevada's ceiling comes from the Transportation Authority; Arizona and California status from ADOT and the California DMV. Charlotte preparation is reported by Axios. Every allocation, probability, curve and conversion is R40 arithmetic as of September 4, not Tesla guidance. Registered vehicles are not necessarily carrying paid rides.
