How fast will robotaxis scale?

Author: Bern Grush

Date Published: May 24, 2026

In The End of Driving (Grush, Niles, Miller, 2nd edition), one of our key assertions was that a fundamental limit exists on robotaxi diffusion, driven by suitability to task, family composition, user preference, and other social factors. That limit likely differs by region: roughly 40–50% in the US, somewhat higher in the EU given denser cities and stronger transit access, and higher still in Asia where the current ownership baseline is much lower and robotaxi access is more likely to leapfrog personal ownership.

This blog post explores likely robotaxi diffusion in US cities assuming a 50% passenger miles travelled (PMT) asymptote, using the following baseline assumptions: a U.S. starter fleet of 3,100 vehicles at the start of 2026; each vehicle services 22 trips/day with an average of 1.5 passengers per trip (33 person-trips/vehicle/day); 80% of the US population is addressable at saturation; users make 3.37 trips/day on average; robotaxis generate a 20% increase in trip demand relative to 2026; and the fleet size plateaus as it reaches 50% of passenger demand serviced. Population growth is not modelled.

The U.S. Robotaxi Fleet: 2026 to Saturation 

A Logistic Diffusion Model for Municipal Planners 

Purpose

This model estimates how long it will take for robotaxis to account for 50% of motorized passenger surface mobility in the United States — a scenario in which automated taxi fleets grow to service half of all personal trips while the remaining half uses privately owned vehicles or residual non-automated public transit. The central finding is that even under an optimistic scenario this threshold is not approached until the mid-2050s, and under a conservative but plausible scenario it extends to the mid-2060s. The 50% figure is a modelling anchor, not a prediction: it sits halfway between the status quo and the extreme claim that every trip will eventually be automated. The exercise is intended to replace hopeful optimism or knee-jerk dismissal with a quantitative basis for long-range municipal planning. The shape of this change is very likely as shown in figure 1, but its time axis is less certain.

Model Structure

Growth follows a logistic (sigmoidal) function — the standard model for technology diffusion, and the framework underlying Everett Rogers’ Diffusion of Innovations. The curve has three phases: a slow initial ramp among early adopters; rapid acceleration through the inflection point when the early majority is being served (approximately 2047–2054 depending on scenario); and a gradual deceleration as the remaining accessible market — the late majority and laggards — shrinks toward zero.

The model is demand-driven, not supply-constrained. Manufacturing capacity is more than sufficient to produce and refresh the 16+ million driverless vehicles needed at peak. At a four-year turnover cycle, this would be roughly 4 million new vehicles per year. The true constraints are regulatory pace, price trajectory, user perception, and behavioral adoption, all of which lose their volatility over time. These are collapsed into a single logistic growth rate, calibrated by specifying the year at which each scenario approaches saturation (97% of plateau).

The shape and structure of this model are grounded in how diffusion works; the speed at which the system unfolds is inherently uncertain. Three scenarios are offered to bracket plausible outcomes. All assume U.S. national averages; in practice the technology will cluster in high-density areas, so some municipalities will experience much larger relative fleets — and much earlier disruption — than others.

The Three Scenarios

All three scenarios share identical starting conditions and plateau. They differ only in the pace of adoption, comprising the combined effect of regulation, price, perception, and behavior:

Optimistic (2055): Regulation advances smoothly; prices fall to personal-vehicle-competitive levels in urban geographies by the mid 2030s; vehicle turnover accelerates. Inflection point, late 2040s.

Base (2060): A cautious, muddling-through scenario. Regulatory friction is persistent but not blocking. Prices become ridehail-destructive in the late 2020s and ownership-competitive by 2040. Inflection, 2050.

Conservative (2065): Sustained political resistance (job concerns, liability debates, risk fears), slow price descent and strong personal vehicle culture in lower-density metros. Inflection mid 2050s.

Note: 2055 is already optimistic. A personal vehicle purchased today may not be retired until ~2040. Even if no new privately owned vehicles were sold after 2030, the existing fleet would not be fully turned over until the mid-2040s. And new personal vehicles will continue to be sold throughout this period.

Assumptions and Data Sources

Starting Fleet (2026 Baseline)

The commercial US robotaxi fleet at the start of 2026 is estimated at 3,100 vehicles. Waymo accounts for approximately 3,000 operating in a half dozen cities. Zoox accounts for ~100 vehicles. Other operators currently contribute negligibly. Tesla’s fleet is excluded here, since the majority still carry human safety monitors and do not yet constitute an autonomous commercial service.

Addressable Population

80% of the US population lives in metropolitan areas sufficient to support a minimum commercial fleet as the technology improves and costs drop. This maps to metros of approximately 50,000 people or more, covering roughly 400 US metropolitan statistical areas. Rural and exurban residents are excluded from the addressable market.

Plateau Fleet

At the 50% PMT asymptote, approximately 550 million robotaxi person-trips must be served daily (272M x 3.37 trips/day x 1.2 induced demand x 50% PMT share / 33 person-trips/vehicle/day). This yields a plateau fleet of approximately 16.7 million vehicles generating just over 200 billion person-trips per year.

Trip Demand Baseline

3.37 person-trips per capita per day (all modes, all ages) from the 2017 National Household Travel Survey — the last pre-pandemic survey conducted under normal conditions. The 2022 NHTS recorded a 37% decline attributable to pandemic disruption and was not used. Vehicle occupancy of 1.5 persons per trip is confirmed by the 2022 NHTS.

Induced Demand

Robotaxi users are assumed to make 20% more trips than the current per-capita average, reflecting the elimination of the driving burden, falling per-trip cost, and mobility unlocked for non-drivers. This 20% uplift is applied to the addressable population.

The 50% PMT Asymptote

50% PMT reflects a long-run equilibrium in which approximately half of all personal travel — measured in person-miles — occurs in robotaxis, while the other half remains in privately owned vehicles: families with children, people carrying equipment, rural residents, those who simply prefer ownership, and other personal reasons. The true national figure will vary by context: the US, with ~850 vehicles per 1,000 people, is likely to land lower than Europe (~650/1,000) or Asia (~400/1,000). 50% is used here as a reasonable working hypothesis to cap the model in Figure 1.

What This Means for City Governments

The S-curve carries a specific policy warning: the curve looks flat and manageable, before accelerating sharply. Under the base scenario, this model shows the US robotaxi fleet remaining modest for several more years. This might tempt planners to defer action. But at the inflection point, the fleet is doubling roughly every two years, and only after that does growth gradually decelerate towards the plateau of 16.7 million vehicles. Cities that wait for this acceleration to become evident before preparing will find themselves designing policy under pressure and partially blind. Worse, if Tesla resolves its current, remaining issues, this timeline relative to municipal curb management will likely compress still further.

Specific preparation priorities, with timing dependent on local conditions:

Curb management frameworks, PUDO zone design, and data-sharing agreements should be drafted now. Early decisions shape desirable outcomes; indecision will compound current problems.

Revenue replacement strategies for parking and transit integration plans should be operational before autonomous fleets enter your city. As fleets grow, they materially impact the curb in dense urban cores.

At the steepest part of the curve — halfway to 50% saturation — cities without robotaxi infrastructure will face acute curb congestion, unmanaged PUDO activity, and parking revenue collapse.

---

Sources: NHTS 2017 (trip rates, DOT/FHWA); NHTS 2022 (vehicle occupancy, DOE/FHWA); Waymo fleet data (TechCrunch Feb 2026; Automotive World Sep 2025); Zoox commercialisation (CNBC Jan 2025; Fortune Jun 2025; NHTSA Aug 2025); Vehicle ownership rates (OICA/World Bank). Model: logistic diffusion, demand-driven. URF, May 2026.