Robotaxi Hype Has Left Cities Unprepared

Author: Bern Grush

Date Published: January 25, 2026

For as long as we can remember, cities have been told the robotaxi revolution is imminent. In 2017, RethinkX's Seba and Arbib predicted that by 2030, 95% of U.S. passenger miles would shift to autonomous vehicles, collapsing the American car fleet from 247 million to 44 million. Urban planners, dutifully trying to prepare, commissioned studies, debated parking requirements, and sketched new street designs. Then 2020 came and went. So did 2025. The pervasive transformation has not materialized.

One by one, the bold predictions collapsed. Elon Musk's 2019 claim of one million Tesla robotaxis by 2020 missed by 100%. GM invested over $10 billion in Cruise before shutting down operations entirely in late 2024. Boston Consulting Group's 2015 forecast of fully autonomous vehicles by 2025 proved fantasy. And city planners, burned by false alarms, learned to tune out the hype.

After years of crying "revolution," the industry has trained exactly the people who need to prepare for robotaxis—municipal officials, traffic engineers, urban planners—to stop listening. Like the classical fairy tale of Peter who cried "wolf" too many times, cities may no longer be listening now that the risks are real.

The irony is brutal: just as the technology finally appears ready to scale, cities are least prepared to handle it. Waymo now provides over 250,000 paid, fully driverless rides weekly across multiple U.S. cities with roughly 100% annual growth. China's Baidu operates at a similar scale and likely faster growth rate. The inflection point appears to have arrived, and our history of hype-casting has left us dangerously unprepared.

What Failed Predictions Cost Cities

Transportation planners work on decade-long timelines. They need reliable signals about when transformative technologies will arrive to coordinate infrastructure investments, update regulations, redesign streets, and prepare traffic management systems. When those signals prove consistently wrong, planners learn—rationally—to discount them.

Consider what happened in cities that took early predictions seriously. They held community meetings, explored curb redesign, debated zoning changes. Then the robotaxis didn't come. Budgets were applied elsewhere. Political capital was spent. The institutional memory becomes: "We've heard this before."

McKinsey's surveys reveal the problem's scope. Expert predictions for Level 4 robotaxis slipped by two to three years between 2021 and 2023[i], then slipped another one to two years by 2025.[ii]  We're not preparing—we're simply rediscovering that autonomous driving is harder than anticipated. This pattern of compounding delays teaches city officials exactly the wrong lesson: ignore autonomous vehicle planning until vehicles are already on the streets.

The Deeper Problem: We Measured the Wrong Things

Forecasting failures go beyond misstated timelines. The entire framing of what cities needed to prepare for was fundamentally flawed.

Industry projections obsessed over vehicle capabilities: miles between interventions, safety comparisons to humans, SAE automation levels. This encouraged a dangerous assumption—that because vehicles are "smart," they'll figure out integration themselves. Call it the "PUDO" fallacy: the belief that pickup and drop-off (PUDO) operations will somehow self-organize at the curb without municipal planning or coordination.

Reality works precisely opposite. The more capable robotaxis become, the more critical municipal preparation becomes. A city with 1% of trips handled by robotaxis can absorb impact through ad-hoc adaptation. A city approaching 10% faces systematic challenges: curb management, traffic flow optimization, equity questions about who accesses increasingly scarce curb space and when.

Almost no forecasts addressed what actually matters to cities: what percentage of total passenger-kilometers will shift from private vehicles to robotaxis? when will that shift occur? and what municipal infrastructure adaptations does each stage require? Instead, we got "market size in dollars" and "rideshare market share"—metrics that sound impressive but tell planners nothing about how and when to rethink curb management.

The denominator problem reveals the mismatch. The United States sees roughly seven trillion passenger-kilometers of motorized travel annually.[iii] Waymo's 13 million annual rides, even assuming 1.5 passengers per trip at 10 kilometers each, represent approximately 195 million passenger-kilometers—less than 0.003% of U.S. travel. Even with Waymo's impressive 100% annual growth sustained for several years, they'll reach perhaps 0.2-0.3% of passenger travel by 2030.

This isn't criticism of Waymo's achievement—going from zero to commercial viability represents a genuine technological triumph. It's recognition that "viable technology" and "urban transformation" operate on vastly different scales and timelines.

China faces a similar calculus. With more cars, but each traveling fewer kilometres, China experiences about five trillion passenger-kilometers annually,[iv]  Baidu's comparable service also represents roughly 0.003% of passenger travel. These are meaningful businesses serving real customers. They are on the cusp of reshaping urban mobility systems.

The Constraint Nobody Modeled

Here's what almost every forecast missed: the binding constraint on robotaxi growth isn't manufacturing vehicles or mapping cities. It's how fast cities can adapt curb infrastructure and traffic management systems to handle robotaxis at scale.[v]

Waymo can double its fleet annually. Few cities can double their curb management capacity annually. Municipal pickup/drop-off coordination, traffic flow optimization, and curb allocation policies require planning processes, community engagement, regulatory frameworks, and physical infrastructure changes. These move at institutional speed, not venture capital speed.

Cities that need to prepare for 5-10% robotaxi penetration in 3-5 years have been conditioned by years of false alarms to deprioritize that preparation. The planning cycles required to adapt urban infrastructure can't start "when robotaxis arrive"—they need to start long before. And years of unfulfilled promises have systematically undermined the credibility needed to trigger those planning cycles.

What Credible Forecasting Actually Looks Like

After fifteen years of failed projections, who should cities trust? Only entities with transparent commercial operations and realistic denominators.

Waymo and Baidu have earned credibility—not because their future predictions will prove accurate, but because they're solving real problems at commercial scale. When they project expansion timelines, those projections rest on operational data rather than PowerPoint optimism.

But even credible operators deserve skeptical questions. When someone projects "billions in revenue" or "millions of vehicles deployed," planners should be asking: what percentage of passenger-kilometers does that represent? If they can't answer, they're selling machines, not transportation solutions.

For cities trying to plan, here's what evidence-based forecasting suggests: If Waymo maintains 100% annual growth for three to four years then growth declines gradually toward automotive industry norms of 1-2% annually, and if two or three competitors emerge with similar trajectories, combined robotaxi penetration might reach 10% of U.S. passenger travel between 2035 and 2042.

But such a national statistic masks where the transformation will actually occur. The first 10% of national robotaxi penetration will concentrate entirely in early adopter cities—e.g., the 100 largest U.S. urban areas that account for roughly 20% of its total population. This means that when robotaxis reach 5% national penetration (which could happen as early as 2030), these major cities will already be the focus, experiencing 10-15% penetration or higher.

The planning challenges, curb management issues, and equity questions won't arrive gradually across a decade—they'll hit the largest cities within the next five years. Cities, such as San Francisco and Austin, confront these challenges today.

This concentrated urban impact fundamentally changes the planning imperative. A city planner in any of these larger metropolitan areas can't wait for 10% national penetration post-2035. They need to prepare now for 10-15% municipal penetration in 2-5 years. The transformation timeline for urban planners isn't next decade, its already here.

Ridership penetration in China could move somewhat faster given government support, with major cities perhaps reaching 10-15% local adoption even sooner. Europe, starting from near-zero and facing stricter regulations plus strong transit alternatives, presents a more complex picture—a focus on roboshuttles-as-transit could enable some EU cities to reach significant penetration despite these obstacles.

The estimates I offer here carry only moderate confidence—perhaps 50-60%—because robotaxi forecasting history offers so little credibility. But they're grounded in actual performance data rather than aspirational S-curves, and they recognize that urban transformation happens in cities first, not uniformly across entire nations.

What Cities Should Do Now

The autonomous vehicle industry spent fifteen years teaching the wrong lessons. It focused on vehicle intelligence over urban systems. It promised transformation on venture capital timelines while ignoring municipal planning cycles. It measured markets in dollars while cities needed answers in passenger-kilometers and infrastructure requirements.

Most destructively, it cried wolf so many times that city planners learned to ignore the warnings. Now, as working technology finally emerges, we face the consequences of squandered credibility.

Cities should not wait for "proof" that robotaxis have arrived at scale—by then, it's too late to plan and budget infrastructure. They should not assume smart vehicles will handle coordination at the curb—self-managed PUDO leads to curb chaos, as early adopter cities have already experienced. And cities should not trust forecasters without an operational track record, no matter how sophisticated their models appear.

Instead, cities should demand specific answers: What percentage of our passenger-kilometers will shift to robotaxis? on what timeline? and what curb infrastructure, traffic management systems, and regulatory frameworks do we need at each threshold? Cities must start planning now for 3-5% penetration in 5-10 years, recognizing that planning processes take nearly as long as technology deployment.

The gap between working technology and reshaping 10% of motorized travel remains vast. We should be deeply skeptical of anyone claiming to know exactly when it will close. But we should be equally skeptical of the opposite error: assuming that because past predictions failed, cities can safely ignore preparation.

The wolf has finally appeared. Cities that learned to ignore the warnings because of years of false alarms will pay the price in congestion, inequity, and lost opportunities to shape how this technology serves public purposes rather than simply investor ROI. The question is no longer whether robotaxis will scale—it's whether cities will be ready as they do.

REFERENCES

[i] McKinsey Center for Future Mobility (2024) "Autonomous vehicles moving forward: Perspectives from industry leaders." https://www.mckinsey.com/features/mckinsey-center-for-future-mobility/our-insights/future-of-autonomous-vehicles-industry-2024

[ii] McKinsey Center for Future Mobility (2026) "Where to next? Insights from autonomous-vehicle experts." January 5, 2025. https://www.mckinsey.com/features/mckinsey-center-for-future-mobility/our-insights/future-of-autonomous-vehicles-industry

[iii] https://enotrans.org/article/americans-drove-1-0-percent-more-in-2024/ [iv] Ma, D.; Wu, X.; Sun, X.; Zhang, S.; Yin, H.; Ding, Y.; Wu, Y. (2022) The Characteristics of Light-Duty Passenger Vehicle Mileage and Impact Analysis in China from a Big Data Perspective. Atmosphere 2022, 13, 1984. https://doi.org/10.3390/ and China Statistical Yearbook (2024) 16-13 Passenger-kilometers by region (2023) https://www.stats.gov.cn/sj/ndsj/2024/indexeh.htm

[v] https://www.intertraffic.com/news/smart-mobility/urban-traffic-and-robotaxi-orchestration