Post 58: Robotaxis Revisited - Part 2

Autonomously driven taxis (AKA “robotaxis”) are currently popping up in cities around the globe and I have little doubt that the robotaxi expansion will continue for the foreseeable future.  This observation, however, begs two important questions:  (1) How fast will this expansion take place? and  (2) What impact will this expansion have on urban life and urban form?  The answers, of course, are not precisely known and the forecasts vary widely.

Elon Musk, for example, thinks that there will be millions of robotaxis from Tesla alone on the road by 2026.  I think that hallucinogens were involved with that particular prediction, but he has been wrong so many times in the past that he is bound to be right sooner or later.  Optimists see robotaxi expansion as a revolutionary advancement in transportation technology that will transform our lives for the better.  They see robotaxis as so prevalent, so reasonably priced, and so useful that families will give up their second or third cars, and perhaps eventually all of their privately owned cars.

Others, however, see robotaxis in more apocalyptic terms.  They see hordes of empty vehicles clogging city streets in search of their next rider, similar to zombies wandering the streets searching for their next victim.  To them, robotaxis are a “solution” that only makes the problem of urban transportation worse.  They worry that the initial public excitement over robotaxis may be the final nail in the coffin for mass transit systems, but that robotaxis will inevitably fail to deliver an affordable and scalable option for a large swath of urban residents.

My goal with this post is to bring some sanity to this topic.  Yes, there are distinct advantages that robotaxis bring to the table that will continue to fuel their expansion.  But there are also clear constraints that will slow that expansion and limit its eventual penetration of the urban transportation market.  Those advantages and constraints will mutate over time, but it will not be a particularly quick process.  As I have pointed out before, urban forms and systems (and human behavior) change slowly.

Near Utopia

The key for those who see explosive expansion is a rapid decline in the price per mile for a typical robotaxi ride.  Optimists think that the average cost per mile could be as low as $0.25 to $0.50 by 2030, or soon after. [1]  That is a significant savings over the cost of a typical Uber or Lyft ride which is currently estimated to average $1.25 to $2.00 per mile (and often much more in congested cities, or during surge pricing periods).  The aggressive cost per mile estimates are based on falling vehicle costs as mass production kicks in, the lower maintenance costs and longer life spans of electric vehicles typically used for robotaxis, and the absence of a human driver.

Volkswagen Robotaxi Prototype

No wonder that robotaxi evangelists foresee the rapid replacement of human-driven Uber rides with autonomously driven rides.  In fact, at the cost per mile rates estimated above, a robotaxi trip is arguably cheaper than driving your own car, particularly if you have to pay for parking at the trip destination. If robotaxi trips start replacing a broad range of trips in privately owned vehicles then the potential market does indeed explode.

Broader mobility.  The most obvious advantage of this Utopian vision is that many people who can’t (or shouldn’t) drive their own car suddenly have a reasonably priced and highly flexible transportation option.  The elderly, teenagers, people with certain disabilities, or even the poor may find that robotaxis give them a level of mobility that is far better (more precise, more private and less stressful) than their current options.

Cleaner and quieter.  Since most robotaxi rides would replace a ride in an internal combustion vehicle with one in an electric vehicle, urban environments would benefit from less ground-level pollution and less vehicular noise.  Although electric vehicles are not pollution free, there would likely be a reduction in total greenhouse gas levels.

Better land utilization.  If private car usage drops, then that also means that parking demand would drop as well.  As I have pointed out in previous posts, most urban areas need a greater level of tax revenue per square mile of service area in order to achieve a reasonable level of fiscal stability.  Replacing low-value surface parking lots with more intensive, high-value buildings would be an important step in the right direction.

Productive time.  City driving is generally more stressful than enjoyable, so a competent autonomous driver would free us to do other things instead of navigating through a chaotic urban environment.  People might choose to send emails or texts, check their social media accounts, or just relax by listening to music or playing a game.  Whatever option you choose, a robotaxi ride could recapture time that previously was lost to driving.

Paradise Lost

Robotaxi naysayers see just another variation on our love affair with the automobile, and arguably one which is worse for urban life than what we have now.  After all, robotaxis will simply replace a human-driven trip in a car with a computer-driven trip in a car.  There is no particular reason to think that the number of occupants per vehicle will suddenly go up, and cautious robo-drivers might actually slow vehicle throughput down.  In addition, robotaxis are likely to cause the total vehicle miles traveled (VMT) to go up which will worsen current congestion levels.  A robotaxi trip (or any ride-hailing trip) isn’t just from my house to my destination, it is from some random starting point to my house and then to my destination.  This increase in trip length (not to mention empty trips waiting for the next rider) might not be a problem during off-peak hours, but it is a potential disaster during rush hours in dense urban areas. [2]

What big cities desperately need is a transportation option that increases passenger density and robotaxis do not currently offer that type of solution.  The core problem is that urban transportation is inherently “peaky” and concentrated.  People do not move from random point to random point at evenly distributed times throughout the day.  There are distinct peaks where transportation demand is focused on certain times (e.g. rush hour) or certain places (e.g. an office district, a factory, a school, etc.).  This pattern creates an obvious physical problem – not enough street space to accommodate vehicles carrying just one or two people – but it also creates a financial problem that undercuts robotaxi profitability.  

If there are enough robotaxis to meet peak demand, then there will be far too many during off peak hours.  Robotaxis will only be profitable if they have high utilization rates, which means that the number of robotaxis that the free market will provide will always be well short of peak demand.  This, in turn, means that during peak hours robotaxis will be in short supply which means that surge pricing will kick in which means that the magical $0.25 to $0.50 per mile price point goes out the window for the most important times of the day and the busiest places.  Thus robotaxis will be just numerous enough and just inexpensive enough to stay busy the vast majority of the time, but not numerous enough or cheap enough to be a paradigm-shifting transportation option for the next decade or so.

In addition, a robotaxi is not a good fit for many of the trips that we take on a daily basis.  Do I really want a robotaxi to drive me to the supermarket, come back in 20 minutes to pick me up and take me to the dry cleaners, then inch through the drive-thru lane at Chick-Fil-A to get some lunch, and then wait patiently at the curb while I carry all my stuff into the house?  And if my trip includes small children, do I have to carry car seats with me and figure out how to install them in various robotaxi vehicles?  Not ideal.  The fact of the matter is that most of us treat our cars like a combination of a rolling storage locker and a school bus – we haul things and people in all sorts of combinations and situations, some of which we can’t always anticipate when we start our trip.  For example, I have my golf clubs, shoes and an assortment of hats in my trunk from March to November in case a driving range and a free hour of time happen to coincide.  Despite its faults, a personal car gives people a great deal of flexibility that reliance upon a robotaxi would not.

One last problem to burst the robotaxi bubble:  think of the curb space chaos in front of the departure gates at most major airports, with vehicles stacked two or three deep as people jump out and try not to get run over.  Now extrapolate that mess to every office building, factory or school as hundreds of robotaxis fight for space to drop off or pick up their riders all at the same time.  Airports are just now figuring out how to handle pick-up and drop-off lanes in a rational manner, but cities are nowhere near ready for that kind of transportation adjustment.  Curb space is already in short supply because of demand for parking spaces and delivery zones – robotaxis will be queued up around the block waiting for what limited curb space exists to become available for their needs or people will get frustrated and disembark in the middle of the street.

The Likely Middle Ground

In order to get a realistic grasp on what the robotaxi world will be like over the next five to ten years, I think it is important to keep three things in mind.  First, autonomous driving is really hard.  This is a problem that the smartest minds in the business thought would be solved ten years ago and yet true autonomous driving (known as SAE Level 5) is not available in any production car or in any operational robotaxi. [3]  The best that is currently available (and likely to be available for the next few years) is Level 4 which provides autonomous driving without any user engagement, but only in limited circumstances.  

This is why all robotaxi services to-date operate in limited geographic areas that have been extensively mapped and carefully vetted so that vehicles avoid confusing or dangerous routes.  This approach, known as “geo-fencing,” is why a Waymo ride in Phoenix won’t take you to some of the outer reaches of the metro area and why the route will often avoid using interstate highways even when that would be most efficient.

Second, no one is making money providing robotaxi services and no one is likely to make money for several more years.  This is why all the leading players are huge corporations with very deep pockets that can afford to lose money while they perfect their service and build market share.  This is an industry that is going to experience serious consolidation at some point in the future which is why the big players are desperate to have enough market presence to remain standing when the mergers and bankruptcies start to happen.

Eventually, of course, shareholders and corporate activists are going to insist that robotaxi services turn a profit.  This pressure to be profitable is why, in my opinion, the rosy forecasts of average ride costs below $0.50 per mile are mostly fantasy for at least the next five to ten years.  The pressure to be the absolute lowest cost provider is not something that happens in a young, rapidly growing industry, but rather in a mature industry where growth is hard to come by.  Thus I expect the cost for robotaxi rides to hover just below the cost of a human-driven Uber ride until robotaxi services are routinely profitable.  This is low enough to cannibalize business from human-driven ride-hailing platforms but not so low that people will ditch their cars for the robotaxi lifestyle.

Third, the regulatory environment is a complicated, patchwork mess.  Some states have lenient laws that favor autonomous driving, but most either have very strict regulations, vague regulations or regulations that include limits which are unfriendly to robotaxi businesses.  Major cities are generally interested in being part of the “leading edge” so that they are seen as progressive and tech-savvy, but they are absolutely paranoid about being so compliant that they become the “bleeding edge” where the blood is literally from their citizens.  As a result, no jurisdiction is anxious to cede control to a higher level of government unless mandated by State law.  No one is fast-tracking an overriding Federal standard for robotaxi regulations because most local governments want to remain “hands on” in protecting the safety of their residents.

This means that robotaxi expansion is going to be a slow (and expensive) city-by-city slog rather than a tidal wave that washes across the country.  It also means that expansion is likely to be labor intensive because cities where service is new are going to want plenty of personnel on hand to (1) deal with operational glitches that block traffic, (2) provide tele-operators monitoring each ride so that citizens feel safe, and (3) keep vehicles clean so that the ride experience is as positive as possible.  This is great for making the rollout of robotaxi service seem successful from a politician’s point of view, but it is another obstacle on the road to profitability.

The Impact on Cities

The most obvious initial impact is that robotaxis will steal rides away from human-driven taxis and riding-hailing apps.  Uber, Lyft and other ride-hailing companies will be willing participants in this process because their main source of profit comes from being the platform that arranges the ride and processes payment – they have no innate preference for humans versus computers when it comes to who is driving the vehicle.  How fast this transition takes place depends on a variety of factors, but mainly upon how quickly costs per mile can fall so that robotaxi services can turn a profit, or at least break even.  My guess is that profitability is still a few years away so the transition will be relatively slow to keep losses from becoming so huge that they are unsustainable.

Still, in the next two or three years I expect there to be robotaxi operations in 15 to 20 cities in the U.S. alone.  There will be four or five major players (perhaps more if the Chinese can break into the market) and the total robotaxi fleet will be in the tens of thousands of vehicles.  That may sound like a lot, but it is really just a drop in the proverbial bucket.  Uber by itself is currently estimated to have roughly a million U.S. drivers.

If my guess is correct, then the initial impact on most urban areas will be fairly minimal.  Robotaxi service won’t expand the ride hailing market substantially until the cost drops significantly.  If the cost does drop substantially (and this is a big “if”), the result will be a dramatic increase in robotaxi trips which in turn will increase total vehicle miles traveled and overall congestion levels.  Depending on where this happens, the result could vary from mildly annoying to disastrous.  Think of the impact, for example, of adding thousands of new trips to the Chicago loop if a door-to-door robotaxi ride becomes cheap enough to woo the more affluent commuters off of the “L.”

The trick will be figuring out at what point on the path of “falling average-price-per-mile” will people shift in significant numbers from trips in their own personal car or trips on mass transit to a robotaxi trip.  That tipping point could potentially have a huge impact on urban life as some of the negative side effects discussed above (e.g. increased congestion and insufficient curb space) become readily apparent.  At that point there is likely to be some public pushback on the robotaxi movement, quite possibly leading to local regulations that limit the number of allowed robotaxis (or ride-hailing services generally) and their allowed distribution throughout the metro area.  If that happens, the market will raise the price of robotaxi rides that now are under-supplied relative to demand.  This in turn will scale back demand, but the damage to urban transportation infrastructure may already have been done.  If middle-class commuters are lured into robotaxi rides from the subways and buses, will they ever return to mass transit even if costs and travel time creep back up?  Most mass transit systems are on thin financial ice already so even short term shifts could be catastrophic.

Cities (or metro areas) will be searching for the right balance, but are likely to find that the right point is hard to pin down and shifts over time.  Unfortunately, people may come to view robotaxi service as the urban transportation version of Pandora’s Box, with “evils” being released before the lid can be slammed back into place.  The hard truth is that most urban areas during rush hour would be hard pressed to handle even a 10 percent increase in VMT without commute times spiraling to unacceptable levels.

Another Option

There is potentially another path that would allow substantial robotaxi expansion without as many negative side effects.  As mentioned above, what urban areas need are transportation options that increase passenger density – in other words, more people being moved per lineal foot of available street lanes.  In the U.S., nearly 75 percent of all commuters drive alone which is 15 to 18 feet of lane space (ignoring the space between cars) per person.  The average for all automotive trips is not much better at roughly 1.5 occupants per vehicle (or approximately 10 to 12 feet per person).  A bus on the other hand, might be just 1 to 3 feet of lane space per person.  Ride-hailing vehicles are currently no better than other automotive trips (excluding the driver).  Precise numbers aren’t available, but estimates are that only 10 percent of Uber rides have three or more passengers.  But what if cities could push robotaxi companies (and ride-hailing companies generally) to increase average passengers-per-trip numbers?

Zeekr Robotaxi Prototype

Current designs for custom-made robotaxis such as the Zoox (Amazon) or Zeekr RT (Waymo)  are 12 to 15 feet in length and can comfortably accommodate 4 people.  If occupancy levels during peak times (or in peak areas) could average 2.5 or 3, then lane space per person might fall to 4 to 6 feet.  Doubling or tripling passenger density versus privately owned cars would be a potentially huge benefit in terms of roadway congestion.

I know what you’re thinking:  “No way am I sharing my robotaxi ride with a stranger,” but hear me out.  Obviously many robotaxi trips will continue to be similar to a typical riding-hailing trip where you (or you and a friend) go from point A to point B, but I think there are a variety of scenarios where a robotaxi “trip” could include multiple stops and/or include passengers that you either don’t know at all or know only tangentially.  

Uber is already experimenting with this type of service with an option called Route Share, which is currently available in half a dozen cities.  This service, targeted toward commuters, has a vehicle running a fixed route every 20 minutes during the morning and evening rush hours.  Riders can reserve a space and select their pick-up and drop-off location.  Since it is not a door-to-door service, riders may need to walk at either end of the ride, but in return they can save up to 50 percent off the cost of an UberX. [4]  Although it is being tested now with human drivers, this service would be ideal for an autonomously driven robotaxi since the route could be mapped in great detail and optimized for safety and efficiency.

I envision a more flexible service that is actually door-to-door so that it feels less like a bus ride and more like a limousine service.  It should take advantage of the ride-hailing platform’s computing ability to optimize routes, adjust for exceptions, and handle billing.  What if, for example, your employer joint-ventured with a robotaxi company to provide rides to work.  You could pick which days you wanted to participate and approximately when you would like to arrive.  The robotaxi’s computers would find other employees near your home who also want a ride, select the best route, and notify you of your pick-up time.  Since each trip would be limited to three or four people, the added inconvenience would be minimal and your cost per trip would drop substantially – and perhaps even be subsidized by your employer as an incentive to return to the office!  Need to stay late for an emergency meeting or leave early because your daughter is sick at school?  No problem; just cancel your spot on the shared ride, arrange for an individual ride, and pay a few bucks more.

There are a wide variety of situations in which this “super car-pooling” approach to robotaxis would be useful.  A retirement community, for example, might arrange a ride that drops a group of seniors at the shopping mall and returns an hour later to pick them up.  Parents might arrange for their 14 year old son and two of his teammates to be picked up from Middle School and taken to swim team practice (including sending a video clip to verify safe drop-off).  Or perhaps you arrange for a robotaxi to take you and your BFFs out clubbing, and then return at 1 AM to take you home.  Since virtually everyone has a smartphone, these rides would be easy to set up, easy to adjust and easy to pay for.  Phones could be used to notify you when your ride is approaching so you don’t forget or keep others waiting, and to verify identities so that your fellow riders are vetted by the robotaxi company.

The real payoff for sharing your ride, of course, is that you might actually approach a cost-per-mile threshold where ditching your car saves you money without imposing a great deal of inconvenience.  Plus there are the added bonuses of being able to relax in a comfortable seat, avoid the stress of urban driving, go online while traveling, and skip the hassle of searching for a parking spot.  The end result might be a transportation option that fills the gap between driving your own car and taking mass transit.  It will not, however, meet so many people’s needs that robotaxis take over the world – but it might make their expansion mostly beneficial rather than mostly detrimental.

The Bottom Line

Robotaxis are coming and my guess is that they will be operating in the vast majority of the top 50 metro areas within five or six years.  This gives cities some time to prepare, but planning for this change needs to start now.  In particular, cities need to ramp up lobbying efforts at the state and national levels to make sure that they retain the ability to impose reasonable restrictions on robotaxi operations.  I don’t favor the ability to ban robotaxis (or autonomous vehicles generally) but some local control is essential.  Robotaxi regulations may end up being similar to the way cities restrict electric scooter-sharing or ebike-sharing businesses.  In areas with an effective regional government, cities might need the ability to cede some operational control because a regional set of rules makes a lot more sense than city-by-city rules.

Cities also need to be able to require data sharing from robotaxi companies.  Details such as pick-up point, drop-off point, time, date and number of passengers will be essential for integrating robotaxis into transportation systems and infrastructure.  Locational data can be generalized to protect privacy, but the more data cities have the more likely robotaxi expansion will be positive rather than negative.  Robotaxi companies are likely to resist by claiming this data is proprietary or that it violates their customers’ right to privacy, but cities need to insist on data sharing even if they have to sign some type of nondisclosure agreement.

Cities also need to insist on clear lines of communication with robotaxi operations.  There will inevitably be problems with the rollout of this type of technology, but effective communication channels can minimize major issues.  As self-driving software improves, there will be less and less need for human monitoring but robotaxi companies need to demonstrate that their vehicles can be safely integrated into an urban area before that happens.  Police Departments, in particular, need to have communication protocols in place so that accidents and stalled vehicle incidents can be resolved quickly and effectively.

In the long term, cities need to build the ability for transportation infrastructure to share data back to robotaxis and other autonomously driven vehicles.  Congestion levels, signal timing data, accident locations, and construction zones are examples of types of information that vehicle computers will eventually be able to process in order to optimize route and speed decisions.  A truly “smart city” will share as much data as possible as transparently as possible.

I am generally an optimistic person when it comes to new technology, but I must admit to having some concerns when it comes to robotaxi expansion.  I fear that an unfettered rollout could lead to an unmitigated disaster.  The potential is high but the dark side can’t be ignored.

Notes:

1. Kyle Harrison; “The Trillion-Dollar Battle To Build a Robotaxi Empire;” April 2025; Contrary Research; https://research.contrary.com/deep-dive/the-trillion-dollar-battle-to-build-a-robotaxi-empire

2. Jeral Poskey; “The unseen environmental costs of autonomous cars;” February 2025; Smart Cities Dive; https://www.smartcitiesdive.com/news/robotaxis-environmental-costs-ghg-sustainability/740947/

3. “SAE Levels of Driving Automation Refined for Clarity and International Audience;” May 2021; SAE International; https://www.sae.org/blog/sae-j3016-update

4. “Introducing Route Share: Uber’s More Affordable, More Predictable Commute”; May 2025; Uber Blog; https://www.uber.com/blog/route-share/#:~:text=What%20is%20Route%20Share?,with%20more%20cities%20to%20come.

Post 57: Robotaxis Revisited - Part 1

 Autonomously driven taxis – generally referred to as a “robotaxis” – are nonexistent in midwestern and most east-coast cities, but they are a relatively common sight in San Francisco, Phoenix, Las Vegas and a few other test-bed locations.  I actually rode in a Waymo robotaxi in early 2023 and wrote about it a couple of months later (Post 34).  Those guinea pig cities have allowed the robotaxi companies to work out the details behind turning an interesting experiment into a viable business, and now the stage is set for expansion across the country and around the world.

Zoox Robotaxi Prototype

There are wildly different visions for how that expansion will play out and for the impact robotaxis will eventually have on urban life and the form of our cities.  Elon Musk, for example, has predicted that:

“By the middle of next year, we’ll have over a million Tesla cars on the road with full self-driving hardware, feature complete, at a reliability level that we would consider that no one needs to pay attention.”

And:

“From our standpoint, if you fast forward a year, maybe a year and three months, but definitely next year for sure, we’ll have over a million robotaxis on the road.” 

That prediction took place in April of 2019 and Tesla has still not initiated any public robotaxi service although they are apparently close to doing a test service with 10 to 20 cars in Austin, Texas.  In addition, the Tesla cars that ship now with autonomous driving hardware and software are not certified for driving without the driver being ready to take back control at any moment.  Elon Musk, of course, has made so many outlandish predictions that never came true that he isn’t a very reliable guide to the future. [1]  My goal with this post is to give you a factual assessment of where the robotaxi industry stands and what is likely to happen with its expansion over the next few years.  In part two of this post I will look a little further out to try to predict the long-term impact of robotaxi technology on urban transportation and urban form.

The Players

The companies that are making the biggest waves in the robotaxi industry have shifted somewhat from my previous post.  In particular, Cruise (General Motors) has apparently dropped out of the business entirely after a checkered experience with robotaxi service in San Francisco.  In addition, several Chinese companies have become major players in this market, particularly overseas.  I’m going to focus here on the U.S. market primarily, although I will mention other markets simply because what happens there will eventually affect the U.S.  This is also not going to be an exhaustive list of companies with their fingers in the robotaxi pie.  Instead, I want to emphasize the companies that I think are going to shape the future of robotaxi service and the ensuing impact on urban form.

Waymo (Alphabet).  The clear leader in domestic robotaxi service is Waymo, and the company is starting pilot programs in Japan and France with an eye toward building a worldwide presence.  While many competitors are still in the testing or limited service phase, Waymo has been providing paid service to the general public for several years in a variety of cities.  The Waymo fleet of robotaxis has now exceeded 25 million miles of autonomous driving and currently provides well over a million paid rides per month.  Operations are currently limited to San Francisco, Phoenix, Los Angeles and Austin, but Waymo has plans to expand to 10 additional cities over the next couple of years including Atlanta, Miami and Washington, DC.

Waymo adapts other production vehicles for robotaxi use rather than build its own custom robotaxi.  The Jaguar I-Pace has been the vehicle of choice for the past few years, but the Hyundai Ioniq and a Chinese vehicle from Geely are apparently in contention for the next generation Waymo product.  Each vehicle is fitted with numerous cameras, as well as Lidar and Radar sensors.  In the near term, Waymo expects to expand their fleet from 1,500 vehicles to 3,500 vehicles, but the eventual goal is to produce “tens of thousands” of robotaxis each year from their conversion factory in Mesa, Arizona.

Uber.  The ubiquitous ride-sharing company has abandoned plans to build its own robotaxis in favor of partnering with a variety of robotaxi producers in the U.S., Europe, Asia and the Middle East.  Waymo’s expansion into Austin, for example, is being handled through Uber instead of Waymo’s own ride-hailing app.  The Waymo/Uber partnership will expand to Atlanta later this year.  Uber has the advantage of having the most widely used app and millions of daily customers which simplifies expansion plans for robotaxi producers.  

Uber not only provides the internet platform, it will also handle fleet operations, including cleaning, maintenance and charging.  This partnership approach makes so much sense that I expect Uber to expand its robotaxi operations as rapidly as possible.  Uber CEO Dara Khosrowshahi has stated that the roughly 100 Waymo vehicles in Austin are busier than over 99% of Uber drivers in terms of trips completed per day.

Zoox (Amazon).  The approach that has been taken by Amazon-owned Zoox focuses mainly on a purpose-built vehicle rather than on adapting traditional cars, although it initially tested its autonomous driving software using Toyota Highlanders.  The all-electric, bi-directional, 4-passenger Zoox vehicle has no steering wheel or control pedals so it can’t be used for anything other than robotaxi service.  However, it does have wide doors, comfortable seats, 4-wheel steering, and high-tech interior controls which make the riding experience as pleasant as possible.  

Zoox has been doing intensive testing in Las Vegas, San Francisco and Foster City, California for several years and is reportedly near to providing public service.  They recently signed a deal to be the “official robotaxi partner” for Resorts World in Las Vegas, although it is not entirely clear exactly what that means.  In any case, Zoox has a unique product but needs to get real-world, paid-passenger experience if it is to keep up with the competition.

Tesla.  The wildcard in the robotaxi world is Tesla which hasn’t actually provided any public service, but which has sold millions of cars, many of which included their “Full Self Driving” (FSD) software and hardware package.  FSD is somewhat misleadingly named because it doesn’t actually provide fully autonomous driving – the human behind the wheel is still expected to take back control in certain situations.  Robotaxi service is expected to start this month in Austin, Texas, although it is likely to be just 10 to 20 cars operating within a tightly limited area and closely monitored with remote “teleoperators.”  Still, if the debut goes smoothly, Tesla hopes to be able to ramp up operations quickly and many analysts expect it to be a major player in the next few years.  There are, unfortunately, question marks galore surrounding Tesla’s robotaxi capabilities because it has been so hard to separate the hype from the reality.

In my opinion, the odds of Tesla rapidly becoming a dominant robotaxi service are about equal to the odds that its service falls flat on its face.  The more likely middle ground is that Tesla eventually masters the intricacies of providing robotaxi service, slowly gains traction in the robotaxi race, but ends up being just one of many providers.  The idea that current Tesla owners will convert their cars to part-time robotaxis strikes me as complete fantasy, and the company is years behind Waymo and other providers in building the corporate capacity to run robotaxi operations in multiple cities.

Tesla Cybercab Prototype

Tesla’s strength is their manufacturing capability and they have unveiled a slick, two-seat “Cybercab”product that is expected to enter production in 2026.  The initial robotaxis in Austin will likely be adapted Model Y vehicles, however, and the actual timeline for Cybercabs to play a significant role in the robotaxi service is up in the air.

The Chinese.  The U.S. has made it nearly impossible for Chinese companies to compete directly in this country’s robotaxi market for reasons that range from national security and data privacy concerns to global trade tensions.  That translates to less publicity in the American press, but it does not mean that they are lagging behind American technology.  Companies such as Baidu, Pony.ai and WeRide have already deployed thousands of driverless robotaxis in various cities in China and are actively expanding into the Middle East and Europe.  In addition to car-like vehicles, Chinese companies are leaders in driverless trucks, buses, and specialty vehicles such as streetsweepers.  Although their direct impact in the U.S. may be muted, their global aspirations means that Waymo, Tesla and Zoox may find it difficult to dominate the worldwide robotaxi market.

Safety

Advocates for autonomous driving technology have long predicted that computers linked with advanced sensors will eventually be much safer drivers than humans.  As robotaxis move from test deployments to a full-fledged public transportation option, the question remains whether they have reached the “safer than a human” tipping point.  Only Waymo has a large sample of real-world rides and a history of safety transparency at this point.  Other companies either don’t have many vehicles in public use or are being very tight lipped about their safety record.  Yet nearly everyone agrees that the public perception of safety is a key to robotaxi success.

A study was recently issued by insurance giant Swiss Re based on the 25 million miles of autonomous driving completed by the “Waymo Driver” and the resulting claims history for both property damage and bodily injuries.  The study included Swiss Re’s data from over 500,000 claims and 200 billion miles of exposure.  It found that compared with human drivers, the Waymo Driver had an 88% reduction in property damage claims and a 92% reduction in bodily injury claims.  Over its 25 million miles of driving, Waymo had just 9 property damage claims and 2 bodily injury claims.  For the same amount of driving, human drivers would be expected to have 78 property damage and 26 bodily injury claims.  This pattern holds true even when compared with late model cars equipped with advanced driver assistance systems such as emergency braking, forward collision warning, or blind spot warning. [2 ]

These results are largely due to Waymo’s extreme focus on passenger safety.  Their vehicles, for example, are typically equipped with 29 cameras, 6 Radar sensors and 5 Lidar sensors.  In addition, prior to providing public service Waymo does extensive testing and mapping of new service areas, and route selection is biased away from confusing intersections or dangerous roadways.

In contrast, Tesla relies primarily on cameras for the FSD system and Elon Musk has derided sensors such as Lidar as a “crutch” and “stupid.”  Although there are some recent indications that Tesla is now using Lidar for its FSD development and testing, their vehicles are still reliant primarily on cameras.  Tesla is just now starting actual robotaxi operations, but their prior FSD systems in the cars they have sold have been linked to a variety of serious accidents, including fatalities.  

Most other robotaxi companies are much closer to Waymo’s approach than to Tesla’s, although no one else has been as transparent with accident data as Waymo.  Still, I think most companies realize that just a few high-profile accidents could put a serious crimp in their business plans, so safety seems to be a high priority for just about everyone.

Profitability

So far, no one is making any money as a robotaxi provider.  Yes, there isn’t a driver to split the revenue with, but building and operating autonomous vehicles is expensive.  The prices charged for robotaxi rides are typically competitive with or slightly cheaper than a typical Uber ride, but each ride is losing money.  How much money is being lost is hard to say because no company is releasing that data, but the estimates are pretty high.

If robotaxis are losing money, why are so many big companies chasing this market?  The answer is that the eventual market is expected to be enormous and the cost of providing service is expected to fall dramatically over time.

Currently, the cost of running a robotaxi business is substantial.  There is the cost of the car with all its cameras and sensors, of course, which most analysts place at well over $100,000 per vehicle.  And then there is the cost of all of the people that are needed for cleaning, recharging, maintenance and monitoring.  Yes, the driver is gone but there is still a lot of labor involved, including “teleoperators” that monitor robotaxi operations.  They don’t generally operate the vehicles remotely unless there is a major issue, but they do track robotaxi operations continuously.  As autonomous driving software improves, the number of monitors will decline significantly but it will not go away any time soon because regulators will likely not allow it until the robotaxi safety record is not just “better than a human” but nearly spotless.

This ongoing need for teleoperations support despite 10 years of autonomous driving software development and testing underscores the difficulty in deploying truly autonomous operation in a complex urban environment.  It is not just hard, it is really hard. [3]  Is the ratio of active robotaxis to teleoperators 10 to 1?  20 to 1?  No one is really saying and it probably varies from city to city and provider to provider.  In fact, the teleoperations ratio is probably a significant negotiating point when providers are seeking approval to operate in a new city or state.  In any case,  it means that the dream of fully autonomous robotaxis is as much mirage as reality.

Plus, there is a lot of overhead involved with ride scheduling software and fleet maintenance, along with mundane stuff like marketing and insurance.  Again, all of this will decline in terms of cost per ride as volume scales up – perhaps to half or even a third of what it is now.  Robotaxi vehicles, for example, might end up costing $40,000 to $50,000 per unit.  But robotaxis rides will never be mostly profit, because while the costs will drop over time, the cost curve will flatten out eventually and it won’t be anywhere near zero.

On the flip side of this equation, just how big is the potential market?  Well, for starters Uber provides roughly 40 million rides per month in the U.S. alone, and Lyft probably accounts for another 10 to 15 million.  Robotaxis won’t be able to steal all of those rides, of course, but I think taking 50 percent is a realistic target five years out in the cities with robust robotaxi operations.  The real gold mine, however, is the roughly 20 billion trips Americans make every month in their cars.  If just five percent of those trips were converted to robotaxis that would be a billion trips per month.  Expand those numbers globally, and the potential market is staggering.

The Bottom Line

Very smart people are investing billions of dollars in the robotaxi race so I have little doubt that Waymo, Tesla, Zoox, et al will be spreading across the country.  Advocates believe that robotaxi service will usher in an era of convenient, safe, and stress-free transportation that will improve both our cities and our lifestyles.  The only questions seem to be when will they get to your city and will you be brave enough to trust them to get you from point A to point B?

Well, perhaps that’s not quite true.  There are actually dozens of unanswered questions surrounding how the robotaxi revolution will play out and it is certainly not clear whether it will lead to urban nirvana or whether we will need to amend Dante’s Inferno to include a tenth circle of hell.  I’m generally optimistic but that optimism is tempered by the knowledge that urban systems and human behaviors are inherently resistant to change.  As always, reality is likely to fall somewhere between nirvana and hell.  Check out part two of this series for my take on what is likely to work and what is likely to disappoint.

Notes:

1. Carlton Reid; “There’s a Very Simple Pattern to Elon Musk’s Broken Promises”; May 2025; Wired; https://www.wired.com/story/theres-a-very-simple-pattern-to-elon-musks-broken-promises/

2. “New Swiss Re Study:  Waymo is safer than even the most advanced human-driven vehicles”; December 2024; Waypoint, Waymo blog; https://waymo.com/blog/2024/12/new-swiss-re-study-waymo#:~:text=It%20found%20that%20the%20Waymo%20Driver%20demonstrated,damage%20claims%20and%20two%20bodily%20injury%20claims.

3. “Fleet response:  Lending a helpful hand to Waymo’s autonomously driven vehicles”; May 2024; Waypoint, Waymo blog; https://waymo.com/blog/2024/05/fleet-response#:~:text=Much%20like%20phone%2Da%2Dfriend%2C%20when%20the%20Waymo%20vehicle,for%20additional%20information%20to%20contextualize%20its%20environment.&text=Fleet%20response%20provides%20the%20Waymo%20Driver%20guidance,street%20and%20make%20way%20for%20the%20truck.

Post 56: Transportation Pricing and Behavior

 All of us realize that there are costs involved every time we travel from one place to another.  Some of those costs are obvious and we think about them almost every time we decide whether to travel or not.  For example, if I am traveling to another city I think about the expense and time involved with flying (e.g. buying a ticket and moving very fast) versus the expense and time involved with driving my car (e.g. buying gas and moving much slower).  

Other costs are more subtle or indirect and we give those a lot less consideration because we tend to form habits around those other costs that save us from having to think about them each time we travel.  For example, if I am going to the store to buy groceries I could drive my car, arrange for an Uber ride, walk to the store, or ride my bike.  All of those options aside from the Uber ride have zero out-of-pocket costs but they do have different costs in terms of time.  So if I’m in a hurry, the time cost of walking may be too great for me to take that option seriously.  But there are also indirect costs like exertion which people might value positively if they like exercise and fresh air, or negatively if they dislike those things or are disabled in some way.  

Driving a car has a boatload of indirect costs such as depreciation, maintenance and taxes, which we all understand that we pay but which are not associated with any given trip so we tend to minimize their impact on our decision.  My point is that while we understand these costs we tend to form transportation habits that eliminate the need for us to think about them for every trip.  Unfortunately, these habits can result in sub-optimal decisions.  On a beautiful day when I am unscheduled, a walk to the store might actually be my best option, but I generally end up driving my car without really considering the walking option because I am in the habit of doing so.

Finally, there are costs that are “hidden” because most people rarely think about them or even connect those costs to their own transportation decisions.  If I drive my car to the store to buy groceries, for example, I add to the congestion of the public streets that I share with other drivers and I add to the pollution of the air which all of us breathe.  These are costs that I impose partly on myself, but mainly on others in the community.  While real, these costs are abstract enough and my contribution is tiny enough that I am likely to write them off as too insignificant to worry about if I even think of them at all.

Although we might ignore most of these costs when we make individual decisions, city governments (and State and Federal governments as well) should be giving all transportation costs a great deal of thought before they make decisions about transportation improvements and their operating budgets.  Unfortunately, governments fall into habits just like we do as individuals which often leads to sub-optimal decision making – and in some cases not just sub-optimal but downright awful.  My goal with this post is to shine some light on transportation costs – particularly direct out-of-pocket costs and time – in order to shift the way we think about the options we have for moving from point A to point B and to potentially shift the way governments set transportation policy.

Place Matters

Before I get into the nitty gritty of transportation pricing, it is important to point out that there is no one-size-fits-all solution.  Your location at the time you are making a transportation decision will have an enormous impact on what decision is best.  A common example in midwestern cities is the proximity to transit.  Midwestern cities are so spread out that the distance from my starting point to the nearest transit stop and the distance from the transit drop-off point to my destination can be so great as to disqualify the transit option as a serious contender.  If the beginning and ending walk is just two or three blocks, then transit might well be a smart choice but the odds of that happening in most midwestern cities is small.

In very large cities such as New York, Chicago or Boston, the cost of parking – both the time-cost of finding a space and the out-of-pocket cost of the parking itself – is so large that driving is far less viable than in smaller cities.  Parking can be so expensive and inconvenient that even owning a car is financially unattractive which means that driving is seldom considered as a transportation choice.  Keep this point in mind as you read the rest of this post.  I’m writing primarily from the perspective of someone who lives in Kansas City, and while that might be relevant to many other midwesterners, it will be less applicable to those who live in much larger or much smaller locations.

Technology Has Changed Pricing

The second caveat to keep in mind is that technology has changed the entire pricing landscape in transportation (and most other industries as well).  The payment platform that has developed around the internet and our smartphones has not only made it more convenient to pay for transportation, it has also changed how much we pay and what exactly we are buying when we travel.  Over the past thirty years, the rise of internet commerce has largely eliminated the role of the travel agent which has reduced the cost of flying by cutting out an unnecessary middle man.  Now smartphones have largely eliminated paper tickets and have enabled the travel industry to shift to “al a carte” pricing.  Paying for checked luggage, snacks, drinks and WiFi access with a few taps on our phones has made possible a concept that previously would have been too convoluted to be feasible.  We can now customize our transportation choices and pay for exactly what we want.  This trend is so powerful it is essentially forcing Southwest Airlines to change their entire business model.

Technology is also changing the way we drive.  Tag sensors, online toll accounts, and now license plate readers are rapidly eliminating the need for toll booth operators and are increasing opportunities for governments to charge for the privilege of driving on certain roads or in certain lanes.  Many locations implemented High-Occupancy Vehicle (HOV) lanes to encourage car-pooling as a way to reduce congestion but those lanes are often underutilized.  So some HOV lanes are being converted to High-Occupancy Tolling (HOT) lanes which use modern electronic tolling systems to allow single occupant vehicles to use the restricted lanes simply by paying a toll.

Even the lowly parking meter has evolved to allow payment via your smartphone instead of requiring you to dig for loose coins.  The change (no pun intended) has not only allowed cities to charge more for parking, it has also made enforcement easier and enabled convenience features such as phone notifications when your time is almost up.  Collecting small payments, such as for metered parking, used to be so inefficient that it often was not worth the effort.  Now, collecting a small payment is just as easy as collecting a large one which opens up new potential revenue streams.

An underappreciated aspect of all of this technology is that it allows dynamic pricing of just about every transportation option.  This changes both how we travel and when we travel.  Pricing that changes instantly in response to demand means that transportation infrastructure can be used more efficiently by smoothing out normal peaks and valleys.  Many airports, for example, are busy in the early morning hours not because people like flying at 6:00 AM but because that’s when the cheapest flights are offered.  The result is a win-win:  people spend less money and airports spread out demand to avoid congestion.

The Hidden Cost of Free

We all like a bargain, and there is no better bargain than free.  Except, of course, that very few things are really free – someone ends up paying and that someone is almost always you.  That is particularly true with transportation choices, some of which appear to be free but really aren’t.  The misconception of free is sometimes beneficial or at least harmless, but it often skews our decision making in ways which are detrimental to ourselves and our cities.

There is a basic concept in economics that states that when people are given unfettered access to a finite but valuable resource, they will tend to overuse it to the point where it is damaged or destroyed.  It is not rational for any individual to exercise restraint because someone else will simply supplant them, but the inevitable result is bad for everyone.  This is often referred to as the “tragedy of the commons” and I wrote an entire post about it roughly eighteen months ago (Post 39).   

Mass Transit.  In the world of transportation, there are three primary examples of things that appear to be free but really aren’t.  The first is the occasional mass transit service that is provided at no cost to riders.  For example, I recently rode Miami’s Metromover which is a 4.4-mile elevated tram that loops through the frequently congested downtown area and a few adjacent neighborhoods.  I found it to be useful, although as a tourist it seemed too limited in its geographic coverage.  The locals must find it more useful because it averages more than 20,000 riders per day which is trending up over the past several years, but still a little short of the pre-pandemic year of 2019.  As I understand it, the system is funded primarily by a special sales tax levy, supplemented by general tax revenue and federal grants.

Similarly, Kansas City has a streetcar line that is also free to ride.  It follows a two-mile out-and-back route that primarily follows Main Street.  Construction is nearly complete on two extensions to that route which will triple its length to just over 6 miles.  Ridership has grown steadily over the past several years and now averages over 4,700 rides per day.  As with the Metromover, Streetcar ridership dropped steeply during the pandemic and is just now approaching 2019 numbers.  Funding comes from additional sales tax and property tax levies within a Transportation Development District that surrounds the streetcar route.

Transit services on a limited route such as Metromover and KC Streetcar may have occasional riders who are simply satisfying their curiosity, but generally they are free of the “overuse” issues that often plague free services.  There is no economic or personal advantage to be gained unless you actually want to travel to some point along the route, which is a naturally limiting set of people.  Despite the surface appearance, however, the rides are not free as I pointed out above.  In theory, the businesses and property owners who pay the added levies required for these transit services get a reasonable return on their investment from increased business or from increased property valuation, although that is certainly up for debate.  As with nearly every tax, my guess is that some reap benefits that exceed what they pay while others fall short.

Public roads and highways.  The second transportation service that has the perception of being free is the use of public streets and highways.  Toll roads and bridges are obvious exceptions, of course, but generally people can drive pretty much anywhere without an out-of-pocket expense.  In contrast to limited-route transit, the free use of public streets does influence behavior in ways that often causes significant congestion and thus damages the value of the street or highway as a public asset.  The phenomenon is known as “induced demand” and I discuss it at some length in the earlier post mentioned above.  I won’t go into detail here but all of us have experienced bumper to bumper traffic that approaches gridlock even on highways that have recently been widened.

Recent events in New York City provide an interesting illustration of the impact of “free” streets versus streets that have an explicit out-of-pocket cost.  In January of this year, New York City implemented a fee known as “congestion pricing” on vehicles entering Manhattan south of 60th Street.  The amount charged is $9 for cars, SUVs and light trucks during peak hours (Monday through Friday, 5 a.m. through 9 p.m.).  The fee falls to $2.25 in off-peak hours.  Specialty vehicles such as large trucks have separate pricing.

I have a modest level of experience with Manhattan traffic because I frequently travel to visit my daughter and her family in Hoboken, New Jersey.  I typically fly into LaGuardia Airport and take an Uber to a hotel near their apartment.  The trip takes me through Brooklyn, into Manhattan via the Midtown Tunnel, across Manhattan at roughly 34th Street, and into New Jersey via the Lincoln Tunnel.  During the day, traffic is heavy virtually the entire trip, but prior to this year, the portion across Manhattan was excruciatingly slow.  The distance from tunnel exit to tunnel entrance is less than two miles but it could often take nearly 30 minutes.  On a trip in late March, the same segment took just over 15 minutes – the difference was amazing.

It is way too early to draw definite conclusions from the congestion pricing program, but early results from the first quarter of the year are extremely promising.  Approximately 10 percent fewer vehicles are entering the congestion pricing zone each weekday, congestion levels are down, and average travel speeds are up.  Ridership levels on mass transit systems leading into the city are also up as would be expected.  What is unexpected is that restaurant reservations, Broadway ticket sales, and general retail sales are also up which means that with congestion down more people are willing to come into the city to do business.  Anecdotally, delivery drivers are wasting less time, emergency vehicles arrive sooner, and complaints about “excessive honking” are down substantially. [1, 2 ] Finally, the system has raised $159 million in the first three months (to be spent on transit improvements), an amount that is within 1 percent of projections.

Parking.  In midwestern cities, parking is typically provided at no direct cost to drivers.  There are exceptions in areas where demand is high such as downtown business districts or event venues, but those exceptions are rare.  The perception that parking is almost always going to be free and plentiful affects our transportation choices – with a clear bias toward driving – although most of us don’t think of parking explicitly when making a trip.  We just assume that parking will be free and readily available at the end of our trip and if it is not we are surprised, annoyed or both.

The result is that city planners, developers and landowners have a difficult time determining the optimal amount of parking to provide.  When a valuable commodity is offered on the open market, the price helps determine the quantity of that product that is needed.  Giving the commodity away for free means that market signal is lost and the result is that both demand and supply are likely to be overstated.  

I have written an entire article on the problems associated with over-supplying parking (Post 52) so I won’t go into great detail here.  The bottom line is that the perception of free parking is a form of market distortion that warps our behavior and harms our cities.

The Collective Good Versus Individual Choice

When we buy a ticket to fly from one city to another, most of us understand that the price of that ticket helps cover the cost of the airplane, the aviation fuel, the pilots, and a variety of other direct costs, as well as leaving a little money left over for profit that is distributed to the airline shareholders.  What most people don’t understand is that commercial air travel is subsidized by the government.  In fact, virtually every form of transportation is subsidized by the government.  Even people who walk to work are walking on sidewalks that someone else paid for.

That subsidy is not necessarily a bad thing.  After all, being able to move people and goods from one place to another is essential to our economy and our way of life.  However, many people have the perception that they are “paying their own way” when they make transportation choices when, in fact, they are not.

There are two issues related to our system of transportation subsidies that deserve more attention than they typically receive.  First, the amount of subsidy varies wildly from one form of transportation to another and from one place to another.  How do we know if we are subsiding in the right amounts?  Unfortunately, transportation subsidies come from so many sources and in so many forms that it would be difficult for anyone to accurately calculate the subsidy per passenger mile or similar statistics – and certainly not the average user.

Second, the indirect costs and benefits associated with each form of transportation also vary wildly and are often difficult to quantify.  Obviously there is a value in moving people and things from place to place and there are direct costs in terms of time and money.  But the deeper you dig into the topic, the fuzzier and fuzzier the issues become.  Mass transit, for example, has been shown to reduce both traffic congestion and air pollution. [3]  How exactly do we place a value on those benefits?  Building bike lanes is thought to provide health benefits, reduce bicycle fatalities, and reduce greenhouse gas emissions.  Is anyone doing the math in a systematic way to balance the subsidies against the advantages?  Not that I know of.

To make matters worse, transit, bike lanes and sidewalks all increase the mobility of people who can’t afford to drive a car or who can’t drive a car due to age or disability.  There is a definite value to both our economy and our society that flows from enabling that segment of our community to be independently mobile.  How much subsidy is that benefit worth?  I have no clue and I doubt anyone else does either.

The Bottom Line

The end result is that all of us make transportation decisions every day without even a rudimentary understanding of the cost (both obvious and hidden) of the options at our disposal.  Most of us, in fact, not only don’t know the answer, we don't even take time to think about the question.  Would knowing the answer change our behavior?  Maybe or maybe not.  The majority would probably choose whatever option seems most convenient regardless of the true price, but there is clearly a significant minority who are sensitive to cost and would adjust their behavior if the system were less opaque.

In addition, there are leaders at every level of government that routinely make decisions on infrastructure improvements and financial support of various transportation systems without a true accounting of the efficiency or fairness of those decisions.  They are practicing what former Yale professor Charles Lindblom called the “science of muddling through” – a decision making process based on making incremental changes to what has been done in the past rather than a comprehensive analysis. [4]  This process acknowledges the complexity of many policy decisions and the limitation of human analytical abilities, but it rarely questions the efficacy of past patterns or whether changes to underlying conditions might warrant a radically new distribution of resources.  Instead, we muddle through from year to year with only small changes at the margins, and even those small adjustments are driven more by the constituencies with the loudest voices rather than careful study.

Given the huge impact that transportation has on city building, this decision to muddle through means that our cities are largely stuck in the same form that they have had for the past hundred years or more.  Think of the enormous changes that have occurred in technology, the economy, and society in general during that time span.  If we were designing cities from scratch would we still copy patterns from a hundred years ago or would we balance transportation options differently?  Muddling through may be politically expedient but I suspect that it is making the goal of strong, resilient cities harder to reach rather than easier.

Notes:

1. Christopher Bananos; “How Well Is Congestion Pricing Doing?  Very.”; April 2025; New York/Curbed; https://www.curbed.com/article/100-dayscongestion-pricing-mta-results.html

2. “Clearer roads, faster trips:  The data behind NYC’s congestion pricing success”; April 2025; TomTom Blog; https://www.tomtom.com/newsroom/explainers-and-insights/the-data-behind-nyc-s-congestion-pricing-success/

3. Robbie Webber; “Contrarian research:  Transit relieves congestion, but park-and-rides do not”; April 2013; State Smart Transportation Initiative; https://ssti.us/2013/04/08/contrarian-research-transit-relieves-congestion-but-park-and-rides-do-not/#:~:text=Michael%20Anderson%2C%20an%20economist%20at,congestion%20would%20continue%20to%20drive.

4. Charles E. Lindblom; “The Science of ‘Muddling Through’ “;  19 Public Administration Review 79; 1959.