The future timeline of robocars -- 2020s land rush, 2030s maturity

There will be a "land rush" on a scale to rival the last one

One of the most frequent questions people ask about robocars is when they will arrive for the public. The answer is difficult and not simple for many reasons, but one of the most important points is that it won't happen at the same time in every places. Some pilot towns will have them very soon; other places may not see them for decades. Predicting dates for future technologies is always extremely risky, so take much of this with several grains of salt, but here is my best guess for the timeline.

My short summary answer to the question is as follows:

  1. The next 2-3 years will see pilot projects in various cities around the world, starting with Waymo in Phoenix, and others in San Francisco, Tokyo, Singapore and a few others.
  2. In the early 2020s one or more pilots will demonstrate safety, regulatory consensus and as time passes, business models.
  3. Once these are established, particularly a working business model, the first player will strike out to dominate the most lucrative and "doable" cities.
  4. The second player will start somewhat later, going mostly after different cities
  5. As more players come online (2023 timeframe) a "land rush" begins as players grab virgin cities. The land rush will cost hundreds of billions of dollars, but the players have that much money
  6. Once the most lucrative cities are claimed (mid-late 2020s) the most lucrative cities become competitive. Other countries with more virgin markets are developed
  7. By 2030, most major cities have some level of service. Major ones have competition. In addition, private cars are deployed in many of these cities.

Independent of this timeline for robotaxis, some car companies will sell cars for private ownership that have self-drive abilities. This may begin on highways but will move to urban and rural street operation. This service begins after the first robotaxi deployments, since it's actually more challenging to build a car to the level where you can sell it to a customer and not have it under your regular control, however it is not too far after, and highway service could begin sooner. This is governed by how quickly the companies can deploy good map and other data services for each territory where driving is to be available.

Land Rush: The no-competition era

One of the most striking realizations is we may see an era of minimal competition in the 2020s. Deploying a fleet of robotaxis to a city is a very non-trivial thing. You need to:

  • Fully map the driving zone
  • Customize the software to local driving rules and unwritten practices
  • Build good relations with the local governments and police, of which there may be several
  • Spend several hundred million dollars on the fleet
  • Hire a local staff
  • Arrange for human drivers for service outside the robotic operation area
  • Contract or set up with local providers for maintenance, charging, fuel, cleaning, parking
  • Do initial promotion -- though PR will be fairly easy
  • In some cases, localize apps and documentation in local languages
  • Put systems in place for updating maps and other databases

For some companies, the hundreds of millions of dollars will be a barrier. For other companies, with ready capital, it will be the management bandwidth that will be an issue. This is not like Uber, which can come to a new city with a tiny staff. Even a company the size of Waymo or GM won't be able to deploy in more than a few cities at once.

Thus, with most of the cities unclaimed, companies will probably prefer to deploy in unclaimed cities, rather than go head to head with a competitor. A few cities will get competition at the start -- probably the San Francisco Bay Area (because so many companies are based there) and Manhattan.

During the no-competition period, the only competition will be companies like Uber and Lyft on the high end and personal cars on the "low" end. Pricing at first will be lower than theirs, but possibly not that much lower without any competitive pressure. Those services run $2/mile. A robotaxi service quickly out-competes them at a bit over $1/mile. Driving a personal car is about 40 to 60 cents/mile, but involves a lot of work and hassle and up-front investment. Riding transit varies wildly in cost and is highly subsidized. Typical transit tickets are $2 to $3 and thus are more expensive on short trips, and in most instances less convenient.

Even at $1/mile, companies will lose money in the early years, not because the raw cost of goods sold is too high, but because they still will have large non-recurring expenses for development, refinement and market building. Uber takes 50 cents/mile from drivers and provides no physical services except insurance for that, and it loses money.

The Land Rush

With all these factors, companies will decide which cities to go after. Of course, they will also evaluate how easy it is to drive in that city -- Phoenix is much easier than Boston -- and the presence of other factors like bad weather, snow, or bizarre driving rules. The density of the city will play a factor, the strength of its transit system and the wealth of its citizens. The friendliness of the city government will also play a factor.

Companies will evaluate all these terms, as well as competition (present or future) and pick the next city on their list, then deploy. As fast as they can, but not too fast.

Getting people out of private cars

Companies will be eager to lose money to claim territory, though. The brass ring is in sight, and worth spending on. They might even price it closer to the eventual prices (which are lower than private car ownership) just to get people used to the idea, and to take them away from private car rides into the robotaxi.

The early years will see pricing experimentation as well. Pricing by the mile is natural, but people with private cars don't pay for car travel that way. They pay a monthly leasing or loan fee, pay for insurance and do maintenance twice a year, licence it once a year and fill up the tank once or twice a week. The only thing they pay per trip is parking -- which you won't see with robotaxis.

Since the long term game involves getting people to give up private car use for robotaxi use, companies may be quite aggressive trying to see what it takes to make that happen.

First "mover" advantage?

There will be a land rush because most people believe in the first mover advantage. They will spend a lot to be the first to deploy in a city in the hope they can keep it. Even though they know that sometimes that doesn't work, they also know that sometimes it does, and it's the only bet to make in the early days.

Even in the ride business, the thing we know of as "Uber" (really UberX) was done first by Sidecar and Lyft in San Francisco. Sidecar is dead now and Lyft is the much smaller company. At the same time, even once Uber came to dominate, Lyft was able to make itself grow faster and become a threat to Uber again.

In the technology world, the first mover often fails. One way it fails is that the second mover comes in with a whole new generation of technology, since generations are so short in this game. So even though Waymo is #1 and will take over several cities, 3 years later they will face somebody with cars designed 3 years later.

Earlier I wrote some detailed analysis of factors in robotaxi competition. When the time comes for competition, there should be many factors to compete on. Even so, being the first and most well known is going to be an advantage people want to have.

There will also be a network advantage. Visitors will be heavy robotaxi users since they have zero chance of owning a car and rarely use transit. Visitors may use the service they already know and have an account with, even if it's not the first mover in town.

The 2030s -- the maturity era

In the United States and a few other countries, the 2030s will be the more mature era, with 3rd generation cars present and competition in many cities. In other parts of the world, things may just be getting started. Once a city gets enough service that people are getting ready to give up private cars, people will sell those cars to folks in cities and countries that don't have robotaxi service.

In addition, rural and exurb areas will not have robotaxi service until much later. They will see deployment of privately owned robocars, but there in many cases the turnover cycle will be slower. They might even be buying the minor glut of used cars from cities taken over by robotaxis.

Technological breakthroughs

The course of this will be altered by the timing of certain technological breakthroughs.

  • Low cost LIDAR seems all but assured for the early 2020s, making vehicle cost reasonable
  • Some day, computer vision will be sufficient, allowing a car to be built with just cameras and radar, which will lower costs even more
  • Some people hope for cars that can "drive without a map" -- meaning minimal mapping and just the use of the type of maps already found in your phone. This would allow deployment to a new city to be much cheaper.
  • eVTOL vehicles -- "flying cars" -- will also start arriving. They will have an obvious superiority for a growing subset of trips, competing with, but also working with the robotaxis. (Because the eVTOL can't land everywhere due to trees, powerlines, noise and other factors, people will fly it to a place where a robotaxi is waiting.)

The 2040s -- you still drive a car?

While human driving will not get completely banned for decades to come, in the mature cities it will become the more odd choice by around this time. Almost like riding a horse was a few decades after the car became mainstream. Certain zones (such as downtown core zones) will have moved to not allow human driving without a special permit.


I have much less understanding of how this will play in China, where the government will exercise much more control. China may designate cities as the first to use robocars, and may even alter local city laws to facilitate. It will probably select a small number of (Chinese) companies which will get licences to operate in those cities. Growth will be large, and eventually those Chinese companies will be ready to enter the land rush in the rest of the world. I am less certain of the dates.

Who wins

This timeline suggests that there is not just one winner. As long as there are virgin cities to deploy in, new companies can continue to spring up. Of course, there can and will be consolidation after that, but probably not as much on the global scale. Countries will consider the robotaxi system to be part of critical infrastructure and enact laws to help local players and prevent total dominance, especially foreign dominance.

This also means that, even though Waymo is very far ahead of everybody else, it is unlikely to win the entire game. At a certain point in the game, their primacy will no longer be a competitive edge. Their maturity will offer them the claim of greater safety, but after a certain point, all competitors will be seen as of adequate safety or they will leave the field. Other competitive factors will dominate.


Two totally unrelated questions: is snow a major impediment to robotaxi deployments? If so, late deployments in rustbelt cities could cause them to fall even further behind the rest of the country.
Along the lines of selling jeans, picks and shovels to the gold miners, what do you see as the business opportunities in this "land rush" aside from the core technologies?

Snow is an impediment in several ways. Mainly, during the "just get it to work" phase most people have not worried about snow, as there are lots of cities without it. However, there has been lots of research on it. Snow creates 3 problems:

  • You can't see anything of the ground or road if it's got fresh snow, so you can't figure out where you are from that. But you can use the vertical edges (in cities) and ground penetrating radar.
  • You have to be able to drive on slippery frozen roads, which means going a lot slower. Humans are stupid and go too fast.
  • If the snow is heavy, it can actually block your sensors in the air, by snowing on them, or even block your movement.

The kicker with snow is this. If you want to offer car replacement, you can't just say, "Too snowy today, not operating!" Unless it's crazy bad snow. That's one of the advantage trains have. And subways even more (at huge cost.)

Many people are already on these pick and shovel businesses. There are companies in all the obvious niches -- mapping, local service, cleaning, maintenance, testing, simulation, sensors etc. I don't know if there are companies specializing in the politics yet, but there will be law firms and lobby firms that do that.

An obvious picks and shovels play is NVIDIA, however as far as the stock price perspective you might have missed the boat on that one.
Personally I don't think snow will present an enormous problem. Cars will rely on millimeter GPS so they will know exactly where the underlying road is. Of course there will be edge cases - plows piling up snow on the side of the road etc. Most of these scenarios will be worked on beforehand in a virtual world (reference: Exciting times...and good article btw..timeline seems realistic.

Nobody wants to depend on GPS. In addition, one of the issues with all the hard localization methods is that while it is good to know where the road is, usually "lanes" get made by human drivers that are not where the real lanes are, but you want to drive in them. Robots, if they make the first lanes, will use the real lanes, I guess.

While they will test in simulator, this is something you need real world for even more. Anyway, nobody thinks it can't be solve, it just has not been on the list for version one for most teams.

There are a hell of a lot of ways that a city will change because of robocars, some of them should be relatively easy to pick.
Lots of the changes are industries that are likely to downsize or disappear.
Car parking operators.
The entire car manufacturing and components industry will suffer as the fleet is downsized both in number and vehicle mass.
Construction companies will suffer from less roading construction, though there may be a temporary up tick if it eventuates that Brad is wrong and that some infrastructure helpful to robocars is useful to speed their adoption, or rail conversion to busways occurs, or conversion of parking buildings is a big thing.
crash repair.
Medical suppliers for A&E.

Companies that should do well out of it.
Amazon, robo delivery vehicles should make delivery cheaper and more convenient.
Electricity suppliers, power plant and transmission companies should benefit from the move to electric vehicles that the robocar helps to hasten.
Pubs and bars.
Events promoters, sports teams, and events that cater to the elderly or the tweens/teens.
Travel companies and other tourist related businesses.
Bicycle manufacturers.
Almost certainly a whole lot more.

"While human driving will not get completely banned for decades to come" - err, wouldn't this be a massive assault on our personal freedom and a hugely regressive step? It sounds like some awful orwellian dystopia.

I'm not sure human driving will be banned at all, but not for many decades if it is. You would be surprised how many people think the banning of human driving is just around the corner.

Brad thanks for your excellent blog, it's become a source of inspiration.
It won't require a change in law to remove human drivers. You can drive uninsured currently in the US is my assumption (guess). Why would any insurance company insure a human driver once the statistics show self driving vehicles vehicles to be much safer. Once you can't secure insurance you will no longer be legally allowed to drive except on tracks set aside for such recreational activities. Other countries have much less issues curtail individual freedoms. I expect Singapore will abolish private vehicle ownership as soon as it's feasible. Cars are already purchased based on a limited number of 10 year licenses that are auctioned. Once it's shown to be possible I could see places like New York City and European Capitals quickly using the precedent set. They will use a rule like you can use your own car but you have to fill the number of available seats in order to do so. Not enough passengers and you will need to hire the appropriate vehicle for your journey

Insuring risk is what they do. Human drivers will not be more dangerous in the future, so they will cost the same, or less to insure as today, and companies are happy to sell insurance to them today.

It's possible some countries might try to ban it. Politically, that's a tall order in the USA and many other places. Maybe not Singapore.

It’s never too soon to sign up for the Human Driving Association:

Add new comment