$16 billion spending on robocars is a drop in the bucket compared to the trillions to be made


Recent coverage summed up robocar spending as about $16 billion to-date. Many have wondered how this can be worth it, since nobody is shipping. When you look at other analysis of how much the winners stand to gain, it's a drop in the bucket. I analyse the numbers in a new article on the Forbes site:

Companies Have Spent Over $16B On Robocars. It’s A Drop In The Bucket


How much of the appeal of air taxis is that it is computationally easier, in a simpler environment? No pesky random pedestrians or stalled vehicles in the airways. Autonomous operation is already a reality for private aircraft (Google "Garmin autoland") So the primary challenge is building a VTOL aircraft that is operationally economical. Whom do you see winning that race?

Well, there is no question that automating basic flight is much simpler. There are still some big issues to resolve, including fully safe landing in high winds, emergency landing in case of airspace shutdown, and airspace management.

However, other problems are larger, including noise, battery weight and public acceptance.

But even though these will fly vastly closer than existing aircraft do, their gaps will still be much larger than cars use, allowing much higher safety margins. Other than birds, there is nothing else up there but other craft. No pedestrians or bicycles or etc.

The trillions are in gross revenues, not net profits, aren't they?

Are margins going to be high? You kind of hit on this when you ask about first mover advantages. I think the big question there is how well the first movers will be able to protect their IP. And I think the answer to that is going to be not very well.

Nobody is making profits of trillions off anything in the world as far as I know.

There will be some winners. And those winners will capture much of this value. This may or may not be the first movers, but people still want to be the first movers.

Margins will be high during the early, non-competition phase. Gross margins anyway. Margins will narrow as things mature and you get competitors in the major markets. But if you can be in a new market at $1/mile or compete in an existing one at 50 cents/mile, you will pick the former.

But if you can be in a new market at $1/mile or compete in an existing one at 50 cents/mile, you will pick the former.

If both are profitable, why not pick both?

What kind of margins are you going to have at $1/mile? And who's going to pay $1/mile?

Margins will be non-existent during the early, funded-by-wall-street phase. And by the time that phase is over, all the car manufacturers will have self-driving tech. And who's going to pay even $0.50/mile for a robotaxi ride to the bar or airport when their own car can drop them off and then drive back?

Nobody is rich enough to deploy everywhere at once. So you pick cities one by one. At first, there are other motives, but eventually the main motive is how profitable is the city.

True margins will be low to start, though gross margins (which impress investors and bosses) will be decent. COGS should be in the range of 15 to 50 cents/mile, so you can have a good margin at $1/mile, even at 30 cents/mile for the cheap vehicle. Once the other costs mature.

It's not going to be everywhere, as some places will be more viable than others for reasons having nothing to do with lack of competition. Some places will have higher demand, or more beneficial laws, or more predictable traffic, or better weather, or whatever. But if you've got a city where you can charge $0.60/mile and only have to spend $0.50/mile (more on that below), and you know you're going to eventually expand to that city anyway, why not deploy there immediately? "Nobody is rich enough to deploy" somewhere where there's a profit to be made? How much does it cost to deploy to a new city? What exactly are you spending money on, anyway? Uber raised $8 billion in its IPO. How many additional cities can you deploy in for $8 billion once you have something that works in one city?

Get something working and making a profit in one city, then raise a bunch of money in an IPO and deploy to as many cities as possible all at once. Obviously you pick the most profitable first. But the most profitable cities are going to be the most profitable for everyone.

I'm not sure what you're counting in COGS, but marginal cost of 50 cents/mile is way too low for the early years of this business, when dead miles will be high (not enough customers), idle time will be high (not enough customers), manpower requirements will be significant (lots of need for human intervention; low car/human ratios), and the cost of the vehicles themselves will be high (costs will come down over time, but initially there won't be a "cheap vehicle").

Even if you do get it down to 50 cents a mile, though, who's going to spend that much on a robotaxi? Not much more than the people who currently are willing to pay for an Uber. And in the long run, even that will dry up, as people will just take their personally-owned self-driving vehicle to go to the bar, or go to the airport, or go to the concert, or whatever. The top reasons that people who own cars use Uber are to get drunk and to go somewhere where it's hard/expensive to park. Both of those reasons disappear when people own a car that can drive itself.

You are thinking of companies like Uber which can expand to a new city trivially -- they just send a team of 2 to 4 people for a few months to set up things. And even then, with oodles of money, they have tons of places they don't operate.

For a service like Waymo One to go to a new city, the costs are very high, and the number of trained people to make it happen is limited. Even with Alphabet's money there is a fairly low limit in the number of new cities you can be bringing on board. You have to map, make nice with the city, learn all the quirks of driving in that city, recode, test, test, test, test, test. Then the usual stuff like marketing and set up local offices and services and depots. And add 5,000 vehicles to your fleet.

So yes, you want cities where the business volume will be high, but if it is, your fixed costs dwindle on a per mile basis, especially amortized over the expected life of your vehicles.

I believe a lot more people are willing to take robotaxi rides for 50 cents/mile than will take Ubers at $2/mile. I guess my big question is, why don't you think that? At that price, a lot of people will drop from 2 cars to 1, and many down to zero. Particularly in areas where parking costs money or is difficult (even if subsidized.)

Note as well that since building a privately owned robocar (other than a highway only one) is much more difficult than a taxi, for a long time you can't get the private car to compete with this.

Uber is maybe on one end of the spectrum (though the cost of complying with new regulations and finding a new batch of suckers to work for rock-bottom wages isn't cheap). But for a service like Waymo One, how much does it cost to go to a new city? How many cities can you go into if you raise $8 billion in an IPO?

I think that would be the way to go, not to try to conserve money up front while your competitors come in and beat you to deployment.

Maybe the answer is that it is very expensive - hundreds of millions or even a billion dollars - for Waymo One to expand to a new city. But if so, that suggests to me that Waymo One has an awful strategy for building a self-driving car. They are relying far too much on infrastructure, if they can't deploy to a new city without spending a billion dollars.

I believe a lot more people are willing to take robotaxi rides for 50 cents/mile than will take Ubers at $2/mile. I guess my big question is, why don't you think that?

I can't think of any reason why they would. The incremental cost of driving the vehicle you already own is much less than $0.50/mile (and the perceived cost is even less than that). (Okay, a couple reasons; see below.)

At that price, a lot of people will drop from 2 cars to 1, and many down to zero. Particularly in areas where parking costs money or is difficult (even if subsidized.)

Who would do that? Where are these areas? There are areas where parking is very expensive/difficult, and cost-sensitive people don't have cars. Maybe in those places people with a short commute will use a robotaxi for $0.50/mile rather than mass transit, though those same places are likely to have to impose congestion taxes precisely to stop people from doing that. And probably within a few years of having robotaxis available you'll be able to buy a car that can drop you off and park itself somewhere where parking is cheap.

Maybe a family with two cars, with one family member that does little driving, and with another family member that does an average amount of driving? Might make sense, though probably not if there are kids in that family. (I wonder what the fee will be when your kid spills his drink or barfs in your robotaxi. Maybe it'll be spread around over all the customers...)

Building a privately-owned robocar is much more difficult than a taxi? I don't agree at all. If you want to narrowly define "privately-owned robocar" to only be a vehicle with no steering wheel, then it's about equally difficult. (There's no reason a privately-owned robocar can't be monitored from a central location, if that's what you were implying.) But the great thing about a privately-owned robocar is that it can be a robocar in some places and an ADAS car in other places. And I don't just mean "highway-only." In fact, as I allude to above, one of the most useful features of privately-owned robocar will be advanced autopark/summon. A privately-owned robocar gives you the best of both worlds. You have a robotaxi when you need it, and a driver-assist vehicle when you don't. The car companies could even charge a small fee per mile whenever you want to use robotaxi mode, and I wouldn't be at all surprised if they wind up doing so.

A private robocar that can only self-drive on things like highways might be easier to build than a taxi. Once you start wanting it to drive the regular urban streets it needs to be as complex as the taxi. But unlike the taxi it has to work and be supported even though owned by a customer and living in the customer's garage rather than coming back to base every night. A car that is sold to a customer has to be super reliable and problem free, a taxi does not. (If a taxi is having problems, you just take it out of service for a while. If a private car has problems, the owner is highly inconvenienced and pissed.)

On top of all that, if the sensor suite costs $5K on the BOM that's adding a lot more to the retail price. For a taxi, that extra cost is not so big a deal due to the much higher utilization.

None of what you said implies that it's any more difficult to build a private robocar than to build a robotaxi.

I can't imagine what type of maintenance a robocar would need that would require coming back to base every night. But a privately-owned robocar could drive to base every night too. There's nothing about private ownership that limits what you can do. (There are things about taxi company ownership that limit what you can do. For instance, you can't park a taxi in places where commercial parking is prohibited.)

There's no need for a private car to be any more reliable and problem free. Pay for the owner's robotaxi fees during the time the private car is out of service, and their only inconvenience is the inconvenience of having to use a robotaxi instead of having their own car. Better yet, have a loaner car drive itself over to them while their car is out of service. This way they don't have to deal with all the inconveniences of having to use a robotaxi to get around.

(Do you think robotaxis are likely to be unreliable and need frequent service?)

Yes, robocars are going to cost more than non-robocars. At least, at first. Over time the opportunities for cost savings (cheaper parking, cheaper insurance, cheaper fuel, more fuel-efficiency, much lower need to ever hire an Uber, the ability to rent your car out when it's not in use, etc.) might outweigh the cost of the sensor suite. And if you can't afford $5,000 extra for a self-driving car, you're probably not the kind of person who's going to be attracted to paying $0.50/mile.

Sure, you can offer a private owner regular maintenance and robotaxi service any time they need it. But you've blurred the lines from the old model of, "I buy a car, it's mine, and I never interact with the manufacturer again" which is how it works for 99% of buyers. That's the thing which is harder than making a taxi service. Hybrid models can work. In fact, you can also do a taxi service that lets you effectively have guaranteed access to a vehicle, the ability to leave your stuff in it all day when you want etc. But these are all unlike traditional car sales.

Will they be unreliable? Because being unreliable is not a problem, they will be unreliable if you can make them a lot cheaper by making them less reliable, yes. (Not safety issues, just mechanical ones.)

Whatever companies make the self-driving car software (which may or may not be the same as the companies that make the rest of the car), will have more ongoing contact with the customer than traditional car companies. I think Tesla has already moved well in the direction of that model, and is well positioned to move even more toward the model of a software company as the need arises. Then again, Tesla might just become a robotaxi company. I believe Elon Musk has already claimed they will, though it won't be 100% his choice, and maybe he was just blowing smoke. Other car companies might have more problems moving to that model. I wouldn't be surprised if some of them opt to focus on creating the cars and let someone else create the software. Cross-car ADAS software already exists (see, e.g., openpilot), and I think that is a crude preview of one possible future we could see. Analogizing to PCs, Tesla is Apple, Openpilot is Linux, Waymo is Google. Nvidia maybe is going with the Microsoft model, though they're doing software and hardware, which is probably the way to go in the near term, and the near term might last quite a long time.

I don't think the need for hardware maintenance and service, beyond things that can be done by the owner (like cleaning the cameras), will be much greater, if at all greater, than for regular cars. And even if it is, relying on the manufacturer to perform those services is only one possibility. I provided it as a simple way to show that anything that can be done by a robotaxi company could be done by a car manufacturer. But I don't think it necessarily will be done that way. For the simple things that can be done by the owner, like manually cleaning the cameras, and like cleaning the interior, the robotaxi model is going to be a significant added expense. Not to mention what a nightmare it's going to be when a pandemic hits.

"a taxi service that lets you effectively have guaranteed access to a vehicle, the ability to leave your stuff in it all day when you want etc" is just a rental car that delivers itself. I think that is a model that will succeed. I also would say that it's not mutually exclusive with traditional car sales. It's possible that the car manufacturer will also run the rental car service. It's also possible it won't. It's quite possible that antitrust law enforcers wouldn't even allow such a thing.

I don't see how you can "make them a lot cheaper by making them less reliable." At least not in ways that wouldn't have an impact on passengers. (If they break down in the middle of a ride too frequently, I'd say that's unacceptable.)

Society might spend $8 trillion on transportation, but the total profits are very small, if there are any. So if there's competition in the market, where is the excess "value" to be captured? Uber & Lyft keep recording losses, even though they've been in this business for years. Then everybody decided to do food delivery, even though there's no profit margin there either. So those companies are shopping around for merger partners or someone to bail them out.

Journalists keep talking about the size of a market in terms of gross money spent, but pay no attention to whether there are any actual profits, either currently or likely in the future.

My marginal cost to drive my car is $0.10 per mile. Adding insurance, depreciation, and repairs brings that up to 20 cents per mile. Why would I want to pay 5-10 times more to get a ride in an AV?

Tons of companies are making tons of profit. Mature industries have smaller margins but they are doing just fine, thank you. Startup industries lose money as investors are willing to do that to get early dominance, as happens at Lyft and Uber.

What car has a marginal cost of 10 cents/mile? That's a fairly rare car. But as you know, you had to buy that car. Robotaxi services want to offer a car ownership alternative which is cheaper in total cost than owning a car. And they can, though not on day one.

If you've got a 7+ year old car that gets 35+ miles per gallon, your marginal cost per mile is going to be pretty close to 10 cents.

That's not super-common, but 2013 was the peak sales year for hybrids, with over half a million sold, so it's not rare, either.

A few more cents a mile and you can add a 7+ year old version of the best selling vehicle of all time (the Toyota Corolla) to the list.

You say that "Tons of companies are making tons of profit." But if we're talking about just transportation/mobility, who are these companies that are making tons of profit?

I am talking about the car companies, the oil companies, the insurance companies, the car dealers, the car service centers, the car rental companies, the gas stations, the car washes, the auto lenders, the parking lots and more. All the people you don't use any more if you switch to robotaxi use.

Oil companies & auto companies (other than TSLA) are some of the worst performers in the stock market and have profit margins of less than 5%. I don't know about the privately held businesses.

Those companies are performing badly in the markets not because transportation is a declining industry, but because of investors fearing those companies will lose out to waves of the future -- including robotaxi service, electric cars, and Chinese carmakers.

Roboxtaxis build themselves, don’t use fuel, don’t need to be insured, never need repairs, never need to be cleaned, don’t cost any money, and never need to be parked?

2002 Acura RSX Type-S, 6 speed manual, averages around 30 MPG combined, could get 40+ at cruising speeds. Gas here sells for $2.399 per gallon so that equates to 8 cents per mile (@ 30 mpg).

The cost of gasoline is just a fraction of the incremental cost of operating a car. The real number is debated, but you need to add depreciation, maintenance, parking and tires at least. Some debate how much of insurance you should add -- many insurance companies now offer variable pricing based on your miles, some even have almost purely mile based. While you can buy flat rate miles-independence insurance, you should not. Parking is not included in most quoted numbers because it varies greatly. Some pay to park with every trip, others get subsidized (free) parking.

As I originally wrote: "My marginal cost to drive my car is $0.10 per mile. Adding insurance, depreciation, and repairs brings that up to 20 cents per mile." I even rounded those numbers up, my real costs are probably less. If I choose to drive my car 1 mile, my only direct marginal cost for that 1 mile is 8 cents for gas.

5 years ago I paid less than $4,000 for my car and I could sell it for almost that much today, so my depreciation is very low. I don't pay much for insurance as I forego collision coverage. 20 cents per mile more than factored in all my costs.

Yes, if you are willing to ride in a $4,000 car, it is cheaper to ride. Though there is a floor, because with cars of that age, repairs and other factors start overwhelming the savings on depreciation. Eventually it starts going up again, and you junk it.

So you want to compare the cost of driving a late model car. For a brand new car, the cost is ridiculously high as they depreciate fast in that first year. Most companies calculate 5 year total cost of ownership -- you will find those from many sources around the web.

What you won't find so much is a good breakdown of the total cost into the annual cost and the per mile cost. There are costs you pay just to have the car sit in the garage -- time based depreciation, time based maintenance, non-mileage based insurance, annual licence fees, interest on the capital (whether you took a car loan or not) and to some extent, the cost of having a home parking space.

Then there are the costs that come when you drive it.

However, there are arguments that this is a silly division, because you don't pay all those annual costs on a late model car to leave it in the garage. And you don't pay them at all if you switch to a mobility service (of the future.)

Add new comment