GM’s “Cruise” robotaxi unit announced the price they will now charge for robotaxi rides, now that they have received permits to operate commercial service all day in San Francisco. There will be a $5 base fee (”flag drop”) plus 90 cents/mile and 40 cents/minute, plus tax.
The real cost is the 40 cents/minute. Taxis in a dense city like San Francisco during daytime traffic only average about 12 miles/hour, so the hourly fee will add about $2/mile to the price bringing it over $3/mile. You’ll do better late at night, which used to be the only time Cruise offered rides to the public. Cruise isn’t allowed on faster roads like freeways at present.
This pricing is pretty similar to Uber, Lyft and taxis. Waymo has announced similar policies to Cruise, but not yet named the specific numbers, but they are expected to be similar.
And that’s a disappointment. Waymo and Cruise have been eager for permission to charge fares for rides. It’s not because they need the money — both plan to lose money on their pilot services for some time to come, lots of money. Charging fares allows them to experiment with business models to see what will work in the future. It also helps control demand. Both have long waiting lists. They want to track demand when it’s not free and when the novelty of a “future taxi” has worn off.
At the same time, the long term goal for the robotaxi is not just to be a clone of Uber, or even a modestly cheaper Uber. The demand for Uber and taxi services at these prices is quite well understood, and needs no grand experiment. There is some value in understanding the effects of removing a driver, and the eventual effects of having a specialty vehicle like the Cruise Origin or Waymo Zeekr, both of which are yet to go into service. Some value, but not the real game.
Waymo has reported that may riders enjoy the privacy of having nobody up front. (According to some reports, a few riders are enjoying that privacy a little too much and are getting more intimate than one might expect in a windowed box where the interior is being recorded on cameras. For many years, I’ve made jokes about this, suggesting that the “Vision Zero” initiative which hopes or zero deaths in moving cars, might have to be come Vision Negative as people are created in moving cars. But it turns out there are solutions to this if it becomes a problem.)
More reasonably people are enjoying having no pressure to converse with a driver, and the freedom to have private phone conversations and phone calls. There’s also no need to tip. This is good, but today it comes with vehicles that are still fairly poor at pick-up and drop-off compared with humans, and which can’t help disabled people get in the car. The presence of a driver can also discourage people making messes or even domestic disputes that get physical, according to some commenters at the recent PUC hearings.
Cruise and Waymo say the charges will be based on an “optimal” trip, ie. the sort that Waze might calculate for a human driver (other than freeway I expect.) The vehicles today often take less direct routes, to avoid streets and turns they don’t like making. While it’s good that riders won’t be charged extra because the car still doesn’t handle all roads, the trips will still take longer — sometimes a fair bit longer. Some riders may balk at the delays now that the rides cost as much as a faster ride with a TNC.
All this is good, but the real change happens when robotaxis offer what was never possible in both private cars and taxis, including prices no taxi can beat, entirely different styles of pricing and previously impractical modes of service.
Cruise and Waymo have announced they might raise prices at peak periods to “control demand.” That’s interesting because while this is a factor, the core reason Lyft and Uber say they do this is to encourage the human drivers to start driving and to head with the areas of high demand. 75% of the Uber fee goes to the driver. With robotaxis, there is no “convincing” the driver, who is, of course, a robot. If companies increase prices, it will be either to make extra profit, or simply to manage demand when demand gets too high. If demand exceeds supply, wait times can skyrocket and rides can be simply unavailable, and that can mean a surge makes sense, though it also means rides are allocated to those who can pay more.
A well provisioned robotaxi fleet, however, should have enough vehicles for some decent fraction of its peak. If demand goes higher, it might raise prices, but it might also try to recruit vehicles from subcontractors, or even, in the more distant future, service from privately owned cars that are robotaxi capable. (At present, there is no indication such a vehicle will be on the market any time soon, in spite of Tesla’s regular bad predictions to that effect.)
A better solution might be to implement ride pooling during surges. While Uber and Lyft have never had great success with their ride pool offerings, periods of high demand are exactly when ride pooling is easiest to do, as there are more people who can be pooled together without major inconvenience. Many people report that rides in Uber Pool added lots of delay for inadequate savings. This also reduces congestion on the road.
Robots don’t have the desires of humans. They will, of course, never decline a ride order from their fleet operator. They also don’t “mind” waiting. Today, Uber drivers must sit around waiting for the next fare, and they are not paid for this time. In theory the flag-drop fee compensates for this waiting, as well as the driving to the pick-up point. While fleet managers wish to control costs and not have vehicles sit idle, there are no emotions or boredom to factor into the equation.
This also enables robotaxis to give affordable shorter rides, such as are desired for “first/last-mile” service both to transit and to robotic vans and pooling. Ridepooling has not been a big success because trips are made longer doing pick-up/drop-off for others. If instead the main vehicle can stick to the common routes and the short side-trips can be handled by a smaller robotaxi, a service can deliver pooling with almost no inconvenience. That’s not possible with human driven vehicles, because these short rides involve a lot of work by a driver for little distance and little money. Robots again, don’t care.
Cruise’s recent problems with the crowd leaving a music festival in San Francisco (Uber and Lyft also overloaded) would have been a great place to do smart pooling. Every car leaving the concert should have been full with people going to roughly the same place, with single person robotaxis (that don’t yet exist) taking people the last mile to their homes from the common point. Indeed, for a concert, vans and buses should start the pooled trip. Those not willing to pool would probably have to wait a bit.
Even a taxi-like service, billed by the mile at a low cost that compares to the estimated 50 cents/mile (plus parking) cost of car ownership is not enough. People don’t feel they pay for their private cars by the mile. Instead they buy or lease the car, fill the tank 1-2 times/week, pay for insurance and maintenance twice/year and make repairs at random. As a result they view actual miles as costing little more than the 15-25 cent/mile cost of gasoline, or even the 4 cent/mile cost of electricity. Indeed, once a person has bought a car, to not use it is to waste it, the cost of the actual drive is seen as insignificant.
Cruise reports their costs have been dropping very rapidly, and in their quarterly report, indicated they hope to soon see costs hit $1/mile. That’s presumably a “cost of goods sold,” which is only recurring expenses that go up with every mile driven or day operating, not non-recurring expenses like their massive software development and testing costs. A business like a robotaxi needs a gross margin of at least 50%, so a $1/mile COGS suggests a $2/mile retail cost. Cruise and the others will need to get the COGS much lower in time. That’s possible, particularly if they move to smaller vehicles for solo trips. Once your business is stable, the depreciation of the vehicle itself becomes the largest component of the COGS, and small vehicles can have small cost. The expensive hardware, like all electronics hardware, becomes cheap in volume.
To compete with car ownership, alternate pricing models are needed which can be easily compared to car ownership, and shown to be equal or superior. That actually should not be that hard — estimates suggest the average car owner pays about $8,000/year plus parking for all costs. The problem is mainly that they don’t perceive this. Passengers don’t pay for parking when they ride taxis, though the small cost of places for cars to wait or be stored is factored into the cost, but shared over many more miles.
The true commercial value of a robotaxi business comes when you can convince customers to replace car ownership with your service, or rather a suite of offerings centered around your service. Just being a cheaper Uber is a start, but the taxi ride business is a tiny fraction of the private car business. Of course, you want to replace their car with something better and/or cheaper. It will have a few downsides, like short waits, and no easy ability to keep stuff in the car that takes you around (at least at the base fare) and it will be more expensive for certain activities, but otherwise it should offer a great suite of advantages. Naturally, at first, companies will try to sell this at a premium, but in time it will need to meet people’s needs for less — and it can.
At first, demand can be artificially high. People are rushing to try out robotaxi services just for the novelty. They are already a tourist attraction in a couple of the places they exist, and this will be true for some time. In order to experiment with how people will use it in real life, you have to offer them the real prices you hope to have once in full production, and you must also give similar wait times and service levels. Until your fleet is very large, the only way to do that will be to limit who can ride so that vehicles stay free and wait times are low. There’s still a lot the companies have to learn about how customers will react to the services and prices, but now they will be permitted to do experiments.
They need more, though. People travel outside San Francisco all the time. A customer will need an easy and effective way to do that, and right now the companies don’t have any offering. Waymo has permission to operate in Silicon Valley, and Cruise may seek that before too long. A full service will need to handle most of the Bay Area to fully replace car ownership for a customer, and then offer an easy rental-car for rides outside the service area.