Waymo Can Charge For Robotaxi Rides In SF (So It Gets Interesting)

Waymo revealed they have been given permission to charge money for rides in San Francisco. They have recently been giving them for free to a select group of testers. In Chandler Arizona, they have been charging for rides for several years, so what’s the big deal?

Waymo is widely regarded as the leading self-driving company by a large margin. They have been at it the longest, with one of the best regarded teams and huge resources, and produced the best results. This includes operating in Chandler with no safety driver in the taxi. That experiment in Chandler has helped them understand how riders react to a robotaxi, especially one with no other person in it. As Sun Tzu never said, “no business plan survives first contact with the customer.”

While they charged money in Chandler, at a rate slightly below the cost of an Uber

UBER
, the problem is that the suburban town is not really a reasonable market for robotaxi services. Uber and taxis are rarely used there, transit and walking options are poor, and private car ownership is the norm. Some day robotaxi services will operate in towns created for cars, but not at first.

San Francisco is another story. It is a much older city with lots of transit and walking, and it’s where Sidecar was born, and quickly copied by UberX and Lyft

LYFT
. Short of a seriously car-free city like Manhattan, London or Hong Kong, it’s a great place to experiment with the economics of robotaxi service to find out what works.

Waymo can try a bunch of pricing models:

1. Price it similar to Uber. This just tests how people compare it to such services, and also how many use it because it’s cool, and that takes a long time to wear off.
2. Price at a price much better than Uber, but more than private car use, like $1/mile.
3. Price at a price similar to the all-in cost of private car use, around 50 cents/mile.
4. Price at a price similar to the incremental cost of private car use, which is also the future all-in cost of a minicar taxi — around 25 cents/mile.
5. Price like transit, flat fee inside the city.
6. Various subscription fees, including flat monthly fee, or flat monthly fee plus much lower per ride or per mile fee.

In order for a robotaxi to succeed is must do much more than just compete with services like Uber or transit. It must work as a car replacement. People need to feel they can give up car ownership, either for real, or temporarily as part of an experiment. That’s actually hard for a pure robotaxi service to do — you need to supplement it with super quick carshare for trips out of town and other special travel, and you need to be able to take the rider most places they need to go, with a very short wait time. You gain a big benefit in large cities from eliminating the costs and delays of parking, but that only goes so far.

Waymo faces a burden because they are only operating outside the downtown. Nobody is going to see it as car replacement until that’s fixed. They could just offer you a human driven taxi when you want to go downtown, as long as they can stop doing that fairly soon. Cruise also has ambitions here, but their service area is even more limited and they only operate at night. (Cruise is running with no safety drivers in that limited area, which Waymo does not yet do in theirs.)

Indeed, one can argue that these economic experiments could be done by just subsidizing the price of an Uber like service. You don’t get the full-product experience which combines being alone in the vehicle and paying for it, but you get a lot. Several companies have operated robotaxi services with safety drivers in certain areas. Motional just announced they will be operating one in Las Vegas, though it will be free and only in the “downtown” area, not the strip or convention center.

Go big

To truly test a commercial service you need a big fleet, and sadly a limited set of customers. One of the main measures of quality of a service is wait time, particularly when you compare it with the near zero wait time of a car in your garage.

The wait time is determined from the number of idle vehicles you have per square mile of service area divided by the number of active customers in that region. The higher density of idle vehicles, the closer one will be to you, and the shorter wait. You can reduce that wait time by having more vehicles or having fewer customers. This is how Lyft, even though smaller than Uber, can compete with it on wait times — they are smaller both in customers and in drivers.

Sadly, the best way to test this is to limit the number of people who can use the service. There’s tons of people who want to try it, so perhaps they could let random people have 2 rides in off-peak times to make sure the vehicles are there for the real test subjects to see what it’s like to have cheap, on-demand robotaxi with fast service.

In a way, it would not be a bad idea for Cruise and Waymo to even combine once Cruise expands to daytime service and more area. I don’t think the learnings will be all that secret or proprietary. This probably won’t happen.

More trips

Waymo did all their early development in Silicon Valley, where their HQ is, and until recently their vehicles were a regular fixture on the streets there. They intended a service there but probably decided they did not need to learn more about suburbs at this point.

There is one service I think would be worth experimenting with, however. It would be to have two services, one in San Francisco, and one in the valley, and regular connector vans and buses (possibly driven by humans) between them. As I have described in my articles on the potential future of transit robotaxis offer the potential for door to door transit where the common part of the ride is shared, giving the best of both worlds. You get the door-to-door on-demand service of cars, with the reduced road use and lowered cost of shared rides for most of the miles.

In such a service you could call a Waymo in Silicon Valley and it would take you to a place where a van or bus is leaving to go north within one minute. (You don’t have to send a van every minute, though they could, you introduce modest delays in the pickup that nobody notices so you only have to send a van or bus every 5 minutes.) The vans go to various switch points in San Francisco where other Waymo robotaxis are already waiting, and you get in yours. (Post covid, two people going to similar destinations would share.) Unlike UberPool services, you do almost no detouring out of your way — none if you don’t share the last mile ride.

The same thing repeats on the way back. Because of the shared van service, the cost of this ride can become less than the cost of driving up yourself and parking. In addition, they would let you do rides among places in San Francisco while up there, to simulate what you could do if you drove your own car up. Today, almost everybody going to SF from the valley drives their own car up — the train exists but has much longer travel times and the stations are not conveniently located. (At certain times semi-express trains run, and the robotaxis could go to the train station and pick up there for those specific times, where the train might beat rush-hour traffic, though barely and only for certain regions of the area.)

This permits a robotaxi service even over a large service area, and would actually compete with car ownership even with existing car owners. Some of the vans could also go to SFO airport to add more utility, but this may not be what Waymo seeks to test now. Of course, there is still the issue of the virus when in the van.

The service would also be very popular with Alphabet employees. Alphabet/Google runs a very large private transit system with buses that run from Google HQ to places all over the area, but in particular to San Francisco.

Artificial prices

It’s not out of the question that other companies, including Uber itself, might object to a Waymo experiment which has enough size to deliver meaningful results but which deliberately undercuts their prices by subsidizing an early service that is not yet that cheap. One would hope this would not create an anti-trust violation if it were just for research.

Right now, robotaxi operation is pretty expensive. Not even counting the immense R&D expenses, the vehicles are being made in small volumes and nothing has yet been made to scale. Cruise has just now applied to NHTSA to get the exceptions and approval they need to manufacture a vehicle without a steering wheel. Real operation will require not just vehicles, but operations centers and cleaning depots and charging infrastructure. There’s a lot that is going to cost a lot in the early days, so this is not a game for the underfunded. Many of the players are the biggest companies in the world, so that won’t be a problem for them.

When it comes time to set real prices, I expect we’ll see something around $1/mile. That’s half of what an Uber cost pre-pandemic (it’s gone up a bunch of late) but more than the typical per-mile cost of car ownership. Car owners don’t tend of to think of the cost as per-mile, however, while robotaxi operators must. It’s high enough to make people who already own a car not find it attractive, unfortunately, though low enough to be a bargain if parking is expensive at the destination. The reality is that once players are in it for revenue, they won’t want to charge less than the market will bear. Only once there is competition will prices really drop. Outside of places like San Francisco, competition is less likely. It will be easier to just go target a virgin city with nobody else playing than to go head-to-head in a city where somebody else got in first.

So now it gets interesting.