There have been some blogs and news items recently, including from our good friend Glenn Mercer, which relate to some recent positive developments in the world of Waymo. They just achieved 200,000 rides in one week, double the level just six months ago and up from 10,000 two years ago. At least within the defined areas where they are operating (good weather, planned cities, etc.), they appear to be operating reliably and are winning customer trust. There are fewer reported incidents of the Waymo cars being involved in accidents, than there was with the now-cancelled GM Cruise programme or with Tesla’s ‘Full Self-Driving’ mode. Customers who have the choice between Uber and Waymo apparently tend to prefer Waymo even though the wait time may be longer, as customers are in control of their environment and don’t share the car with the stranger required to drive the car.
Does this then mean that robotaxis have now finally broken through, and we will soon see them displacing Ubers and conventional taxis, and then in turn the privately owned car? The clear answer is no for two fundamental reasons, one that might be addressed, and the other more fundamental. (There are a host of other obstacles from insurance through to regulation, but I’ll ignore those for today to keep life simple).
The addressable issue relates to the technology itself. One of the reasons why you see the trials of robotaxis confined to the sunnier US states and southern China rather than New York or (until very recently, and then only on a very controlled basis) Beijing is that the technology is still weather-dependent. If you put a robotaxi into falling snow or heavy rain, it is blinded and will either hesitate or pull over until conditions improve. If we then factor in the layout of city streets, there is a big difference between the wide open spaces and relative order and discipline that you would find in Phoenix compared to the chaos of Manhattan or London. I do believe that with improved sensor technology, better real-time analytics including leveraging AI, and the potential of additional inputs through car-to-car and car-to-infrastructure (V2X) connectivity, the capabilities will improve. Every time your flight lands using Autoland capability, you are relying on V2X of a fairly primitive form, but it gets you on the ground safely.
The second issue however is more fundamental in my view and will only be partially addressed by technological advances. As is well-recognised, Uber relies on an army of drivers who they have been able so far to treat as self-employed, providing their time and their car entirely at risk with no guaranteed financial return. Speaking to dealers and other renters who specialise in providing cars to Uber drivers, the rents charged are two or three times those charged to a normal customer because of the default rates, and there is a surge of driver activity in the day or two leading up to when payments are due as the drivers need to secure their economic livelihood into the following week. From the Uber perspective, they incur no fixed costs to provide the means of transport and they have entirely variable capacity. If there is no demand, the drivers go home to bed (as I and my fellow travellers from MHA discovered in a snowy Alexandria Louisiana on our way to the NADA Show at the end of January). If there is peak demand, and ‘surge’ rates have kicked in, drivers are happy to squeeze in an extra hour or come back out from their homes. With drivers only getting part of that surge rate, the extra capacity comes at only marginally higher rates to Uber than the base.
For Waymo and other similar robotaxi companies, additional capacity means additional vehicles which today include a technology stack that costs around US$200,000. Baidu in China is reporting a much lower cost of under US$30,000 including the base vehicle, but it is unclear whether this is a like-for-like comparison in terms of capability. However, even if we took the lower end of that cost spectrum as a long term possibility, additional capacity will come at that cost plus insurance, licences and other fixed costs. The inefficient use of fleet capacity will drive up costs, restricting capacity will lead to customer frustration. Unless customers can be persuaded to smooth out their demand with some working, shopping and studying in the early hours of the morning as much as during daylight hours, this problem is insurmountable. This is the ‘lie’ that sits behind the oft-quoted claims by mobility-evangelists that personal car ownership is wasteful because the average car is only used 5% of the time.
Elon Musk (obviously) has an answer to this. His vision is that robotaxis such as that shown in October last year will be so cheap that individuals will buy the cars, and then allow them to be used to provide a driverless robotaxi service when they don’t need the car. We’ll ignore for now the (to him) minor legal obstacles like the fact that Tesla has done under 600 miles of testing with unsupervised self-driving in California since 2016, where he plans to launch ‘later this year’. The idea fails more fundamentally on the willingness of owners to share their cars. Unless the car owners are happy for their cars to be off on commercial work at the same peak hours when they want it, and to clean out the mess left by the last passenger making his or her way home in the early hours, before doing the school run, it seems to me that take-up will be low.
We are therefore left with a situation where capabilities will improve, technology costs will come down, but the combination of demand patterns and fixed to variable cost economics will mean that whilst robotaxis will become more widespread, most vehicles on the road will still have a steering wheel, with a driver behind the wheel, and in most environments their eyes on the road, rather than a movie.