I have often thought of this as well. One option is that they still outsource the financing to local owners, not necessarily your traditional car owners, per say, but more like local companies that buy AVs and place them in service.

However, I think an interesting approach would be to use a traditional Project Finance scheme [0] to obtain cheap, non-recourse loans in order to acquire and deploy large fleets. These schemes are common for infrastructure projects, which occasionally amount to billions of dollars. Under such a scenario, then the actual owners might end up being entities familiar with setting up these deals, and Waymo et al selling the technology/cars/platform (arguably of higher margins).

[0]http://www.investopedia.com/terms/p/projectfinance.asp

> The end game is that Uber, Lyft and taxi drivers are out of a job as chauffeur.

This is true, but as someone who grew up somewhere where there's snow on the ground 6 months out of the year, I'm wondering how we're going to deal with more extreme driving conditions. How reliable are these self-driving cars in ice and snow? Is there even a strong economic reason to solve this problem right now?

My concern is that we'll end up with a fleet of cars that work great 90% of the time, and then flat refuse to drive when they fall outside their design envelope. At that point driving will be a niche market: the pool of drivers for emergency vehicles (which can't just take the day off when it's snowing), and services in rural areas (where detailed maps and standard heuristics may fail) will be the only permanent workforce. Meanwhile anyone else with a licence will have to step in as a chauffeur when conditions get bad.

Considering most of the U.S. (primary market for self-driving cars) experiences snow and ice several months of the year, why wouldn't engineers spend time to make this work well? Is there something inherent about adverse weather that makes humans better at it?

The evidence is that anti-lock brakes and variable traction controls have lead to a reduction in crashes in adverse weather conditions. Driving in snow and ice is hard. Why wouldn't a system that can make much finer and frequent adjustments outperform people? The only thing that is really holding back self-driving systems is the computer vision aspect, which is mitigated with Lidar and by further research into computer vision improvements.

If they'd really want to do that, Alphabet has > $90B in cash.

The spend/build-out is even bigger than that because if these cars have no driver how to they refuel? Unless they've got a robot arm capable of doing that themselves they still need a human involved. As well as somewhere to do repairs and maintenance on the cars. Right now Lyft and Uber effectively out-source that to the owner of the vehicle. So whoever is doing this roll out doesn't just need the vehicles they also need the facilities to house, maintain, and fuel them. That infrastructure is, arguably, even bigger and more complicated than buying up the cars.

If the math works, the capital will flow. Even if the math doesn't work, the capital might still flow.