Engineers are significantly more expensive than chauffeurs, especially since the chauffeurs get paid roughly minimum wage - or less. I too am very curious where pricing settles out.

Only if there is one engineer per car

Well no, it depends on the multiple. Minimum wage is $15000, and an engineer costs Google an average of at least 10X that in base salary, plus another 10X that in equity. I think the break-even is probably closer to 20 cars per engineer.

Currently they have ~250 cars in SF, and 2,500 employees of which about 600 are engineers.

To break even I'd say they probably need at least 12,000 cars, assuming they don't hire any more engineers and the other 1,900 employees are costless. This also assumes that the cost of the Waymo is exactly the same as a regular car, which at least today, isn't even close. And that the server infrastructure is free, and that insurance is comparable. It's probably also too early to know what the maintenance cost of these systems is compared to a vanilla vehicle.

Given how many taxis a typical city has, there's no way that the cost of engineers would come close to the cost of chauffeurs.

In fact, this is the way economics work for many forms of automation: you're replacing a large amount of labor time with a much smaller amount of engineering time. The engineering time is more expensive of course, but the much smaller amount more than makes up for the increased cost.

Eventually when Waymo is deployed globally the expensive upfront engineering will be absolutely dwarfed by the cost savings of not having drivers.

Litigating accidents will cost more because suddenly it's not "partner" at fault but your own people. Also, someone's gotta buy that new yacht somehow...