Then they just sell their car if they don't work for Uber anymore. It's not like cars will suddenly become worthless.

They might drop in value a lot when driverless technology goes mainstream. If the supply of sedans is the same but the demand is way less because driverless cars don't need to be wasted sitting parked 90%+ of their existence, that could make used cars way cheaper for a while.

Uber drivers are likely to accrue higher than normal mileage on those vehicles, which would lead to faster depreciation.

Loan payoff is usually gradual, so I'd imagine most of them from day 1 are underwater as far as what they owe and what the car would sell for on an open market.

Market similarly discounts used taxicabs, used police vehicles and anything that's re-sold by car rental places - such vehicles have higher mileage and are more likely to have been abused.

Hmm, that's an interesting thought. How to evaluate it?

It depends on (1) how long from now you estimate cab driving may be obsolete, (2) how long the useful life of your car for cab purposes is, and (3) how driverless technology will change what the resale value of the car at its end of for-hire life otherwise would have been.

Uber requires fairly newish cars (not sure how reliable this link is, but according to http://www.idrivewithuber.com/uber-car-requirements/ it varies by type and area quite a bit from around 5 to 15 years) so if the car's useful life will only be 5 years (such as in NYC) and you estimate there's still around that much time or more left before drivers can be replaced, then it doesn't really matter. On the other hand, if your car's useful life is more like 15 years, you could be stuck with a useless asset. Maybe a driver/owner can also titrate the risk here by buying an already older car?

The other question is how widespread driverless technology would affect the resale value of these cars. Can a driver/owner just sell their car when they lose their job? On the one hand, for privately owned passenger cars, it's hard to imagine the resale value not being reduced, when car utilization rates can increase dramatically (right now a privately owned passenger car spends almost all its time parked, doing nothing useful for anyone).

On the other hand, for-hire vehicles (well, medallion cabs in NYC anyway, not sure how this compares elsewhere) already have utilization rates close to 100% (they're constantly being rented out to different drivers and are only off the streets when they're being maintained), so IF driverless technology is cheap to retrofit to these older cars, they should hold their value. ...And on the other other hand, right now once a for-hire vehicle can't be used as a cab, at least it can be resold as a private passenger vehicle. If private passenger vehicles are much cheaper because the demand is much lower due to much higher utilization, that stops being a good option. In that case, all you can really do is get as cheap a car as you can now, because it's going to have very little resale value after driverless technology hits.

Finally, a driver can rent a car. There are businesses specifically to do this. The risk is then shifted to the rental business (though presumably they charge appropriately for that). Yellow medallion cabs in NYC are almost all done this way, not sure what percentage of ubers are rented versus owned-by-the-driver.

Anyway, maybe the bigger risk is not the investment in the car, but the investment of time. If you are spending your whole workday driving for several years, you are getting very good at driving, and not getting good at any skill that will have market value after driving is automated. And even if there are any marketable skills that are related to driving (eg, let's pretend that driverless technology can't do CDL driving at first, or that a good cabbie can pick up airplane piloting quicker than average -- just silly examples), all the other displaced cab drivers will be chasing the same options. I think the more expensive asset at risk is the driver's skill and time than the car.