Uber is [edit: almost] certainly not operating at negative gross margin.
> People tend to underestimate the true ownership cost of their cars, and Uber uses that to make them believe they're making a lot more per hour than they actually are.
That isn't an argument about its efficiency though.
Maybe the marginal value of a car self-driving itself somewhere is $1 and the marginal cost of that driving is $0.99—it's still more efficient for that car to drive than not. Whether the owner thinks their profit is $1 or 1 cent doesn't really matter.
The number of full-fare (no subsidy to either rider or driver) Uber trips surely absorbs these incentives, leaving a positive gross margin. (There's also a miniscule computing, bandwidth, and payment processing cost to a marginal ride, marginal being defined as the nominal X+1th ride after all the costs for X rides are already paid for.)
Perhaps absolute certainty is a slight overstatement, but here's one where I'd bet eating my hat on it. I edited to insert an "almost".
Example: a few weeks ago, I paid ~3USD for a 13km ride that took 90 minutes (due to Beijing traffic). The driver was definitely paid more than that for the ride.