Essentially he makes the point that, if you have to own a fleet and maintain it, even if you stop paying driver's wages, you're not making profit compared to market rates for Uber drivers. Sad implication is that uber drivers themselves are underwater after taking into account depreciation and maintenance.
A fleet where Tesla operates a network without needing to own or maintain the capital equipment is a lot more profitable, and they're one of the few companies positioned to pull it off (certainly not Waymo). Essentially, Tesla could make money on the vehicle sale, make money on the maintenance, and make a platform fee on the taxi network. Wonderful economics on paper.
Win-win too: the car-owners supply the capital and in return receive the freedom of their own car plus an offset to their depreciation via the rental cash flow. That would mean that teslas would have incredible resale value compared to 'dumb' cars that don't produce cashflow.
That said, trust in the network is key, and Tesla + Musk aren't exactly winning public trust recently. If this is going to work, each incremental rider needs to be able to trust that
- if they try to book a Tesla robotaxi they will be able to find one available, nearby and in the time they want it
- the car will work (well maintained, enough charge for their trip duration)
- the self-driving will be safe
While all those things are straightforward, to be honest, it's hard for me to imagine how they can provide a better value proposition than human Uber/Lyft in that regard, where in major cities a car is always 10 min away or less, and the driver has both a financial and survival interest in ensuring it's well maintained and safe (the driver's life also depends on it).
Maybe the niche will be suburban or exurban places where human ride shares don't have enough demand to efficiently match with supply, and where people are okay waiting / planning in advance if they're sure a car will come. Could be some overlap with starlink market fit.