Comment by Zigurd
They haven't said it explicitly. But the reason that Waymo can add five cities this year is very likely they are at least at break even on opex. They likely reached that point sometime last year and it seems to have held up.
So I wouldn't call robotaxi service unproven. But I would call the idea that you can claim to be running a robo taxi service without depots, cleaners, CSRs, and remote monitoring that can handle difficult situations in a more sophisticated way than each car having a human monitor it, naïve.
I read that as meaning even the scaled robotaxi service (Waymo) does not throw off enough cash to offset the loss of Tesla's vehicle sales unit. (The putative Tesla buyer they are dissuading from purchase would have to take a whole lot of robotaxi trips to generate the same amount of profit for Tesla. Assuming Tesla can get robotaxis working.)
In the 2000s publishing pivot to the Internet, this was known as "trading physical dollars for digital pennies."