Hence why many investors don't mind that Tesla isn't generating wonderful cash flow. If there's a reasonable expectation that this data collection will lead to full autonomy, then Tesla should price their cars at whatever level is necessary to move their maximum production into consumers' hands, even if it were a loss (which it's not).
And even if they don't use the data to train, they can use it as an argument for how much better they are than humans. That's one hurdle that Waymo could have trouble with, whereas Tesla could point to tens of millions of miles where their sensors would have avoided crashes that humans failed to avoid.
