If you're spending more than you're making, then you're almost certainly projecting LTV.

The big, falsifiable assumption here is that your acquisition sources will keep sending you users of the same quality. Because of the way that ML systems work, this tends to not be true, and if you are using long windows it will both take you a long time to realise this, and cost you a bunch of money.

This is normally how companies go bust/stop growing as a result of LTV models.