I know that accounting-wise this "makes sense," but marginal costs after rolling out the physical network are pretty close to zero, right? How is shedding 2.3 million TV subscribers a profitable move?
But, it's really hard to figure any of this out when talking about ATT. They have 4 (at least) different ways to sell TV service (DirecTV satellite, AT&T TV, AT&T TV Now, and AT&T Watch TV). If those names aren't confusing enough, then I'm sure they still sell U-Verse service somewhere too.
Each of these services was/is losing customers, but at different rates. And the cost for each service varies a lot too, as does the programming available. I actually subscribe to AT&T TV Now, but I'm on a grandfathered programming package that isn't available any longer. Who knows how many other grandfathered packages exist between the varied services.
The price hike reported here is for the AT&T TV Now service, which was their older IP TV offering. It's been a pretty open secret for a while that they have been trying to get as many people off of that service and moved over to the (more expensive) AT&T TV service as possible.
And to go back to your original point, each additional customer does create variable costs. At&t has to pay carrier fees for every subscriber. A customer pays $6/month for ESPN, $0.60 for NatGeo, etc. Maybe the old price tier wasn’t covering the carriage fees + equipment + labor to install
However, I do suspect that as the number of cable TV cord cutters continue to rise, you're left with more and more customers who are less price sensitive and more set in their ways. So it's really not a bad strategy to just give up trying to compete on price and milk an existing customer base for as long as you can.
You're going to lose the people who are ready to move to online streaming and just haven't gotten around to it soon enough anyway.