The labels are in. Apparently, it's free trial as presented (pre-reversal) in terms of label compensation. Basically, the labels are rich enough in this deal to weather whatever hits they could take to their current cash flows from iTunes during the free trial. OR, as others have put it, they can afford to take a longer view of the upside beyond 3 months and just roll with sharing in the risk by offering their stable of content rent-free for 3 months.
In this reversal, it's the slivers of what the labels are paid in that 72% cut that goes to the artists and THAT will apparently be covered by Apple during the trial (or probably even less than that since some clarification comments makes it sound like free trial comp will be based on what is actually played during the trial). If so, this really doesn't cost Apple much at all, as the labels deals with their artists in general are thoroughly exploitative relative to what the label keeps for the value it delivers, so those of us worrying about what this will cost Apple is really worrying about very little hard cash... easily recouped with even moderate success after 90 days, if a fair amount of those trials start paying.
In another thread one guy thought this might cost Apple about $18 million. Do the math to see how much Apple could make each month AFTER 90 days: my own guess is about 25 million paying subscribers (which is apparently only 5 million more than Spotify has now) at $2.80 per month (which is Apple's cut right off the top) = $70 MILLION PER MONTH. Now that's not all profit but I believe there is plenty of profit in there (in spite of the much slung infrastructure and R&D argument, else if the latter is so expensive that there's hardly any profit in this, why is fat-profit-margin-hungry Apple messing around with this business?). And of course, if one assumes that Apple can't get more paying subscribers than Spotify even though Apple is going to do a Microsoft IE embedded in the OS play here, pick your own number of paying subscribers and multiply by $2.80. How many months does it take to wash $18 million?
Of course, the $18 million number could be wrong too. But at my $70 million/month guesstimate or your own guesstimate more or less than that, even $50 million in free trial cost is going to get washed in a few months. After that month or a few months post-trial period, that $70 million or your number just keeps on coming, month after month. About the time the free trial ends for millions of us, iOS9 rolls out with this service as a default for EVERY SINGLE iDevice user on the planet. So it seems favorable in the math and revenue potentials that whether it's going to be my guesstimate of $70 million/month or your number, whatever the number, it's more likely to go up than down. The hard cost expensed for this free trial- whatever it is- can only go down since it doesn't accumulate indefinitely.
So even if the most pessimistic among us can find a way to get that total cost to- say- $100 million, my guess of $70/million recoups it in revenues in 2 months and my guess that there are healthy profit margins in this for Apple implies maybe 3-5 months to recoup it in profits. If your guess is <$70 million, even vs. $100 million, it shouldn't take that long for Apple to wash the expense... and then your revenue or mine will just keep on pouring in... for (potentially) years to come.
Looking forward from this point through our short-term lens, all we see is some ambiguous level of cost for Apple with no revenue. But hop in your time machine and jump forward 12 months. Then, look backwards at the 3 month trial vs. what Apple has made for the 9 months since. That will create an entirely different view. Hop forward another 12 months and take another look at 21 months of revenues vs. the 3-month trial. Make up your own numbers. Apple will do great here unless pretty much nobody wants to pay for this streaming music service that will be the default in iOS9.