How much does it cost to run a streaming music service? If you’re Spotify and you’re streaming billions of songs every month, as the company disclosed in a blog post on Wednesday, you’re spending millions in royalties while seeking to find the right offering to connect with consumers.

Let’s run through some simple math in a realistic scenario (just for demonstrative purposes). First, let’s assume Spotify is streaming two billion songs per month. Second, let’s assume 90% are interactive and 10% are non-interactive. (Although it’s better known for streaming on demand, the service does have a radio function.) Third, let’s use royalty rates typical for the U.S. Lastly, let’s ignore the threshold at which a percent-of-revenue payment to rights holders may kick in (instead of paying a per-stream fee).

For the interactive royalty rate, let’s use 0.75 cents per stream (it’s the midpoint of a typical range of 0.5 to 1.0 cents per stream). For the non-interactive rate let’s use the 0.093 per stream rate set earlier this year by the Copyright Royalty Board for non-interactive webcasters (such as Pandora and Last.fm). A blend of the two royalty rates, weighted at 90/10 for interactive/non-interactive, is 0.6843 cents per stream.

Two billion songs per month generates royalties of \$13.7 million per month, or \$164 million per year, given the assumptions mentioned above.

Next, let’s say the company wanted to operate on a 40% gross margin (meaning the cost of generating each dollar of revenue was 60 cents). A 40% gross margin would require about \$274 million in total revenue. How would Spotify generate that much revenue? Signing up 1.52 million paying subscribers at \$15 a month would generate that much annual revenue. Or it could run 13.7 billion audio ads at \$20/CPM. (Which is not only a lot of ads at a high CPM, but would mean there would be an audio ad every 1.75 song streams. Spotify runs only three minutes of audio ads every hour.)

Subscribers who do not pay the \$15 per month would use the free, ad-supported version of the service. And Spotify has far more non-payers than payers. In fact, it has been reported that only 2% of Spotify users have upgraded to the premium, paid service. So the actual revenue mix, which is the stated plan of the company’s executives, is a mix between paid and free users.

All of these numbers are offered just to show the costs involved in streaming music and the revenue that must be generated to turn a profit. They’re hypothetical numbers, based in reality, that point to a clear conclusion: streaming music isn’t cheap.

But those examples are for the more costly interactive streaming rates. What if Spotify looked at the royalty burden and decided to change its product in order to save costs? Let’s say the company focused on technological innovations that made a non-interactive service work so well that users didn’t mind not being able to select songs. Switching to a non-interactive business model would save Spotify millions. Royalties paid would drop to less than \$2 million per year, assuming the 0.093 per stream rate mentioned above.

As licensing deals are currently set up, a company that chooses to offer interactive streams is choosing the higher of two royalty rates. But it is able to offer a better user experience and probably generate interest from more avid (and thus more valuable) music fans. The hitch is the extra value offered must be converted into extra revenue in the form of subscriptions (or through selling downloads, tickets or other items). It’s a brutal numbers game that requires innovative products and massive scale to succeed. Maybe somebody will make it work one of these days.