How Streaming Royalties Flow: Who Gets What After You Press Play

Ask what a stream pays and you'll get a number: a third of a cent, half a cent, four-tenths. Every one of those numbers is real in the sense that someone divided a payment by a play count and got it. None of them is a rate. Nobody at a streaming service ever agreed to pay a third of a cent, and no contract anywhere contains that figure.

That distinction is the whole subject. Once you see that a play is a claim on a pool rather than a purchase of a unit, the strange features of music streaming stop being strange — why the number drifts down in good years, why two artists with identical play counts get different checks, why songwriters and performers are paid by entirely different institutions on entirely different schedules.

Follow one month's subscription payment all the way through and the machine becomes legible.

Stop one: the pool

Your subscription fee does not go anywhere in particular. It joins every other subscription fee and every dollar of advertising the service sold that month. From that gross figure the service sets aside a royalty pool — the portion owed to everyone whose music was played — and keeps the rest for its own costs and margin.

The pool is then divided by streamshare. Spotify's public description of its own model is unusually blunt about the arithmetic: "Artists get paid based on their share of total streams. If your music accounts for 1% of all streams, you get roughly 1% of the royalty pool." For scale, the company reported that in 2025, artists who accounted for one in every million streams on the platform generated over $11,000 on average.

Two consequences fall straight out of that formula, and they explain most of the confusion.

First, your individual listening does not determine where your individual money goes. Under a pro-rata pool, a subscriber who plays one obscure record on repeat all month contributes to the pool, and the pool pays out according to global play shares. Your money is diluted into everyone's.

The fork: one recording, two copyrights

Now the pool has to be split, and here the system stops being a single system.

The U.S. Copyright Office's 2015 study of music licensing lays out the structural fact that everything else depends on: a recorded song is two separate copyrighted works. There is the musical work — the composition, the notes and words as written, typically owned by songwriters and their publishers. And there is the sound recording — the specific captured performance, typically owned by a record label, with the featured artist paid out of the label's share under contract.

Those two copyrights are licensed by opposite mechanisms.

One play, two copyrights

A single stream triggers two separate licenses under two different legal regimes

The recording side: a private negotiation

For on-demand services, the sound recording is licensed in the open market. There is no statutory rate. Each service negotiates directly with each label — the three majors and hundreds of independents — and the terms of those deals are confidential.

That opacity is why so much of the public argument about streaming pay is conducted with inferred numbers. It is the same problem that shows up whenever a negotiated wholesale layer sits between a consumer price and a producer, as it does in the interchange fees behind a card swipe.

Where U.S. recorded music money came from

Retail value of U.S. recorded music revenue by format, 2024 ($ millions)

The composition side: a rate somebody else sets

The song is handled the opposite way. Under Section 115 of the Copyright Act, a service can obtain a compulsory license to reproduce and distribute musical works without the publisher's permission — the publisher cannot refuse — and in exchange the price is set by a federal tribunal, the Copyright Royalty Board, in five-year proceedings.

The rates in force are codified at 37 CFR Part 385. The headline figure is a percentage of the service's revenue, stepping up across the term: 15.1% in 2023, 15.2% in 2024, 15.25% in 2025, 15.3% in 2026 and 15.35% in 2027. Two refinements matter mechanically. The pool is calculated as the greater of that revenue percentage or a figure keyed to total content cost — what the service pays for sound recordings, at 26.2% of that cost for standard subscription offerings — so a service cannot shrink its songwriter obligation simply by reporting less revenue. And the amount is an "all-in" figure: the public performance royalties the service already pays are subtracted from it, with a per-subscriber floor (for a standard portable subscription, $1.10) setting a minimum below which the mechanical share cannot fall.

That structure is the same species of negotiated-then-frozen arrangement described in the old statutes that quietly price ordinary purchases: once a rate is set by tribunal for a fixed term, it becomes a fact of the industry until the next proceeding rather than something the parties can revisit.

Getting that money to the right songwriter used to be the harder half. The Music Modernization Act of 2018 created a blanket license, available from January 1, 2021, administered by a single designated body — the Mechanical Licensing Collective — which collects the mechanical royalties, maintains a public database of works and owners, and matches recordings to compositions. As a condition of the liability protection the Act offered, services also handed over the royalties they had never been able to match: $424,384,787 transferred to the MLC on February 16, 2021, covering uses before 2019.

That number is worth sitting with. It is not a scandal so much as a measurement of an infrastructure problem — for years, nobody had a reliable, industry-wide map from a recording to the humans who wrote it. Matching is a database problem of the same shape as the identity-matching described in how credit bureaus assemble a file on you: the hard part is not the money, it's the join.

The third system: SoundExchange

There is a third path, and it is the only one where federal law names the performer directly.

For noninteractive digital transmissions — internet radio and satellite radio, where you cannot choose the next song — sound recordings are covered by their own statutory license, and the money is collected centrally by SoundExchange. Section 114(g)(2) of the Copyright Act then splits the receipts by formula rather than by contract.

The one split written into federal law

Statutory allocation of receipts from the noninteractive sound recording license, 17 U.S.C. 114(g)(2)

Fifty percent to the copyright owner, 45% paid directly to the featured artists, and 2.5% each into escrow funds for the session musicians and backing vocalists who played on the record. The featured artist's 45% does not pass through the label and is not reduced by an unrecouped advance. That is a meaningful structural difference from the on-demand side, and it is entirely a product of the fact that Congress wrote a formula here and left the on-demand market to contract.

Whether a musician is paid by statute or by contract is not a detail. It determines whether the money arrives directly, and whether an old advance can absorb it.

Where the arguments actually live

Once the plumbing is clear, the disputes sort themselves into three, and each is a design question rather than a moral one.

Pro-rata versus user-centric. The pool model divides by total plays across the whole service. The alternative — usually called user-centric — would divide each subscriber's own fee among only the artists that subscriber played. Advocates say it matches money to intent and helps artists with small, devoted audiences. Critics say it would raise administrative cost substantially, shift money in ways that are hard to predict, and reward listening volume differently rather than more fairly. Both sides are describing real effects of the same change.

Thresholds and filters. Some services have adopted minimum play counts before a track earns, and demonetize material they judge to be functional noise or fraud. The stated purpose is to stop a growing tail of automated streaming from draining a fixed pool. The objection is that a rule aimed at fraud also catches very small legitimate artists, whose income was already the smallest.

Statutory versus market. Publishers point out that they cannot say no under Section 115 and so cannot bargain; services point out that a compulsory license is the only reason a catalog of tens of millions of songs can be cleared at all. Both are true, and it is why the rate proceeding, not the marketplace, is where that fight happens.

Nobody at a streaming service ever agreed to pay a third of a cent, and no contract anywhere contains that figure.

Reading the number when you next see it

So when a per-stream figure appears, three questions make it interpretable.

The system isn't hiding a rate. It genuinely doesn't have one. What it has is a pool, a share, and two legal regimes that were never designed to meet.

Primary Sources

1. U.S. Copyright Office, Copyright and the Music Marketplace (2015)
2. 37 CFR Part 385, Rates and Terms for Use of Musical Works Under Compulsory License for Making and Distributing Phonorecords
3. 17 U.S.C. 114, Scope of exclusive rights in sound recordings
4. U.S. Copyright Office, The Music Modernization Act - Section 115 Blanket License
5. RIAA, 2024 Year-End Music Industry Revenue Report
6. The Mechanical Licensing Collective, historical unmatched royalty transfer