Miso - The Financial Rails for Music

A dollar spent on music today reaches the people who made the music somewhere between two months and two years from now. On the way it passes through six or seven organizations, and every one of them charges twice: a percentage of the money, and the use of the money while it waits.

The industry argues about the first price. The second one is the reason the first one never falls.

Music never got a rail

A rail is the shared infrastructure that value settles on. It decides how money moves, when settlement is final, what it costs, and what gets recorded. Visa and Mastercard are rails. ACH is a rail. Bank wires are rails. FedNow is a rail. Your bank app, Venmo, and Stripe are not rails; they ride on top of them.

Every asset class that needed settlement to understand what was moving eventually got infrastructure built for itself.

US equities clear and settle through DTCC¹: NSCC nets and guarantees broker-to-broker trades, DTC holds the securities and moves book-entry positions between owners.
US Treasuries are held and transferred over the Fedwire Securities Service², which moves the security and the cash together in one delivery-versus-payment instruction.
Foreign exchange settles through CLS³, whose payment-versus-payment design settles over $8 trillion a day across 18 currencies.

One rail per asset class, and in each case the rail knows the shape of the asset. The position, its owner, and the matching cash leg move in a single operation. No separate industry has to reconstruct afterward who the trade belonged to.

Music never got one. It runs on general-purpose payment rails with an accounting industry bolted on top.

A payment rail knows how to move a dollar. It has no idea that a recorded song is two separate copyrights, the recording (the master you hear) and the composition (the underlying song), that each is divided among writers, producers, and performers in fractions negotiated years earlier, and that the two halves are usually owned and administered by different companies. The rail is blind to every part of that.

So the money settles first, in a lump, into someone’s bank account. Only then does the work start: matching plays to works, works to owners, owners to percentages, percentages to payment addresses. Every one of those steps is a separate company with its own database, its own accounting period, and its own bank account holding the money in the meantime.

The whole royalty administration industry lives in the gap between “the money arrived” and “the money reached its owners.” So does the money.

Follow the dollar

Because a song is two copyrights, the money follows two pipes. The recording flows from the platform to whoever licensed it, then to the artist under their deal. The composition splits again into two kinds of royalty. Performance royalties are owed when a song is played in public, and are collected by performing rights organizations, or PROs (ASCAP, BMI, SESAC, GMR), which sell radio stations, bars, and streaming services a licence to play anything they represent. Mechanical royalties are owed when a song is copied, which a stream counts as, and in the US are collected by the Mechanical Licensing Collective. From there the money goes to publishers, to their local partners abroad, and finally to writers. The work is one thing. The ledgers are six.

The published numbers, hop by hop.

The composition side (a US stream, per dollar of music revenue):

Hop	Who holds the money	How long it sits	What they keep
1. Platform	The streaming service	Pooled monthly	~30% of music revenue⁴
2. Mechanical (copying)	The MLC	Monthly cycles; pays members monthly	An admin fee it does not publish
3. Performance (public play)	A PRO	3.3–5.5 months after the quarter it was earned in, by BMI’s own published schedule⁵	~15% (BMI, after its 2022 for-profit conversion⁶; it passed ~90% before)
4. International	The foreign collecting society → the writer’s home society → the publisher’s local partner	12–18 months, commonly cited⁷	An admin fee in each country, plus a local partner’s commission (commonly 15–25%)
5. Publisher	The publisher	Their own calendar	Half the writer’s share in a standard deal (15–25% if the publisher only handles the paperwork)
6. Unmatched (“the black box”)	The MLC, when it cannot work out whose song it is	Until at least 2028, then split by market share	Nothing, but the money stops belonging to a particular song

The master side (a US stream, per dollar of music revenue):

Hop	Who holds the money	How long it sits	What they keep
1. Platform	The streaming service	Pooled monthly, then divided by share of total plays	~33% of music revenue⁴
2. Label	The label	Its own statement cycle, and nothing at all until the label has earned back the advance it paid	On paper the artist gets 10–25% of the retail price, though it is often calculated on the lower wholesale price. Subtracted first⁸: money held back in case shops return copies, old contract clauses that pay on only 90% of copies sold, and the unpaid balance of advances on past albums
3. Distributor	A self-serve distributor	Their payout cycle, and only once you are owed at least $2 to $50⁹	Anywhere from 0% of sales (for a flat fee of about $25 a year) to 9%⁹
4. Stranded	Any step where the balance is under its minimum	Indefinitely	The middleman and the bank

The spread between best and worst case is enormous. An independent artist using a flat-fee distributor, correctly registered with the MLC, sees most of the money in one to two months. An artist on a major-label deal that also takes a share of touring and merchandise, with publishing collected abroad, gets nothing for years while the label earns back its advance, then 10–25% of a wholesale price after all the deductions above, while the songwriting side takes 18 to 36 months to come back through local partners. That there is no single answer is itself the finding: the rail is not doing the work, so the outcome is whatever the contracts on top of it happen to say.

The third column is the one nobody argues about. Every row of it is a company holding other people’s money on a calendar it sets for itself.

The second price

Money in transit is not idle money. Whoever holds it earns on it. That is the ordinary, well-understood economics of float¹⁰, and it is why insurance companies and payment processors make as much from investing the money they hold as from the business they are nominally in. Nothing about music suspends it.

Several of the collecting bodies say so in their own filings. SoundExchange nets the interest¹¹ it earns on royalties it holds against its own operating costs. ASCAP folds investment income¹² into the pool it distributes. The MLC holds mechanical royalties it has not yet matched to an owner and invests them; critics reading its tax filings¹³ put its securities position past a billion dollars, a figure the MLC’s defenders note commingles matched-but-undistributed royalties and operating reserves with the unmatched funds. The commingling is the point: nobody outside can tell which dollar is which, and the return on all of it accrues to the holder.

Labels and distributors publish no equivalent disclosure, and I am not claiming they slow-walk statements on purpose. The structure is the claim: every layer here holds a pool of money it does not own, for a period it sets, and keeps the return on it.

Put numbers on a single release. It earns $50,000 over a year, and each dollar sits about six months on average before it reaches the people who made the music. At roughly 4%¹⁴, about what a three-month Treasury bill pays, that delay is worth around $1,000 to whoever is holding the balance. A microphone, or ten hours of studio time, transferred from the artist to an intermediary purely as a function of time.

Now scale it to a catalog. $100 million held for six months at 4% throws off about $2 million a year, every year, with no additional work by the holder and no line item on anyone’s statement.

The artist’s loss is larger than the interest, because the interest is only the safe, boring version. The real cost is optionality. Money that arrives in September cannot fund the record you were going to make in March. Artist income turns over unusually fast: it goes straight back out into recording, touring, marketing, and paying collaborators. Delaying it does more damage per dollar than delaying almost any other kind of income.

There is a sharper version of this. A holder who wants more than the safe rate can put the balance somewhere it cannot be pulled out quickly. When an artist’s payment comes due, it gets funded out of the current cycle’s inflows, and the books square later. Everyone gets paid, on time, with money that belongs to the artist behind them in the queue. This is maturity transformation¹⁵, it is what banks do, and it is legal when the assets are real and the arrangement is disclosed. It shares one feature with a Ponzi scheme, dependence on new inflows to meet current obligations, and two things separate it from one: real returns, and honesty about the structure. An artist can verify neither from a royalty statement.

The third price: money that never arrives at all

Some of the money does not arrive late. It does not arrive.

Spotify hosts well over 100 million tracks. Tens of millions of them get between 1 and 1,000 plays a year and average $0.03 a month. Spotify’s own royalties guide says plainly⁴ what happens next: because labels and distributors require a minimum withdrawal, usually $2 to $50, and banks charge $1 to $20 per transaction, “this money often doesn’t reach the uploaders.” The aggregate is about $40 million a year. Starting in April 2024, Spotify’s fix for its least-played tracks was to stop generating the royalty at all.

The same mechanism runs at every hop, and nobody publishes a ledger of what was stranded or where it went.

Unmatched money follows a different path to the same place. When the MLC cannot match a stream to a registered work, the money goes into what the industry calls the black box. Streaming services handed the MLC $424 million¹⁶ of historical unmatched royalties when it opened. The Copyright Office’s report¹⁷ is explicit about the disposition: after a holding period, unmatched royalties are distributed by relative market share to the publishers in the MLC’s records. Money most likely owed to small and unsigned writers routes to the largest publishers, who hold the biggest catalogues and keep the tidiest records. The worse the industry’s records are, the more the majors collect. The first such distribution is not expected before 2028.

Nobody in the chain is paid to be fast

Suppose a PRO cut its distribution lag from five months to five days.

It would give up the float on a quarter’s worth of collections. It would give up the administrative revenue attached to running quarterly cycles. And it would remove most of the reason a songwriter needs it in the first place, because a body that receives money and forwards it immediately is much easier to compare against alternatives than one whose value proposition is bound up in a process nobody can see into.

Speed is a cost to every layer of this system and a benefit to nobody in a position to build it. Ask any of these organizations to invest in faster settlement and you are asking them to fund a reduction of their own revenue, and each can correctly answer that the delay is not theirs alone: the streaming service reports monthly, the collecting society reconciles quarterly, the partner abroad works its own calendar. Everyone is waiting for someone upstream, and everyone earns while they wait.

This is a coordination failure with a subsidy attached. It does not resolve on its own.

The tell is that they have to lobby

Every layer of this system exists by statute or by contract, and the statutory ones were written at the industry’s request.

The PROs operate under agreements struck with the Department of Justice in 1941¹⁸ that still set their rules, and in 1979 the Supreme Court upheld their central product against an antitrust challenge: a single-price licence that lets a broadcaster play everything they represent for one fee. When digital arrived and recordings had no right to payment for being played online, the industry lobbied one into law in 1995¹⁹, creating a brand-new revenue stream whose rates are set by a board it negotiates with. When streaming exposed a gap in copying royalties, the record industry’s and songwriters’ trade bodies united behind one bill, and the 2018 Music Modernization Act²⁰ created the MLC: a collective written into law, with a database, populated by the labels and publishers.

The spending tracks the statutes. The RIAA²¹ spent between $2.4 million and $6.5 million a year on US lobbying from 2001 to 2020, and more than $2.5 million in the first quarter of 2025 alone. The trade’s own description of PROs²² notes that they “lobby on behalf of rights holders, especially in discussions of legal royalty rates.” CISAC²³, the coordinating body for the roughly 228 collection societies, states its mission as providing “a global voice to policy makers.”

Nobody lobbies for the continued existence of Fedwire. No trade association spends millions a year defending the necessity of DTCC. Load-bearing infrastructure does not have to argue for its own survival, because removing it breaks something immediate and obvious to everyone. So when a layer’s position depends on a law it wrote and defends with an annual budget, that budget is telling you how much of the position rests on the service.

The strongest counterexample is the MLC itself, the industry’s own reform. It pays monthly²⁴ and has distributed over $4 billion, which is real. Then look at what the reform produced: another middleman written into law that holds unmatched money, distributes it by market share, and invests the balance. Offered the chance to fix settlement, the industry built a new intermediary to sit on the money. Its earlier attempt at a shared source of truth, the Global Repertoire Database, cost £8 million and died in 2014²⁵ when the societies that were supposed to fund it withdrew, for the obvious reason that a shared ledger is a ledger they do not control. A second attempt, the Open Music Initiative²⁶, launched in 2016 with the majors, Spotify, and YouTube behind it, and went quiet.

Middlemen should be opt-in layers, not gatekeepers

None of this is an argument against intermediaries. It is an argument about what they are selling, and how it is priced.

Every middleman in music is a service provider. That is all a label, a distributor, a publisher, or a collection society has ever been: a company doing work an artist cannot easily do alone. Getting a record into ten thousand stores was that work. So is fronting capital against an unproven album, developing an artist across three records, working a song at radio, clearing a song for use in a film or an ad, or negotiating with a broadcaster on behalf of two million songs at once. Real services, hard to replicate, worth paying for.

What came loose is the connection between the work and the price. A service provider who has to win the job charges for the job. A service provider sitting in the money’s path charges a share of every dollar, forever, whether or not it is still doing anything. Digital distribution removed the physical problem that justified the position, and the position stayed, because the settlement path still ran through it.

That is what leverage means here, and it is not a metaphor. Leverage comes from having somewhere else to go. An artist who can only be paid through a chain of companies has no exit from that chain, so the terms are whatever the chain says they are, and the layers keep their cut by default rather than by merit. This is also why the arrangement has held for so long: not because nobody objected, but because objecting changed nothing.

Permissionless infrastructure is an exit. An artist who can publish a work, fix its splits, take payment, and have it settle to every owner without asking anyone’s permission does not need a middleman in the path. They may still want one in the room. The difference between those two things is the entire argument.

Once the exit exists, the price has to change shape. A service that has to be chosen gets sold the way services are sold everywhere else: a subscription, a fee for the project, a rate for the work, a percentage that ends when the engagement does. A permanent share of every dollar a song will ever earn is not the price of a service. It is the price of a position, and it only holds where the artist cannot leave.

We think that leads somewhere specific, and this part is a forecast rather than a finding. A layer whose remaining function is moving money has nothing left to sell once the money can move itself, and we expect those layers to thin out. What survives is the contribution: the label that funds a record and takes real risk on it, the people with the taste and the relationships, the team that makes the thing better than the artist would have made it alone. That work is worth paying for, and on a neutral rail it gets paid for what it contributes rather than for where it sits.

That outcome is worse for whoever depends on the tollbooth, which is why they are not going to build it. It has to come from outside the institution set, which is why the answer is a protocol and not a committee.

The workarounds are the evidence

People built three things to route around this system. Each one tells you what is broken.

Cross-border. Everything above is the domestic case. Internationally, roughly 228 societies across 111 countries²⁷ are stitched together by bilateral agreements, some written in the 1920s. A play in Germany routes through the German collecting society, then the writer’s own national society, then a publisher, then that publisher’s partner in each country, each taking a cut and adding months; international royalties commonly take 12 to 18 months⁷ after withholding tax and currency conversion. Europe’s 2014 directive²⁸ on licensing across several countries at once exists because a country-by-country structure cannot handle borderless streaming, and it fixed the problem by brokering across hundreds of territories rather than dropping the model. The internet made listening borderless: a fan in Lagos and a fan in Berlin stream the same file. What stayed national is the settlement. Cloudflare stopped making developers pick a datacenter and treats its whole network as “Region: Earth.”²⁹ Music’s money should settle once, everywhere.

The direct purchase. A sale should dodge all of this, and for the first mile it does. Buy an album on Bandcamp and the money lands in the seller’s account within a day or two, at a published fee of 15% on digital sales, dropping to 10% above $5,000³⁰. Then it stops. Bandcamp pays one account. If four people made the record, or the producer is owed a percentage, or the song and the recording belong to different people, Bandcamp’s own help docs³¹ say collaborators divide the money themselves, outside Bandcamp. The split falls to whoever controls the account, paying everyone else by hand. Even the fastest money in music dead-ends at a split the system cannot perform.

Selling the wait back to artists. beatBread, Amuse, Sound Royalties, Symphonic, Unchained, and SongVest will each hand an artist a lump sum today against royalties they have already earned but not been paid. Amuse alone has done more than 50,000 of these³², and its users come back an average of four times. That is revealed demand, priced by someone who cannot see the future and therefore bets low; the lender keeps the spread. Unchained states the model outright³³: it earns “on the difference between the actual music royalties amount and the projected earnings.” An artist gets their own money sooner, at a discount, and the delay that created the need stays exactly where it was.

Why this is buildable now

The problem is a century old. Three things changed at once.

The ingredients are production-grade. Settlement on a public ledger got fast and cheap enough for per-stream and per-purchase amounts. And value can now move on-chain denominated in dollars rather than volatile crypto, which is what regulated stablecoins made possible. An artist thinks in dollars; the rail can settle in them. Meanwhile the rest of finance normalized fast settlement: US securities moved to T+1 in 2024³⁴, and the Fed launched instant payments³⁵. Next to that, a quarterly royalty statement looks antique.

The users are free to adopt it. Millions of independent artists now operate outside the incumbent pipeline and own their masters. Independents already generate about half of Spotify’s $10B+ in royalties⁴. They can choose a new rail from the bottom up, without a single label or society agreeing to anything. Going around a coordination failure is the only way past one.

Agents need deterministic settlement. Two kinds of AI are arriving in music: one generates tracks, the other runs the business. Administration is the broken part, and an agent is well suited to run it, but an agent runs a loop where each call’s result conditions the next step. That needs operations that are deterministic and time-bound: call it, it succeeds or fails now, with a result you can read. Traditional settlement is the opposite, and wrapping it in an API does not help, because the API returns “pending” over the same swamp.

What we built

Miso is that rail.

A work’s ownership is recorded as fixed-supply shares on a public chain, with the rights structure built into the objects rather than reconstructed afterward. Creating a Recording grants the Composition’s immutable cut as recording-share ownership. A Release stores a flat, ordered tracklist and the exact percentage owed to each party, fixed before any money moves. The ownership graph exists before the first dollar does, which inverts the order everything above depends on: settlement stops being the thing an industry has to interpret after the fact.

When a Release is paid, the payment lands at that object’s address. If its admin capability is held in a Vault and the revenue plugin is installed, anyone can advance the next step: immutable track splits send value to Recording addresses, Recording plugins fold those balances into royalty pools, Composition-owned recording shares can route their rewards into the Composition pool, and shareholders claim from the pools. Each step either completes in full or does not happen at all, and takes its destination from the work itself, so a caller can advance settlement but cannot redirect it.

Two limits, stated plainly. The lifecycle is staged: a bare payment does not push money to every person in one transaction, and claims remain pull-based. And the rail enforces the objects and splits the parties create. It does not adjudicate who wrote the song.

What it does remove is the discretionary gap. There is no layer that decides when your money moves, no calendar you cannot see, no balance held below a minimum nobody publishes, no pool of unmatched funds earning a return for its custodian. The deposit is visible, the next destination is fixed by the work’s own identity, anyone can advance an installed routing step, and holders claim from pools they can inspect.

The rails are the product

Music does not need another intermediary promising to pay a little sooner. It needs rails where the money was never held up in the first place, and where the layers that remain earn their place by being useful rather than by standing in the path.

For a century the obstacle was that a shared source of truth required persuading rival institutions to trust one another’s books, which meant one of them had to be too big to fail, which meant the ledger and the skim were always the same asset. That constraint is gone: a public ledger with rules in code and state anyone can verify is cheap, and splits, routing, and claims are programs that run on top of it. The missing piece was never payment. Payment rails have existed for decades. The missing piece was a rail that knows what a song is.

Wikipedia (DTCC) · Depository Trust & Clearing Corporation · NSCC netting/guarantee and DTC book-entry settlement for US equities. ↩
Federal Reserve Financial Services · Fedwire Securities Service · book-entry transfer and delivery-versus-payment for US Treasuries. ↩
CLS Group · CLSSettlement · payment-versus-payment FX settlement, $8T+/day across 18 currencies. ↩
Spotify · Royalties guide · self-reported revenue splits, the 1,000-stream eligibility threshold, minimum-withdrawal stranding, and the independent-artist share of payouts. (Split percentages are MIDIA Research estimates; the ~70/30 platform framing, the 1,000-stream threshold, minimum-withdrawal stranding, and the independent-artist figures are established.) ↩ ↩² ↩³ ↩⁴
BMI · Creator resources · published quarterly royalty payment schedule. ↩
Music Business Worldwide · BMI’s for-profit margin increase · BMI’s 2022 conversion to for-profit raised its retained margin from about 10% to 15%. ↩
BMI · International royalties FAQ · “This process can take 12-18 months depending on the country of origin.” BMI retired the live page; link is the 2026-05-09 archive snapshot. ↩ ↩²
ASCAP · Money from recording · what a label subtracts before paying an artist: money held back against returns, clauses paying on only 90% of copies sold, and unpaid advances from past albums. ↩
DistroKid · Pricing · flat-fee DIY distribution, 0% commission on store revenue against a competitor set that runs up to about 9%. ↩ ↩²
Investopedia · Float · the general mechanics of float, money held in transit that its holder can invest before forwarding it. ↩
Billboard · SoundExchange 2017 financials (IRS filing) · SoundExchange nets investment income earned on held royalties against its administrative fee. ↩
ASCAP · Royalties and payment · investment income is folded into the credit pool distributed to members. ↩
Hypebot, reporting the MLC’s IRS Form 990 · The MLC, holding $1.2 billion, turning into a hedge fund? · critics’ reading of the filings puts the MLC’s invested-securities position past $1B; the MLC’s defenders say the figure commingles matched-but-undistributed royalties and operating reserves with the genuinely unmatched funds. (disputed, attributed to critics’ framing) ↩
US Treasury · Daily Treasury Par Yield Curve Rates · what a safe short-term government loan pays, about 4%. ↩
Wikipedia · Maturity transformation · the banking mechanism of funding near-term obligations from continuing inflows. ↩
The MLC · MLC receives over $424M in unmatched black box streaming royalties · the historical unmatched-royalty balance transferred to the MLC at launch. ↩
U.S. Copyright Office · Unclaimed royalties: report and recommendations · market-share distribution of unmatched royalties, with a first distribution not expected before 2028. ↩
Wikipedia · Broadcast Music, Inc. · the 1941 DOJ consent decree and the 1979 Supreme Court ruling upholding the blanket license. ↩
Wikipedia · Digital Performance Right in Sound Recordings Act · the 1995 statute creating a digital performance right for sound recordings. ↩
Wikipedia · Music Modernization Act · the 2018 statute that created the MLC. ↩
Wikipedia · Recording Industry Association of America · RIAA’s annual US lobbying spend. ↩
Wikipedia · Performance rights organisation · PROs’ lobbying function on statutory royalty rates. ↩
CISAC · CISAC · CISAC’s stated mission as “a global voice to policy makers.” ↩
The MLC · The MLC · monthly distribution cadence and cumulative distributions over $4 billion. ↩
Music Business Worldwide · Who will build the music industry’s global rights database? · the Global Repertoire Database’s 2014 collapse after roughly £8 million spent. ↩
Wikipedia · Open Music Initiative · the 2016 launch of the open-API rights-data initiative. ↩
CISAC · CISAC Global Collections Report 2025 · 228 member societies across 111 countries and territories. ↩
UK legislation (EU Directive mirror) · Directive 2014/26/EU · the EU’s multi-territorial licensing directive. ↩
Cloudflare · “Region: Earth” · Cloudflare’s framing of its network as a single global region. ↩
Bandcamp · What are Bandcamp’s fees? · 15% fee on digital sales, dropping to 10% above $5,000. ↩
Bandcamp · Do you support split payments for collaborative releases? · collaborators must divide payments themselves outside Bandcamp. ↩
Amuse · Amuse surpasses 50,000 automated royalty advances · scale and repeat usage of Amuse’s Fast Forward advances. ↩
Unchained · Advances · Unchained’s stated model, earning the spread between projected and actual royalties. (estimate) ↩
Investor.gov (SEC) · New T+1 settlement cycle: what investors need to know · US securities settlement compressed to T+1, effective May 28, 2024. ↩
Federal Reserve · About the FedNow Service · real-time gross settlement launched July 2023. ↩