Value – April 2025

Over the past 150 years, the music ecosystem has evolved into a thick forest of technologies for music creation, generation, reproduction, distribution, and performance.  Each new technology has increased the number of ways businesses could create value from music: by playing it in the background in public spaces, broadcasting, selling albums or individual tracks, or streaming to consumers.  

Whatever the technology, value has always been created at the point of contact with listeners, and this has guided the collection of license fees from the business use of music. In this blog I analyze what has changed with the introduction of digital technology: The ability to profile individual customers and to hide the source of value created by music. I describe how digital technology can be used to create transparency in value creation and propose an approach to the latest technology revolution: AI-generated music.  

The technology roller coaster

Before the advent of recording and broadcasting technology, professional musicians generated value by live performances in private homes, theatres, and public spaces. When at the end of the 19^th century, recording was invented, listeners could enjoy music at other places and times than where the performance occurred. And when broadcasting was introduced not much later, they could enjoy live music even when not present at the performance.

This situation remained stable until the early 2000s, when digital streaming was introduced. Listeners could now enjoy a vast catalog of music anywhere anyplace using their own devices. Live and recorded performances could be distributed and played at low cost. 

As a result of all the digital transformation of the music ecosystem, the major source of revenue for musicians has shifted back to live performance for an audience.

But the digital roller coaster has not stopped and today, generative AI companies scavenge the Internet for digital music recordings to train their systems on. A GenAI system can generate music similar to what it has been trained on. 

What has changed, what has remained constant? First, the factors that have remained constant under all changes in technology, including the digital changes. 

1.  Businesses use music to create revenue

Look at the central part of the following diagram. (I will discuss the other parts shortly.)

As always, event organizers generate revenue from entrance fees and the sale of drinks to the audience. 

Radio and TV broadcasters generate revenue from advertisements. 

Bars, discos, supermarkets and other venues where background music is played, use music to attract customers, retain them, and influence their spending. 

Music retailers generate revenue from selling music tracks in physical or digital formats. 

Streaming providers generate revenue from subscription fees and advertising. 

Audio-visual content creators generate revenue by including music into their movies, commercials, and videos, each of which has its own business model.

In all cases, music adds value for listeners, and this value is monetized by commercial users of music. 

(2) Revenue is generated at the customer interface

The second constant factor is that commercial value is generated at the point of contact with listeners. Some people view this as a characteristic feature of the digital platform economy, not just of the music economy. For example, Booking.com, Amazon.com, Airbnb, Uber and other digital platforms inserted themselves between producers and consumers of value. They acquire customers, take revenue from them, and pass some of this on to the owners of the assets that actually create customer value ⎯hotel owners, sellers, property owners, car owners.

However, value generation at the customer interface is a feature of the wider content economy, and it predates digital technology. In the older analog economy too, event organizers, book publishers, music publishers, record companies, analog radio and TV, and other distributors of content sit between the creators and consumers of content. They take revenue from the customer and pass some of it on to the owners of the assets with which value is created. This is a feature of the content economy, and digital technology has not changed this. Digital technology just expanded the scale at which it happens and scope of assets for which it can happen.

In short, in the older analog economy as well in the digital economy, the owner of the customer interface generates revenue. But digital technology does add a twist that enhances money-generating power at the digital consumer interface.

Individual digital consumption can be monitored

What is new in the digital economy is that individual consumption events can be registered. When selling an e-book, the seller can even follow individual reading behavior. When does the reader read the book? Where? For how long? How many pages at a time? How long per page? When do they turn a page? Do they go back to earlier pages?

The same holds for selling music in a digital format, and for streaming music. The consumption of digital radio, digital TV, digital music streams and digital AV content can be monitored individually. These companies not only own the customer interface, but they can also monitor individual listener behavior. They can measure for each listener to which tracks they listen, when they listen, where they listen, how long they listen, how often they listen, etc. 

This individual data can be monetized by the owner of the customer interface by building customer profiles. The digital economy has been called a surveillance economy, but a more accurate term is profiling economy. In the non-digital content economy, the behavior of individual audience members cannot be monitored. In the digital economy, individual behavior can be monitored.

The profiling economy allows new kinds of value generation for digital content companies. Targeted advertising, social media engagement, personalized interfaces, individualized music recommendations, personalized listening profiles and other targeted services are great ways to generate revenue and glue consumers to their digital service providers.

Using digital monitoring to create transparency

There is another kind of monitoring in the music ecosystem, namely of the commercial use of music. Intellectual Property (IP) holders, typically artists, authors, and producers of music, need to know when their music is used commercially, in order to provide licenses and collect license fees.  This is very labor-intensive, it has been delegated to Collective Management Organizations (CMOs), who do this on behalf of a large number of IP holders.

CMOs license businesses to use music for a commercial purpose and pass on the license fees to the intellectual property (IP) holders of the music. In addition, as shown in the above diagram, record companies and music publishers also collect some of the license fees, independently from CMOs.

In the analog economy, this is a manual process but with the advent of digital technology, collecting data about music use can be individualized too. Playlists reveal what has been broadcast, and a monitoring technology called fingerprinting can be used to track what is played where and when as background music. CMOs are increasingly using these data collection technologies to grant licenses and collect license fees. So far, digital technology has been labor-saving.

CMOs work regionally, which means that a musician only receives data about music use in the country of the CMO where the musician is registered. License fees from music use in other countries are received without indication of the consumption events from which they originated. The Music360 project makes this transparent, by integrating data about music use from CMOs in different countries. Using the Music360 platform, authors, performers, and producers can monitor the use of their music across all countries covered by CMOs who have joined the platform. 

What about AI?

By training an AI system on a large sample of music tracks, it can generate new music similar to the tracks in the sample. As usual, businesses can use this music to generate revenue, and this revenue is generated at the customer interface. Should gen-AI companies receive a license fee for this? And should they pay license fees to the IP holders whose music they used to train their systems on?

A gen-AI company could argue that, just like a human composer, its gen-AI system has listened to a lot of tracks but creates new work that differs from anything it has heard, and that is its own IP. In this reasoning, if AI-generated music is used commercially, the gen-AI company should be paid a license fee. The company owning the gen-AI system could become a member of a CMO to get paid for the commercial use of its music.

This view is inconsistent with the fact that some music generated by a gen-AI system is in fact very similar to existing music. This is to be expected, for a requirement on gen-AI systems is that they should not “hallucinate” but produce correct answers, i.e. answers that agree with the data they have been trained on. 

Another argument of a gen-AI company may be that they are in the position of AV producers, who use existing creative material, add something creative, and then claim IP over the result. According to this reasoning, they should receive a license fee when the result is used commercially. It remains to be seen whether this is a valid argument.

But at least AV producers pay license fees to the holders of IP of the music they use. And these IP holders can prohibit that their material be used, in which case AV producers don’t use it.

For musicians whose music is included in a training sample, there are two possible ways to pay them license fees.

1. For every use of an AI-generated music track, all IP holders in the training set are remunerated. This requires the ability to trace an AI-generated music track to the set of all tracks that its generating AI system has been trained on.
2. The gen-AI company obtains a “training license” from all IP holders in the training set, just as an AV producer obtains a synch license from all IP holders of music included in an AV production. 

Both options are technically feasible. Obstacles come from the business model of gen-AI companies. In their rush to be first-to-market, AI companies prefer to treat all online music as free, and the training sample as a commercial secret. Regulatory bodies must design enforceable governance to push them out of this extractive business model.