Spotify’s Glass Ceiling: The Secret Plan to Break the Subscription Model

"The Playlist" / Netflix With 761 million monthly users and revenues of about $4.5 billion in the first quarter of 2026, Spotify is cementing its position as the undisputed force in the global streaming market. The Swedish audio giant reports 293 million paying subscribers, far ahead of Apple Music and YouTube Music. That strength is reflected in rare confidence on Wall Street: the stock jumped 15% when the company’s strategic plan for 2030 was unveiled, and a leading American investment bank raised its price target to $625 on the back of double-digit growth forecasts. So what is that plan? Spotify is advancing licensing agreements, including with Universal, that would allow users to create cover versions of familiar songs using artificial intelligence, directly in the app, as a paid premium add-on. The content will remain a "closed garden": no download option, internal sharing among fans only, and a compensation mechanism for the original artists. The move marks a dramatic shift, from a distribution and recommendation platform to a content production platform, with potentially far-reaching economic and cultural consequences. The question is whether there is באמת demand for a secondary market of AI-generated music, or whether Spotify is simply bumping up against a glass ceiling and looking for an alternative revenue engine. And what about the musicians themselves? At the end of all this, are they facing an economic upside, or just another promise? To unpack that question and analyze where the industry is headed, we spoke with two lecturers from "HaKatedra" at the Eretz Israel Museum, one of Israel’s leading enrichment centers, Prof. Lior Zalmanson, senior lecturer in the Coller School of Management at Tel Aviv University and head of the research lab for AI processes, and music producer and radio host Nadav Rabid, who previously managed Galgalatz. Rabid believes Spotify’s move is intended to respond to growing competition from new players such as Suno and Udio, music generation models alongside tools that make it possible to produce a demo from an existing recording, separate stems in songs, sample instrumental passages and alter voices. Spotify / Walla! Technology, Yinon Ben Shushan Suno alone has become a real threat within two years: it has crossed the 2 million paying subscribers mark and reached $300 million in annual recurring revenue, a 50% jump in one quarter. About 100 million users have already tried it, and it generates about 7 million songs a day. In November 2025 it raised $250 million at a valuation of $2.45 billion. "Spotify is essentially eliminating its classic distinction between the creator and the listener, and positioning itself not only as a conduit between them but as a content production platform in its own right. To understand where it gets the confidence to do this, you need to know what Spotify presented at that event under the name 'Large Taste Model', a model the company says is based on 3.4 trillion daily taste signals. In simple terms, Spotify is realizing what we always thought it might do, use the analysis of our behavior when we listen to music on it to build the ultimate LLM of musical taste," Rabid says. Prof. Zalmanson offers a more critical view of the model and points to the risk of shifting power structures: "When we talk about an algorithm that claims to understand the 'ultimate taste,' we have to ask who is really managing whom. Spotify no longer merely reflects our taste, it actively shapes it through recommendations and algorithmic decision-making. Once it becomes a content producer as well, the algorithm becomes a kind of 'boss' that dictates the boundaries of creation and listening. This has deep implications for the authenticity of the human connection we seek in music." But the economic explanation does not end there. Spotify has in fact reached saturation in its basic subscription model, but instead of declaring its end, it is making a strategic split, expanding side revenue streams and moving to a tiered products-and-services model instead of a uniform all-catalog access model for a monthly fee: "Every streaming service today offers the same catalog, and the pace of growth in new subscribers is slowing. The next growth, they are betting, has to come from all kinds of additional services and benefits. In other words, from adding payment layers on top of the basic subscription," Rabid adds. One of the most interesting issues surrounding music created with artificial intelligence is the dramatic gap between the volume of songs being generated and the volume of listening they receive. In just two years, AI music has gone from a marginal experiment to a massive flood: if at the start of 2025 around 10,000 generative songs a day were being uploaded to Deezer, accounting for 10% of the new catalog, by the end of the year the number had risen to 50,000, about a third of uploads. As of spring 2026, that figure has reached a peak of 44%, with about 75,000 new AI songs every day. However, in practice, the share of listening to this music remains low and steady, stuck at just 1% to 3% of all streams on the platform. "I find it hard to imagine a world where people want more and more versions of familiar songs by Queen and Billie Eilish," Rabid said. For Prof. Zalmanson, this gap is not accidental, but stems from a basic psychological principle tied to the perception of effort: "The reason listening to generated music remains low lies in the way we perceive value in work and creation. As human beings, we value art not only for its final product and aesthetics, but also for the human effort, difficulty and investment we recognize behind it. When the audience knows a song was created in seconds by a prompt, the magic and vulnerability that underlie the emotional experience disappear. Without the 'fingerprint' of human effort, music has a hard time generating real resonance." And yet, there is a trend here that cannot be ignored. During the COVID lockdowns, daily listening on Spotify and other streaming services surged, with functional playlists for sleep, concentration and exercise at the center. Investigations revealed that the company filled those playlists with generic, low-cost background music purchased from production houses and distributed under pseudonyms. In effect, subscribers paid billions to listen to industrialized elevator music produced at minimal cost. Is the gap between the supply of AI music, 44% of uploads, and listening to it, 1% to 3%, a permanent state, or an early stage before audiences get used to it and listening rates jump? Rabid assesses: "It’s impossible to know what will happen. I estimate listening will rise, certainly in places where there is functional music. And I also think there will be 'legitimate' artists who find interesting ways to make music with AI, and they will find an audience." And if listening rates do rise, will that stem from genuine demand and the blurring of the line between human and generated songs, or simply from platforms actively pushing the content into playlists to save on royalties? Rabid: "I assume the platforms will try to push cheaper content wherever they can, and then it depends on how much pushback there is from listeners, who according to surveys do not want this. But as the tools improve, they will stop being something special, and more and more artists will use them as part of the creative process. In that sense the boundaries will be blurred, because there will be songs in which some of the parts are, say, generated, and the creator will be much more involved and in control of the final result." "I find it hard to imagine a world where people want more versions of familiar songs." Nadav Rabid / Courtesy of the subjects How will the rise in listening to generated music affect artists’ ability to earn a living, when Spotify pays about a third of a cent per stream, around $3 for every 1,000 streams? Already today, many artists cannot make a living from music, certainly not from streaming. Rabid argues that in percentage terms, only a few can rely on streaming as a source of income: "Clearly, the bigger and more diluted the pool becomes, the harder it will be, but even here we are talking about an intensification of a process that started earlier. Music was never a 'safe career.' The democratization of production tools allows more people to enter the field and create music, but in the end it is a fairly brutal pyramid, and there have always been a few who made it to the top and even fewer who managed to stay there." Prof. Zalmanson offers a more complex view of the idea of "democratization" and explains why it is a trap: "The term 'democratization' is accurate and precise when it comes to removing barriers to entry. Generative tools do indeed give everyone the ability today to create, produce and express themselves musically with unprecedented ease, and that is a welcome trend in itself. However, in practice, this often turns out to be an illusion. While production power is dispersed and reaches everyone, the power of distribution and financial reward becomes more concentrated and aggressive than ever. In the end, the biggest winners of this 'democratization' are the platforms themselves, which profit from flooding the market with endless content at near-zero cost, while human creators are forced to compete with the machine and with each other in a saturated market. The democratization of creation tools, unfortunately, does not bring with it a democratization of wealth distribution, but leaves most of the power in the hands of the tech giants." And who is already paying a heavy price for the spread of artificial intelligence? The most dramatic impact of AI is already being felt very tangibly in fields such as jingles, loops for commercials, music for promotional videos and background music for web videos, which once provided steady income for composers, studios and sound technicians and are now going through rapid replacement. Economic studies of the production music sector confirm the depth of the crisis: AI tools are expected to take over as much as 60% of the industry’s revenue. According to Nadav Rabid, the main victims today are already the professionals who made a living from utility music, such as session musicians, commercial composers and creators of soundtracks for low-budget films. For an independent producer or an advertising agency, the choice between paying thousands of dollars to a human musician and using a cheap algorithmic solution is almost always decided by budget considerations.