Warner Music Group CEO Robert Kyncl published a memo to shareholders yesterday, the day of WMG’s annual stockholder meeting and the day before he speaks at the Morgan Stanley TMT conference later this afternoon. Universal Music Group reports its Q4 and full-year 2025 earnings tomorrow. In the continuing major label beauty contest, this is Kyncl laying out his pitch before big brother Lucian gets his turn on stage.
The memo is topped with an artfully posed, moodily lit headshot - open-collared shirt over black tee, kind eyes, a hint of a Mona Lisa smile. What a handsome bugger. He looks like a man who knows something you don’t. Which, really, sets the tone for the 2600-odd words that follow.
With UMG’s share price circling its post-IPO low, CEO Lucian Grainge’s Streaming 2.0 vision gathering dust and the previously much-hyped superfan tiers nowhere to be seen, the field is wide open for someone to step forward as the man with the plan, and Kyncl wants your vote.
The plan, in his case, is AI. And like all good beauty contest contestants, he wants to save the kittens and cure cancer - AI will make artists richer, fans happier, shareholders wealthier and the music business bigger. Everyone wins. Nobody loses. The headshot says “Trust me: I’m the man who can”.
The memo does the usual things, wheeling out big numbers to wow the room - 5.1 trillion global streams, 800 million-plus subscribers, Goldman Sachs projecting 1.5 billion subs and $55 billion in revenues by 2035. He says the industry is “around 35% above 1999 peak”, which is true if you don’t adjust for inflation, and much less true if you do.
He says there’s “clearly more share of the wallet left for music”, because the average American household spends $14 a month on recorded music versus $69 on video streaming. Though, that $69 reflects a rather different market to the “all the music in one app” model we’re used to.
Video streaming is a series of walled gardens, where the value of each subscription is the content that they and only they have. People pay for Netflix and Disney+ and Max and Prime and Apple TV+ simultaneously because each has exclusive shows and films the others don’t. That’s what pushes the number up. Music streaming is one subscription for everything - there’s no exclusive content to chase. Worth remembering that comparison, though, because it comes back later.
But the big numbers are the warm-up act. The main event is AI.
AI as the growth play
Kyncl opens his pitch for the future with a claim that sets up everything else. AI, he says, will “enable fans to more deeply interact with the music they love, by reimagining it rather than just listening”. When the company went public in 2020, Warner’s model was built on consumption. “Today, it is based on consumption and creation”.
The pitch goes like this. Fans spend about $0.25 per hour listening to recorded music, $0.50 per hour on video, and $5 per hour gaming. The more interactive the medium, the higher the spend. Suno has two million subscribers paying $12.50 a month to create music with AI, on top of whatever they pay for streaming. That’s new money. Warner has deals with Suno, Udio, Stability AI and Klay, all structured on variable economics - Warner grows as they grow. AI is not a threat. It is a new revenue layer.
None of this is a surprise. Warner has been talking up its AI licensing deals for months, and the “new revenue on top of streaming” line has been part of the pitch since at least last year’s Q4 earnings call. But the closer you look at the evidence Kyncl uses to support it, the shakier it gets.
First, the money might not be new. Last week, Suno’s lead investor at Menlo Ventures, CC Gong, said in a post (since deleted) that she had “personally shifted most of my listening to Suno” because she was “so tired of Spotify giving me the same overplayed recommendations”.
If people are making music instead of streaming it, that’s not a new revenue layer. That’s potentially the same money moving somewhere else, probably with a smaller cut going to rightsholders. Gong’s post was deleted, presumably because it directly undermines Suno’s fair use defence in its ongoing copyright litigation, because ‘market dilution’ is one of the factors that weighs against fair use. It also potentially undermines the entire foundation of Kyncl’s AI argument.
Second, it’s not clear that people actually want this. Kyncl’s pitch depends on the idea that fans are itching to “reimagine” the music they love. The data on how people actually use streaming services suggests otherwise.
Catalogue music - tracks more than eighteen months old - accounts for roughly 73% of total US album consumption according to Luminate, and that share is growing. On-demand streaming of current music in the US actually declined by 1.6% in 2025. People are listening to more music, but it’s overwhelmingly old music, familiar music, playlist music. Most people want music to happen in the background while they do something else. They’re not looking to remix it.
Suno’s meteoric rise to $300 million in annual recurring revenue proves there is a market for AI music creation tools. But Suno’s own pitch deck shows only 25% of subscribers are still around after 30 days, and the conversion rate from its 100 million-plus total users to paying subscribers is about 2%. Whether even that lasts or whether Suno is riding a novelty that’s about to peak is an open question - and Kyncl is betting the house on it.
Which brings us to the question of who, exactly, is supposed to be paying for all of this. And - tediously, predictably, yet a-bloody-gain - it’s going to be that much-hyped but somewhat elusive creature, the lesser-spotted superfan. Of course it is.
The superfan has left the building
The industry has been talking about superfans for years. Kyncl cites the standard data point - 20% of US listeners are superfans, they spend twice as much as everyone else.Grainge has been saying the same thing. Former Spotify CEO Daniel Ek talked about it constantly. And yet no major Western DSP has actually launched a superfan tier. Spotify’s long-rumoured ‘Music Pro’ has been delayed repeatedly, and by late 2025 Ek was telling analysts that existing subscriptions were “plenty enough for us”.
What Kyncl has done - and it’s quite clever - is quietly swap the delivery mechanism. The superfan opportunity used to be about lossless audio, exclusive content, early access, all the things DSPs have struggled to package into something anyone will pay for.
Now it’s about “the ability to play with songs in new ways”. Same promise, different wrapper. AI is the superfan play. Which aligns with what Spotify’s new co-CEO double act was as good as saying on their recent earnings call. Only in Spotify’s version, it remains to be seen how - whether - this would benefit rightsholders like Warner.
Which might explain why the most interesting section of the memo isn’t really about AI at all. It’s about deal structure.
The small print
Buried in the memo’s discussion of DSP renegotiations, between the headline AI claims and the beauty contest posturing, is a passage that deserves more attention than it will probably get.
Warner, says Kyncl, has “reached new wholesale terms with our partners that better reflect the ever-increasing value of music, while providing us with greater economic certainty and untethering us from where our partners choose to set subscription prices”. What that means is that Warner has locked in a floor - if Spotify bundles music into something cheaper or slashes prices, Warner doesn’t take the hit. The risk sits with the DSP.
Separately, he writes that they “see the division of the royalty pool using an artist-centric approach as a key and fair component of remuneration. The good news is that we’ve successfully established it together with our DSP partners; now we will focus on evolving it to match the changing landscape”. That last bit - “evolving it to match the changing landscape” - is worth lingering on. The changing landscape is AI - and evolving the royalty pool to match it could mean something very specific.
Read this alongside Warner’s AI streaming licensing deals and an interesting possibility opens up. If Warner has negotiated DSP agreements that exclude AI-generated content from the royalty pool - or that weight AI-generated streams differently under artist-centric models - then the real competitive advantage isn’t in AI licensing revenue. It’s in protecting the existing catalogue from dilution.
This is speculative, but the logic works. “Artist-centric” monetisation routes more money toward established, verified human artists and away from the long tail. That long tail is increasingly AI-generated.
Luminate’s data shows 96.2% of daily uploads to DSPs now come from independent and DIY distribution, with 106,000 new tracks arriving every day. 88% of the 253 million tracks on streaming platforms barely get played. If artist-centric models filter AI noise out of the royalty pool, major label catalogues benefit disproportionately. It’s not what’s in the deal. It’s what’s excluded.
It’s a smart defensive play. But it only works as long as the current DSP architecture holds - and as long as DSPs keep agreeing to terms that favour the majors over the flood.
The direct licensing play
There’s another thread in the memo that hasn’t had much attention. Kyncl talks about the major’s music publishing company Warner Chappell further pushing into direct licensing - pulling more digital rights out of collecting societies in more markets and licensing them directly to DSPs.
To date, music publishers directly licensing DSPs has mainly been restricted to Anglo-American repertoire, because of the way rights are assigned and managed in other markets. It has generally been felt that extending direct licensing to other repertoires can’t be done at scale, because it would require individual writers to withdraw rights from collecting societies, and requesting they do that would be very controversial.
AI potentially changes the pitch that Warner can make to songwriters.
While collecting societies have licensing relationships with streaming services, and some are now in talks about licensing for AI, pretty much every digital music platform has started by negotiating deals with major record labels, and those deals dictate how the business model looks moving forwards. With AI potentially remixing and reworking existing songs, songwriters need a bigger influence over the business model, and it’s the major publishers - working in sync with the major labels - that have that influence.
After all, Warner Chappell is sitting across the table from the AI platforms right now, shaping the product, negotiating the deals, and - if the artist-centric models work as advertised - carving out protections that don’t yet exist in the collective licensing world.
The pitch to songwriters becomes simple. We can protect you from AI, or help you monetise in the age of AI, but either way you’ll need to pull your digital licensing rights out of your CMO and give them to us. We’ve got the insight into the deals, the data and the relationships. Nobody else is in the room.
Whether that’s a genuine service or a land grab dressed up as protection depends on your perspective. But it’s a much easier sell than it was three years ago.
Copyright, platform lock-in and walled gardens
On Monday - the day before the Warner AGM - the US Supreme Court declined to review the DABUS case, which means AI-generated works without human authorship still can’t get copyright protection in the US.
This matters for Kyncl’s vision of fans “reimagining” songs. If a user creates an AI-powered remix on Suno or a future Spotify integration, the copyright status of the output depends on what it is. If it’s a wholly new AI-generated track - something conjured from a prompt with no identifiable source material - there may be no copyright at all, which means anyone can copy it, distribute it, do what they like with it. Not much of a business model.
But if the output is a “reimagining” of an existing song - a remix, a reworking, a derivative - then copyright law already has a framework for that. Derivative works are protected. And whoever controls the rights to the derivative controls where it lives. That’s where the walled garden comes back into play.
If Spotify builds AI remix tools and your personalised version of a song is a derivative work created inside Spotify’s ecosystem, that’s platform-exclusive content. If that “reimagining” lives in Suno, it stays in Suno. The user can’t take their creation to Apple Music. The rightsholder gets a royalty on the underlying work. And the platform gets stickiness it didn’t have to pay for.
That’s where that $69 video comparison comes back. If AI derivatives are platform-locked, and fans start building libraries of personalised music they can’t take with them, you’re looking at exactly the kind of walled garden fragmentation that drives multi-subscription spending in video. Which is great if you’re trying to close that $14-versus-$69 gap. Less great if you’re an artist wondering why your fans are paying three platforms for AI-mangled versions of your songs instead of streaming the originals.
Skateboarding dogs and the visionary schtick
Kyncl positions himself throughout as someone who has seen this all before in his previous roles at Netflix and YouTube. Netflix was “deemed too dependent on the movie studios”. Google was “written off” by Fortune in 2010. YouTube was “mocked as peddling amateur content - think dogs on skateboards”. He was there for all of it, skating-boarding dogs and all, and the sceptics were wrong every time.
The only problem with that positioning is that Netflix, Google and YouTube were technology platforms disrupting incumbents. Warner is an incumbent trying to ride a disruptive technology. These are not the same thing.
There’s also a selectivity in what the memo covers. Kyncl references Deezer reporting 60,000 AI tracks uploaded daily and seven million tracks generated per day on Suno alone, but only as evidence of opportunity. He doesn’t mention that up to 85% of streams on AI-generated tracks on Deezer were fraudulent. And if the majors really have carved AI content out of their royalty pools, the flood doesn’t hurt them - it might even help, because the money inside the pool gets divided among fewer streams. The people who get hurt are the independents and smaller artists who don’t have the leverage to negotiate those exclusions.
But it’s not just the people outside Warner’s walls who should be paying attention. On efficiency, Kyncl cites a 380-basis-point margin improvement over five years and a 42% increase in OIBDA per employee since 2022. Last week, Block laid off 40% of its workforce - more than 4000 people - explicitly citing AI, and the market rewarded it with a 20%-plus share price surge.
Warner has roughly 4500 employees. Kyncl describes AI as “an accelerant” for “increasing efficiency” but does not say what that means for headcount going forward. But if I was working in a role at Warner Music where there was even a shadow of a hint that it could be automated or replaced by AI, I’d be reskilling and looking for a new job, yesterday.
Bob’s big bet
Kyncl is making a big, public bet that AI will expand the music business rather than eat it. His AI deals are structured so Warner grows as its partners grow. His DSP deals appear to be structured to protect the value of human-created catalogues. His compensation, amended in November 2025, ties stock options to total shareholder return thresholds of 8%, 10% and 12%. His conviction and his pay packet are pointing in the same direction.
We’ve seen how that played out for Lucian Grainge, whose share price at Universal Music is barely a whisker above its reference point for the IPO four and a half years ago, and whose bonus threshold triggers of €30 and €38 per share are so far out of reach that it seems unlikely - barring a miracle - he’ll ever see his megabonus payday, having hitched up his skirts and tied his garter so tightly around the maypole of Streaming 2.0, superfans and super premium.
But there’s a question the memo doesn’t ask, and it’s the one that matters most. Kyncl’s entire thesis assumes the current landscape holds - that Spotify and its fellow DSPs remain the dominant consumption endpoints, that the major labels remain the dominant suppliers, and that AI slots neatly into the existing architecture as a new revenue layer on top.
Spotify, meanwhile, has shown no interest in building the superfan tiers that the industry has been begging for. It has shown plenty of interest in AI. New co-CEO Gustav Söderström spent Christmas coding with Anthropic’s Claude and told analysts his best engineers haven’t written a line of code since December.
Spotify is building a language-to-music dataset that works just as well for generating music as it does for recommending it. If Spotify decides that AI-generated content is cheaper to serve and keeps subscribers just as happy, the “new revenue layer” Kyncl is banking on might turn out to be Spotify’s margin play, not Warner’s.
And that’s if Spotify remains the destination at all. A few years ago, nobody could imagine a world where Facebook wasn’t the default social network, or a day when you didn’t check Twitter to stay on top of the news. Both assumptions turned out to be wrong.
Is Spotify going to be the place where Gen Z and Gen Alpha go to consume AI-mangled versions of the artists they love? Or does a generation raised on Suno and TikTok and whatever comes next build entirely different habits - ones where the carefully orchestrated landscape of compliant streaming services and major label dominance looks as quaint as the CD rack at HMV?
Kyncl’s memo is every bit as confident, polished and good looking to investors as his artfully staged-and-lit headshot. The question is whether the music industry’s biggest AI optimist is building for a future that might not include the platforms he’s building on.
