A hundred million people have now made music with Suno. Let that sit for a second. The biggest names in music software have spent decades building their audiences, and a single AI tool blew past most of them in under two years.
My honest reaction is that it’s both impressive and a bit boring at the same time, and I don’t think those two things cancel each other out. Here’s how I see it.
1. Suno is undoubtedly doing ‘well’
Suno reported around 100 million people have used the platform, with 2 million paying subscribers and roughly $300 million in annual revenue as of early 2026. Numbers like that don’t come from existing producers switching tools. They come from people who would never have opened a DAW in their life, typing a sentence and hearing a song come back.
(For reference, some of the largest music production companies have told me directly that they’ve built up audiences of around 11 million, over decades, but that includes dormant customers from 10+ years ago).
I think that’s a good thing. Most of those people aren’t trying to become recording artists, they’re just playing around and being curious. And curiosity is how almost every producer I know got started, usually with a cracked copy of something and no clue what they were doing. If AI music is the thing that gets a new generation curious, with the lowest barrier to entry music has ever had, then I hope it’s a gateway. The thing that makes someone think, maybe for the first time, that they could make this too.
So I’m not in the camp that wants to pull the ladder up. More people making sound is more people falling in love with music. That part I’m glad about.
The reach is enormous. The listening is not.
Here’s the part that gets lost in the hype, though. Deezer now says 44% of everything uploaded to its platform every day is AI-generated, around 75,000 tracks. And yet AI music accounts for only 1 to 3% of actual streams, with about 85% of those streams flagged as fraudulent and demonetised. Nearly half the supply. Almost none of the demand.
People are generating music at a furious pace. But it’s not getting listened to.
2. Easy isn’t the same as satisfying
Typing a prompt and getting a finished song is impressive. It’s also a bit hollow after the third or fourth go, because you didn’t build anything or make a single decision that was yours. You described a vending machine order and a machine handed you a snack.
Producers love being in the driver’s seat. The reward in music production isn’t only the finished track, it’s the thousand small choices along the way: the EQ move at 250 Hz that suddenly makes the vocal sit, the arrangement decision at 2 a.m. that changes the whole feeling of a song. Take those choices away and you’ve taken away the craft. What’s left converts fast but doesn’t teach you anything and doesn’t feel like yours.
I use a simple analogy for this. Just because the Olympics exist, people don’t stop going jogging. The existence of the best in the world doing it at the highest level has never stopped ordinary people doing it for the love of it. If anything, watching the best makes more people want to give it a go. AI being good at making music, even very good, doesn’t remove the reason a human picks up the work in the first place.
AI changes a lot about the inputs. It changes nothing about why we do this.
3. It’s better than beginners. It’s nowhere near the pros.
I’ll be fair to the technology here, because pretending it’s bad helps no one. On the pure technical side, and even on some songwriting, AI music has already gone past a lot of beginners and a fair few intermediate producers. That’s just true.
But it still isn’t close to a great human songwriter or a great mix engineer. And it still has that sound. You know the one. A slightly glassy, slightly smeared, weirdly confident sound that makes a track easy to spot as fully AI-made. Ok but i’m am mastering engineer and a plugin developer who listens to music and knows audio inside out… what about normal people?
Well, a Deezer and Ipsos study found that 97% of listeners couldn’t tell AI from human music in a blind test, which sounds damning until you sit with it. People can’t always pick it out of a lineup, sure. That doesn’t mean they prefer it, or feel anything for it, or would choose it once they know.
They might fix that strange sound eventually. I assume at some point they will. But solving the texture doesn’t solve the deeper thing, which is that people care where their music comes from once you tell them.
4. Look at what happened to diamonds
There’s a useful parallel sitting right next to us in another industry. Lab-grown diamonds are chemically identical to mined ones. Same material, same sparkle, and in 2026 they cost roughly 60 to 85% less. They’ve taken something like half of all engagement ring sales. By every rational measure they should have ended natural diamonds.
Except they didn’t. The market split instead. Lab-grown became the everyday, affordable option, and natural diamonds got repositioned as the premium, the real thing, the one people pay a lot more for on purpose. Plenty of buyers can’t tell the two apart in a glass case, and they still reach for the natural stone, just because of where it came from.
Cheaper and more available did not mean more valued. Music is heading the same way.
Music is splitting along that exact line. Cheap, infinite, machine-made stuff on one side, and human-made work as the thing people actually put value on, on the other. We’re in the loud part of the cycle right now, where the hype says it’ll all be AI and nothing else will matter. But hype is hype, and it always burns brightest right before people decide what they actually want to keep.
Cheaper and infinite has never been the same as ‘valuable’.
5. People are already voting with their ears
This isn’t a prediction I’m hoping comes true. It’s already moving. Deezer started labelling AI tracks at the platform level in 2025, the first major service to do it, and tagged over 13 million of them in a year. Spotify pulled tens of millions of AI spam tracks in a fraud crackdown. Bandcamp banned AI-generated music outright. In that same Deezer study, 80% of people said fully AI-generated music should be clearly labelled, and 52% said it shouldn’t sit in the main charts next to human-made songs.
People want it labelled and they want it kept separate. A lot of them, honestly, find the idea of fully machine-made music a bit icky, and they’re saying so before anyone’s forced it on them. That’s the demand side talking, and it’s telling you it values the human hand.
6. Why we’ve kept AI out of our products
This is the part that’s personal to how we build. My plugin company, Mastering The Mix, hasn’t put generative AI into anything we make, and that’s a deliberate choice, not a gap in the roadmap. We’ve leaned hard into smart algorithms (the kind that do real, useful analysis and heavy lifting) while keeping the producer in the driver’s seat the entire time.
There’s a real difference between a smart tool and a generative one. A smart tool shows you what’s happening and lets you make the call. A generative one makes the call for you and hands you the result. The first kind makes you a better producer over time. The second kind quietly cuts you out of your own track. Everything we ship is built the first way, so the human is still sculpting the sound, still deciding, still putting their taste and their vision into the work.
I’m not doing this to take some moral stand. I just think human-led music is what lasts. Keep things producer-led and organic and you keep the one thing AI can’t fake at scale, which is the variety and the uniqueness you get from millions of people making different choices for their own reasons. Hand all of that to a model and music drifts towards the same confident, glassy average. Keep it human and it stays different, in a good way.
Keep the producer in the driver’s seat, and the music stays unmistakably theirs.
Where I land on all of this
AI music is going to bring millions of new people to sound, and I’m glad it will. It’s going to flood the pipes with cheap tracks almost nobody listens to, and we’ll all learn to filter past that. It’ll keep getting better, and it still won’t replace the reason a person sits down to make something that’s theirs.
The pattern’s already there if you look past the noise. Cheap and infinite on one side, human and valued on the other, and people coming back to the real thing like they always do. My job, and the job of anyone building tools for producers, is to make sure the craft is still here when they come back, and that the tools still treat them like the artist rather than the bottleneck.
So make the music yourself. Not because a machine can’t, but because you can, and that’s really the whole point.
Written by Tom Frampton.
Tom Frampton is the founder of Mastering The Mix, where he builds audio software used by producers and engineers around the world. He started the company to make pro-level mixing and mastering decisions easier to get right, without taking the producer out of the driver’s seat.
Sources
- Suno user / subscriber / revenue figures: Music Business Worldwide and TechCrunch, Feb–Apr 2026.
- AI share of uploads vs streams, and platform labelling: Deezer Newsroom, Apr 2026; TechCrunch, Apr 2026.
- Listener attitudes (97% blind-test, 80% want labels, 52% chart separation): Deezer + Ipsos study, Nov 2025.
- Spotify AI spam removal and Bandcamp ban: Digital Music News / DJ Mag, 2025–2026.
- Lab-grown vs natural diamond pricing and market share: BriteCo / Stacker 2026, The Knot, Paul Zimnisky, Rapaport Jan 2026.
Source link
