Spotify is aiming to launch its new AI-powered remix and covers service in early 2027, according to a report in the Financial Times, although its sources concede that an exact timeline is yet to be finalised, and it’s not yet been decided how much to charge users to access the new add-on service.  

The music industry, and especially Universal Music, is very excited about the commercial potential of using AI to remix and cover existing tracks and songs, rather than using AI to generate new music. However, it’s not yet clear that artists and fans share that excitement. 

Meanwhile Spotify itself has a long history of promising a higher-priced tier with extra functionality to boost payments to the music industry, and then failing to deliver on that promise. 

While Universal, Merlin and Believe have all signed up to participate in Spotify’s AI remixes gimmick, the labels and distributors will need to get artists to opt-in, especially for the voice swapping element, where users can ask the AI to generate a cover version with vocals that imitate a specific artist. 

Universal has been reaching out to some artists – sometimes via management, sometimes via their lawyers – to discuss its various AI deals, though some sources say they are so far struggling to convince many acts to get involved. The FT cites one insider as saying “winning over major contemporary stars has proved more challenging than securing participation from some established older artists”.

However, when asked about that challenge on an investor call yesterday, Universal’s Chief Digital Officer Michael Nash – somewhat testily – tried to put a positive spin on it all, insisting, “our engagement with the creative community is going extremely well”.

“We’ve engaged with thousands of our artists talking about this product model”, he went on, adding, “we’ve already converted a significant percentage of those discussions into opt-in agreements, and we continue to make excellent progress”. 

For artists and songwriters, the various AI deals that have been announced by record labels, music publishers and music distributors pose a lot of questions. 

When it comes to licensing basic AI training – which often involves copying large quantities of music – the majors seem to think they don’t need to get the permission of individual artists and songwriters providing they share any licensing income it generates. Music creator groups around the world do not agree. 

But where existing songs and recordings, and especially an artist’s voice or brand, is used on the output – as with Spotify’s AI tools – artist and songwriter permission is required.

That’s something Universal boss Lucian Grange confirmed during yesterday’s investor call. He explained, while battling bravely through a stonking cold, that “artists and songwriters opt in whenever their voice, image, name and likeness, lyrics or compositions are used, and the creative community shares fairly in the value that’s created”. 

Some artists simply won’t want to take part. For others, it will depend on the specifics of the deal. And the major labels have a long history of being very vague when it comes to their often complex digital licensing deals, and with AI remixes things are even more complicated. 

If Spotify is sharing the extra money it makes selling its AI tool with the music industry, how does that work? As in, how is it split between the music used to train the model, the recordings and songs being covered and remixed, and the artists whose brand is being exploited and voice is being imitated? 

Is Spotify taking its usual 30-35%, or does it want more to pay for operating its AI model? And how does Universal propose to share the money with its artists? 

Sources say that both streaming royalty rates (usually a minority share for the artist) and a 50/50 split have been proposed in different conversations about the major’s AI deals. But when it comes to artist brand and voice, arguably labels are simply acting as agents negotiating a deal around rights they don’t own or control, meaning artists might want a much bigger share of the money. 

Obviously we don’t know how many remix-able tracks and clone-able artists Spotify reckons it needs to launch its new service. On an investor call earlier this year Co-CEO Gustav Söderström said “we would like to have as many artists as possible, obviously, but we don’t expect to have all artists”. 

That said, if the service launches with a small selection of artists and tracks – and if it is predominantly heritage acts – that would likely make the service much less attractive to the average user. And Spotify presumably wants to launch with the best possible offering. 

Plus, of course, even if a decent number of artists come on board, the song rights in the works being remixed or covered also need to be fully licensed. Even if a decent number of publishers sign up, that doesn’t mean their writers will opt-in, and co-ownership of song rights always makes licensing the publishing rights complicated. 

And making it even more complicated, the songwriter collecting societies control at least some of the rights that Spotify needs to exploit by allowing its AI to remix and cover existing songs, so they need to be involved in the licensing and revenue sharing too. 

It may be that offering upfront advance payments can sufficiently sweeten the deal to get a decent number of artists and songwriters onboard. But whoever fronts those payments – Spotify or its industry partners – is taking a risk, given we don’t really know how many people will pay the additional $4-5 that Spotify is reportedly planning on charging for its AI remix product. 

Some are optimistic about potential uptake. The FT cites Ian Moore from Bernstein – a global equity research and brokerage firm – estimating that 7-8% of Spotify’s US subscribers might sign up for the AI gimmick by the end of 2028. 

Meanwhile on the Universal investor call, Grainge said of the AI opportunity, “we are moving from principles to products, and from products to revenue” and “we expect these partnerships to begin generating incremental revenue” in 2027. 

Given under the rules being set by the majors, any AI-generated remixes and covers will only be available within the platform where the AI did the remixing or covering, Spotify has a massive advantage over start-ups building similar products like Udio. 

Because if users being able to share any AI mixes they have prompted is part of the experience, there are a lot more people to share the mixes with within the Spotify ecosystem. Although, if sharing the outputs is central to the product, you have to think AI remix tools within TikTok, YouTube or Instagram have the most potential. 

We won’t really know how big the commercial potential is for AI-powered remixes and covers until someone – probably Spotify or Udio – gets enough rightsholders and artists on board to go live with such a service. 

Many people within the industry are hopeful that a decent number of music fans will be interested in playing around with these tools as they go live, but will the fans pay for those tools, and will that interest be long-lived or are we dealing with a gimmick here? 

If it turns out it’s a short-lived gimmick, that will be a problem for those music companies, such as Universal, that have been pushing digital platforms and AI start-ups go down this route. Even Suno has agreed to add AI remixing to its platform as it’s negotiated licensing deals with Warner Music and others, although you sense that’s only to placate the music industry and its real focus is outputting generic tracks. 

Suno insisting that generating generic music remains core to its offering is reportedly one reason why Universal has so far declined to do a licensing deal with the AI start-up, instead ploughing on with its copyright infringement lawsuit. But if Spotify and Udio’s AI remix products turn out something of a damp squib, Suno – for all of its YouTube scraping and AI slop generation – will have been kind of vindicated. 

​