Spotify AI Pivot Is Becoming an Agent Platform Built on Artist Labor
Spotify's gross margin climbed from 25 percent to 32 percent in three years — and the company says AI is why. Engineers stopped writing code in December 2025, using an internal AI coding system called Honk to handle implementation instead. By the company's own count, that one system merged more than 1,500 pull requests. Spotify is now targeting 40 percent gross margin by 2030, built on the premise that human implementation work can be replaced by AI at scale. The UMG remix deal announced this week is the same bet applied to artists: a paid add-on that lets subscribers generate derivatives of licensed tracks, with no disclosed royalty formula and no answer to who gets paid when a machine remixes a human musician's work.
Spotify and Universal Music Group signed a deal this week that lets subscribers use AI to remix songs from participating artists — and neither company has disclosed how the revenue will be split with the artists whose work is being used. The AI remix product is described as a paid add-on, but no specific price, royalty formula, or artist consent mechanism has been announced. Royalty structures for AI-generated derivative works are not standardized anywhere in the industry. No other major label has a comparable public agreement. Spotify shares rose 16 percent on the announcement anyway.
The financial case Spotify presented at its Investor Day is grounded in numbers the company published that day. Co-CEO Gustav Söderström laid out a new strategic thesis: access to a catalog came first, then personalization through curation, and now Spotify is entering "the era of generation," where AI creates music in real time around each user's taste, context, and intent. The company reported EUR17 billion in 2025 revenue with a gross margin of 32 percent, up from 25 percent in 2022. It is targeting EUR100 billion in annual revenue by 2030, with gross margins between 35 and 40 percent and operating margin above 20 percent. It has 761 million monthly active users and 293 million paid subscribers. The 2030 targets imply the AI features are expected to drive meaningful margin expansion — which makes the unresolved UMG royalty question more consequential, not less.
The labor shift predates this week's announcements. In December 2025, Spotify's best engineers stopped writing code entirely, using an internal AI coding system called Honk — built on Anthropic's Claude Code — to handle implementation instead. According to Spotify's engineering blog, the transition was visible within weeks: engineers returned from holiday to find every terminal in use, no IDEs. The company says it shipped more than 50 new features and changes in 2025, most of them AI-assisted. Spotify shares jumped an additional 13 percent on the combined Investor Day announcements, adding to the prior day's gain.
That same day the company announced Studio, a standalone desktop app that connects to a user's email, calendar, and notes, browses the web, and completes tasks on their behalf. It also unveiled Personal Podcasts, which generates audio briefings from the same data sources. Neither product is a music feature. Both are positioned as daily productivity tools. If Studio gains traction, Spotify enters direct competition with Microsoft's Copilot, Google's Gemini, and OpenAI's agent products — not with Apple Music or YouTube Music. That is a harder competitive surface than streaming audio, populated by companies with deeper enterprise distribution and more established AI infrastructure.
For emerging artists, the UMG remix deal adds a more immediate problem. When listeners can generate AI derivatives of catalog tracks — remixes, covers, variations — on demand, the economic value of a new artist's original work faces pressure from infinite AI-generated variations of already-popular songs. An emerging artist's single, already competing against algorithmically curated catalog, now also competes against listener-generated derivatives of tracks from artists with massive existing catalog presence. The royalty structure for those derivatives has not been defined, which means the artists with the most catalog also have the most to gain from an AI remix ecosystem — and emerging artists have the least.
Personal Podcasts will roll out to eligible Premium users in the U.S. next month, with monthly credits included and the option to purchase more. More than 500 million users have already streamed a video podcast on Spotify, a figure up nearly 50 percent year over year.
What the UMG deal confirms is that Spotify believes AI-generated derivatives of copyrighted music can become a paid product — and that the market will accept it without a public accounting of who gets paid. The company's 2030 targets require that bet to pay off in margin expansion. Artists whose catalogs make the service possible were not mentioned in the deal announcement. Whether they have a claim on the revenue the AI remix add-on generates is a question neither Spotify nor UMG has answered publicly.
What to watch: Spotify has not disclosed when the AI remix add-on will launch publicly or what specific artist consent mechanism will govern which tracks can be remixed. The next signal will be whether UMG's publishing arm or other major labels publish their own terms — or whether the industry defers to Spotify to set the de facto standard.