Suno doubled its valuation to $5.4B with a $400M raise and announced its first fully-licensed model, record labels piled 61,000 more songs onto their copyright complaint, and GEMA's landmark European ruling got pushed to late July.
Suno raises $400M at $5.4B valuation — AI music's biggest bet yet
Suno closed a $400 million Series D on Tuesday, more than doubling its valuation from $2.45B just seven months ago. Bond Capital led, with IVP, Forerunner, Union Square Ventures, Alkeon, and Quiet joining alongside existing backers. The company now has 2 million paid subscribers and $300M in annual recurring revenue.
The scale is worth pausing on. Suno is generating more revenue than most mid-size record labels, and investors are valuing it like a platform company, not a tool. Whether that bet pays off depends entirely on how the licensing story plays out — which brings us to the next two items.
Suno's first fully-licensed model will replace the current one entirely
Buried in the funding announcement: Suno confirmed it will roll out its first AI music model trained exclusively on licensed repertoire "in the coming months," built in collaboration with music industry partners. More significantly, the existing model — the one trained on disputed data — will be discontinued, and all users will be migrated to the new licensed system.
If you're using Suno in production, this matters. The current model's sonic character, its strengths in certain genres, its quirks — all of that changes when the training data changes. Expect a transition period where outputs sound different. Whether "different" means "better" is an open question. The Warner deal from November 2025 provides the template, but the catalog available for training will be limited to labels that have opted in.
Record labels add 61,000+ songs to Suno copyright complaint
Universal Music Group, Sony, and other major labels moved to amend their complaint against Suno in the U.S. District Court for Massachusetts, alleging that more than 61,000 additional songs were used for training without authorization. A key summary-judgment hearing is now scheduled for July 2026. Warner Music Group settled and is no longer part of the litigation, but Universal and Sony are clearly escalating.
The timing is calculated — filing this the same week Suno announced its $400M raise signals that the labels intend to make the legal risk impossible for investors to ignore. For practitioners, the practical implication remains the same: if you're shipping commercial work made with the current Suno model, understand that the legal status of its training data is actively contested.
GEMA v. Suno verdict pushed from June 12 to July 31
The Munich Regional Court delayed its ruling in GEMA's copyright case against Suno from June 12 to July 31, citing internal administrative reasons. The case centers on six compositions — including "Atemlos," "Daddy Cool," "Rasputin," and "Mambo No. 5" — and whether Suno needed authorization to train on them. A GEMA win would be the first major European court decision confirming that AI platforms need rights clearance for music training data.
The delay is frustrating for anyone waiting on legal clarity, but the case itself remains significant. European copyright law differs substantially from U.S. fair-use doctrine, and a ruling here could set a precedent that travels across the EU before the American cases even reach trial.
Spotify's ElevenLabs-powered audiobook creator enters invite-only beta
Spotify's AI audiobook creation tool, announced May 21 with ElevenLabs as the voice engine, is entering invite-only beta this month through the Spotify for Authors platform. Authors upload manuscripts and generate full narrations using ElevenLabs' synthesis models — no studio, no narrator required. All AI-narrated titles get a "narrated by a digital voice" label, and the tool is non-exclusive, so authors can distribute elsewhere.
This is the clearest signal yet that ElevenLabs is evolving from a voice-cloning novelty into core infrastructure for audio platforms. For audio professionals, the more interesting question is what happens when this quality of synthesis becomes available for music vocals, sound design narration, and interactive media — not just audiobooks.
A2IM Indie Week to tackle AI policy for independent labels
A2IM Indie Week runs June 8–11 in New York and is expected to be the most significant independent music gathering this year for AI policy language. With majors cutting bilateral deals — UMG with Udio, Warner with Suno, UMG with Spotify — independent labels and distributors need their own framework for how AI training, licensing, and compensation should work for catalogs outside the Big Three.
If you're an independent artist or label, watch what comes out of this event. The deals being struck right now will define who gets paid when AI models train on or generate music, and independents risk being left out of a system designed by and for majors.
LALAL.AI shipped its first native DAW plugin — offline AI stem separation that runs entirely on-device via their Lyra model. Insert it on a track, pick which stem you want (vocals, drums, bass, acoustic guitar, piano, electric guitar, or full instrumental), and it renders the separation locally with GPU/NPU acceleration. No cloud upload, no per-track limits, no length caps. It works in any VST3-compatible DAW, with AU in beta.
This is meaningful because it removes the workflow friction that made cloud-based stem separation tools annoying for session work. You stay in your DAW, the audio stays on your machine, and processing takes seconds rather than upload-wait-download cycles. The separation quality sits in the same tier as Demucs and RipX — not perfect, but good enough for remix work, sample extraction, and production reference.
Suno's announcement that its current model will be fully discontinued and replaced by a licensed version is the clearest marker yet of a permanent split forming in AI music. On one side: models trained on licensed catalogs, with artist opt-in, revenue sharing, and legal clarity. On the other: open-source and research models trained on whatever data was available before the lawsuits hit. The sonic characteristics of each lane will diverge — licensed models constrained by available catalog, unlicensed models frozen in time as legal risk makes retraining untenable.
For working audio professionals, this means the tools you rely on today may not sound the same in six months. If you've built workflows around the current Suno or Udio output, start testing alternatives now. And pay attention to which catalogs are actually included in the licensed models — a model trained only on Warner's catalog will have very different strengths than one with access to all three majors plus independents. The quality of the licensed era depends entirely on the breadth of the licenses.