AI Music Market: Deep Research
An analytical memo on the growth, competitive dynamics, and regulatory landscape of AI-generated music.
1. Executive Summary
AI music generation has moved from research demo to industry-scale supply shock in less than three years. On the demand side, well-capitalized generative platforms are compounding at rates unprecedented in consumer software: Suno now sits at a USD 2.45B valuation on USD 300M ARR as of Q1 20267,8 and ElevenLabs at a USD 11B valuation on USD 330M+ ARR as of February 20269. On the supply side, Deezer disclosed in April 2026 that 44% of daily uploads to its platform, roughly 75,000 tracks per day, are now fully AI-generated, up from ~10,000 per day when its detection tool launched in January 202516.
Analyst point estimates for the pure-play "generative AI in music" software market remain modest. Grand View Research pegs it at USD 569.7M in 2024 rising to USD 2.79B by 2030 at a 30.5% CAGR1, but industry-body forecasts are an order of magnitude larger. The canonical CISAC/PMP Strategy study of December 2024 projects the generative AI content market in music and audiovisual will grow from EUR 3B in 2024 to EUR 64B by 2028, with Gen AI service-provider revenues rising from EUR 0.3B to EUR 9B over the same window4,5. The gap between "software TAM" and "content-market TAM" is the central strategic tension of the sector.
That growth arrives against a hardening legal and regulatory backdrop. The RIAA-coordinated June 2024 copyright lawsuits against Suno (D. Mass. 1:24-cv-11611) and Udio (SDNY 1:24-cv-04777) are moving through discovery, with UMG and Sony moving in May 2026 to expand the Suno complaint from 560 to 61,026 recordings, pushing theoretical statutory damages from ~USD 84M to over USD 9B11,13. Warner settled with Suno in November 2025, and Universal settled with Udio in October 2025 on terms including a walled-garden "licensed" platform, the first commercial template for AI music12.
Streaming platforms are moving from passive host to active gatekeeper. Spotify removed over 75 million "spammy" tracks in the 12 months ending September 2025, tightened its impersonation policy to ban unauthorized vocal deepfakes, and in May 2026 announced a joint AI-remix product with Universal Music Group behind a paid tier18,19. Deezer demonetizes 85% of streams to AI-flagged tracks17. This memo argues that the sector's trajectory over the next 24-36 months will be shaped less by model quality than by three variables: (1) the terms of major-label licensing settlements, (2) platform-side content policies and detection tooling, and (3) the outcome of the RIAA-led fair-use litigation. Firms that solve for licensing early will command a durable structural advantage.
2. Market Size and Growth
2.1 The Analyst Consensus Range
Third-party market-research firms give strikingly divergent baselines for the AI-in-music market, largely because they draw the boundary in different places. Grand View Research sizes the narrow "generative AI in music" software segment at USD 440.0M in 2023, USD 569.7M in 2024, and USD 2,794.7M by 2030, a 30.5% CAGR through 20301. Market.us, by contrast, uses a broader "AI in music" boundary that includes AI-assisted production, mastering, recommendation, and voice tools, arriving at USD 5.20B in 2024, USD 6.65B in 2025, and USD 60.44B by 2034 at a 27.8% CAGR3. Both point estimates should be read as directional; the sector has moved so fast that any single-source TAM is stale within two quarters.
Goldman Sachs' updated "Music in the Air" report (3 May 2024) does not publish an AI-specific TAM but names AI as one of three structural drivers of its global music-industry forecast, alongside streaming price increases and royalty modernization, flagging AI as both an incremental revenue lever and a potential source of royalty-pool dilution2. The bank's treatment is instructive: it treats AI music as a lens on the existing USD 100B+ recorded and publishing pool rather than as a discrete new market.
2.2 The Industry-Body Forecast: CISAC / PMP Strategy
The most cited forward-looking figures in the sector come from the December 2024 study commissioned by CISAC (the International Confederation of Societies of Authors and Composers) and produced by PMP Strategy. Its headline numbers, drawn directly from the primary study, are as follows4,5:
Gen AI content market (music + audiovisual): EUR 3B (2024) → EUR 64B (2028); implied CAGR ~115%.
Gen AI service-provider revenues (music + audiovisual): EUR 0.3B (2024) → EUR 9B (2028); the study frames this as "doubling each year" (actual arithmetic implies ~135% CAGR).
Music-only Gen AI outputs: EUR 0.5B (2023) → EUR 16B (2028), with the 2028 mix split EUR 10B replacing human music inside traditional streaming, EUR 5B in B2B use, and EUR 1B on Gen-AI-native streaming platforms.
Music-only Gen AI service-provider revenues: EUR 0.1B (2023) → EUR 4B (2028), split roughly evenly between assistive creative software and mass-market complete-output tools.
Creator revenue at risk: Up to 24% of music and audiovisual creators' revenues, or a cumulative EUR 22B over 2023-2028 (EUR 10B music-only, EUR 12B audiovisual). Digital music streaming faces the steepest 30% cannibalization; radio/TV/live/background 22%; CD/video 21%.
Because CISAC is a creator-rights trade body, the study has an editorial interest in emphasizing creator harm; its methodology (~50 industry interviews plus quantitative modelling) is disclosed but not independently audited6. The figures are nonetheless universally cited in trade press: Music Business Worldwide, Music Ally, PRS For Music, APRA AMCOS, GEMA. At time of writing they are, the only comprehensive forward-looking dataset in circulation. For an investor memo they are best read as an upper-bound scenario reflecting the incumbent rights-holder view.
2.3 Where the Money Actually Sits
The CISAC breakdown makes clear that "AI music market" is a compound of three functionally different revenue pools:
Consumer generation tools (Suno, Udio, ElevenLabs Music). Subscription and credit revenue paid directly by end-users. This is the pool where Suno booked ~USD 300M ARR by February 20268 and ElevenLabs booked USD 330M+ ARR (across its full audio stack) by end-20259. Highest growth, highest legal risk.
B2B licensing, stock, and sync. AI music sold into ads, game studios, YouTubers, retail background music, podcast production. The CISAC study projects EUR 5B of the EUR 16B 2028 pool sits here, a channel where AI music is a direct substitute for royalty-free stock catalogs (Epidemic Sound, Artlist) and library music. Margins are thinner but litigation exposure is more contained.
Streaming and consumption revenue. Revenue collected by streaming platforms on plays of AI-generated tracks (whether user-uploaded, synthetic-artist accounts, or platform-native AI content). CISAC projects ~EUR 10B of 2028 music streaming will be AI content substituting for human, a direct dilution of the royalty pool paid to human artists.
3. Competitive Landscape
The AI music generation stack has consolidated faster than most consumer software categories. As of mid-2026, three tiers are visible: (1) a small group of well-capitalized independents with proprietary foundation models; (2) hyperscaler and platform-native offerings that leverage distribution rather than model IP; and (3) a long tail of vertical, legacy, or niche tools competing on workflow specialization.
3.1 Tier 1: Best-Capitalized Independents
Suno. The category leader by revenue and reach. Series B raised USD 125M in May 2024 at ~USD 500M valuation; Series C in November 2025 raised USD 250M at USD 2.45B post-money on USD 200M ARR7. By February 2026 Suno reported 2 million paid subscribers and USD 300M ARR, with over 100 million total users and ~7 million songs generated per day, equivalent to Spotify's entire historical catalog every two weeks8. Consumer pricing: Pro at USD 10/month (or USD 8 annualized) and Premier at USD 30/month (USD 24 annualized), plus a free tier8. Commercial "Creator" tier launched September 2025. Suno settled with Warner Music Group in November 2025 (multi-million-dollar payment plus a licensing partnership; Suno also acquired concert-data firm Songkick from WMG), but remains in active litigation with UMG, Sony, and European collecting societies Koda (Denmark) and GEMA (Germany)12,13.
Udio. Founded December 2023 as Uncharted Labs by five former Google DeepMind researchers; raised USD 8.5M seed (January 2024) and USD 10M Series A (April 2024) from Andreessen Horowitz, will.i.am, Common, and Instagram co-founder Mike Krieger10. Free tier initially offered 600 songs per month. Named as co-defendant in the RIAA-coordinated suits filed June 24, 2024. Udio settled with Universal Music Group on October 29, 2025 (dismissal November 4, 2025) with terms including compensation plus a commitment to launch a "walled-garden" platform trained exclusively on licensed music and offering opt-in artist compensation; existing users were given a 48-hour download window before transitioning to a streaming-based model12. Sony continues to litigate its separate Udio case (SDNY 1:24-cv-04777); on June 30, 2026 a New York judge denied Sony's motion to expand the case from 333 to 30,775 recordings13.
ElevenLabs. Founded 2022 as a text-to-speech company; expanded into music in August 2025 with "Eleven Music" (Merlin Network and Kobalt licensing deals), then Music v2 (May 2026) which supports genre-switching mid-track. Series D closed February 4, 2026: USD 500M at USD 11B valuation (3× the prior round), led by Sequoia, bringing total funding to USD 781M since 2022. Company-reported 2025 ARR: USD 330M+ (unaudited)9. ElevenLabs' strategic posture, building foundational models across the full audio stack (text-to-speech, transcription, music, dubbing, and conversational), positions it as the horizontal AI-audio platform rather than a text-to-song app.
3.2 Tier 2: Hyperscalers and Platform-Native
Google / YouTube (Lyria). Google DeepMind's Lyria model powers both the consumer-facing MusicFX (in Google Labs) and YouTube's Dream Track, the first major-label-licensed AI music experiment, launched November 16, 2023 with 30-second output limits and the licensed voices of nine artists: Alec Benjamin, Charlie Puth, Charli XCX, Demi Lovato, John Legend, Papoose, Sia, T-Pain, and Troye Sivan20. YouTube also runs the "Music AI Incubator" for generative experiments with artists and producers, and deploys SynthID watermarking on Lyria outputs.
Meta (MusicGen / AudioCraft). Open-sourced in mid-2023; primarily positioned as research infrastructure rather than a consumer product. Meta has not commercialized MusicGen as a standalone service but the model underpins numerous downstream tools.
Stability AI (Stable Audio). Stable Audio 2.0 launched April 2024 with the ability to generate up to three-minute tracks from text prompts. Stability commercializes via subscription and API; positioned as the open-model alternative for pro users.
ByteDance (Ripple / Seed). Ripple, a TikTok-adjacent AI music tool, allows creators to generate tracks and vocals for short-form video. ByteDance's Seed research team has published on audio generation and its integration into the TikTok creator toolkit is the largest embedded-in-a-social-platform distribution channel for AI music.
3.3 Tier 3: Vertical and Legacy Tools
A large second tier of specialized tools compete on workflow rather than model quality: Boomy (mass-market song generation with distribution to streaming), Soundraw (royalty-free AI stock music for creators), AIVA (classical/cinematic composition), Mubert (generative streaming and API for developers), Beatoven.ai (mood-based background music for video), LANDR (AI mastering, now bundled with distribution and sample libraries), and Splash Pro (AI vocals and beats aimed at hobbyists and gaming platforms like Roblox). Boomy attracted regulatory scrutiny in 2023 when Spotify removed a large share of its uploads over suspected streaming-fraud attributes, a preview of the platform-side enforcement pattern that would generalize across the sector by 2025.
Table 1. Top AI Music Platforms: Snapshot as of Q2 2026
Platform | Product Type | Launched | Total Funding / Latest Valuation | Reported User / Revenue Metric | Monetization | Notable Legal Status |
|---|
Suno | Text-to-song | 2023 | USD 375M+ / USD 2.45B (Nov 2025) | 100M+ users; 2M paid; USD 300M ARR (Feb 2026) | Freemium subscription (USD 8-24/mo) | Warner settled Nov 2025; UMG/Sony litigation, 61K-work expansion motion pending |
Udio | Text-to-song | 2024 | USD 18.5M disclosed | Not publicly disclosed | Freemium; transitioning to licensed streaming model | UMG settled Oct 2025; Sony case ongoing (SDNY 1:24-cv-04777) |
ElevenLabs | Full audio stack (TTS, music, dubbing) | 2022 | USD 781M / USD 11B (Feb 2026) | USD 330M+ 2025 ARR (self-reported) | API + subscription | Merlin, Kobalt licensing deals; no major-label suits |
Google / YouTube (Lyria, Dream Track) | Foundation model + creator tool | 2023 | N/A (in-house) | Not disclosed | Ad-supported / bundled with YouTube | Licensed cohort of 9 artists; SynthID watermarking |
Stability AI (Stable Audio) | Text-to-audio, open-model | 2023 | USD 100M+ raised prior to 2024 | Not disclosed | API + subscription | Multiple copyright suits vs. Stability (image side) |
Meta (MusicGen / AudioCraft) | Open-source model | 2023 | N/A (in-house) | N/A (not commercialized) | Research; no direct revenue | No specific music-side litigation |
ByteDance (Ripple) | In-app AI music for TikTok creators | 2023 | N/A (in-house) | Not disclosed | Bundled with TikTok | No public litigation |
Boomy | Mass-market song generation + distribution | 2018 | USD ~10M seed | "Millions of tracks" self-reported | Freemium + revenue share | Prior Spotify takedowns over streaming-fraud attributes |
Soundraw | Royalty-free AI stock for creators | 2020 | Not disclosed | Not disclosed | Subscription (USD 17-33/mo) | No known major suits |
AIVA | AI classical / cinematic composition | 2016 | Not disclosed | Not disclosed | Subscription tiers | No known major suits |
Mubert | Generative streaming + API | 2016 | Not disclosed | Not disclosed | API licensing + subscription | No known major suits |
LANDR | AI mastering + distribution | 2014 | USD 45M+ disclosed | 4M+ creators (self-reported) | Freemium + credits | No known major suits |
Sources: company disclosures and press releases; TechCrunch, MBW, Wikipedia public-record entries. Funding figures reflect publicly disclosed rounds only. ARR/user counts are company self-reported and unaudited unless noted.
4. Streaming Platforms: The Distribution Bottleneck
Streaming platforms occupy the pivotal position in the AI-music value chain. They are simultaneously the largest distribution channel for AI-generated tracks, the arbiter of what counts as "legitimate" music for payout purposes, and, through their partnerships with major labels, the venue through which licensed AI experiences will be commercialized. Their policies over the past 12 months mark a decisive shift from passive host to active gatekeeper.
4.1 Spotify
On September 25, 2025 Spotify disclosed that it had removed over 75 million "spammy" tracks in the prior 12 months, tying the enforcement wave explicitly to "the explosion of generative AI tools"18. The same announcement introduced three structural changes: (1) an updated impersonation policy that prohibits unauthorized AI vocal impersonation. Vocal cloning of a named artist is now permitted only with the artist's authorization; (2) participation in the DDEX industry-standard AI-credit disclosure schema, exposing AI-training and AI-generation flags in Song Credits (live in mobile from April 16, 2025); and (3) a new spam filter that de-recommends mass uploads, duplicates, SEO-hack tracks, and artificially short "abuse" tracks (a common streaming-fraud vector where tracks are cut to ~31 seconds to trigger a payable play)18.
The commercial pivot came May 21, 2026 at Spotify Investor Day, where the company announced an AI music licensing deal with Universal Music Group. The deal allows Premium subscribers to create AI-generated covers and remixes of participating UMG artists' songs as a paid add-on. Terms, including price point, launch date, revenue split, participating catalog, and content moderation rules, were not disclosed in the announcement19. The framing ("consent, credit and compensation") is now the de facto industry template for licensed AI music.
4.2 Deezer
Deezer's April 20, 2026 disclosure is the single most useful published dataset on AI supply. Deezer reports: nearly 75,000 AI-generated tracks uploaded daily, roughly 44% of total daily deliveries, up from ~10,000/day when its proprietary detector launched in January 2025 (with intermediate readings of ~30,000/day in September 2025, ~50,000 in November 2025, and ~60,000 in January 2026)16. Despite that supply glut, AI tracks generate only 1-3% of total streams. Of those AI-track streams, 85% are flagged as fraudulent and demonetized, evidence that a large share of AI-generated uploads are being used for click-farm royalty-fraud schemes rather than for genuine consumption17. Deezer's stance is significant because it publishes hard operator data. Most other platforms do not.
4.3 YouTube
YouTube's Dream Track experiment (November 2023) established the template for licensed AI music: DeepMind's Lyria model, a limited-cohort US creator pilot, 30-second outputs, and, most importantly, a nine-artist licensed voice set20. The parallel Music AI Incubator has tested generative style-conversion and riff-from-humming tools with a broader artist cohort. SynthID audio watermarking is deployed on Lyria outputs, providing a technical basis for downstream detection.
4.4 SoundCloud
SoundCloud's February 12, 2024 terms-of-use update quietly permitted user-uploaded content to be used to train AI/ML systems, a change surfaced in May 2025 by researcher Ed Newton-Rex via the Internet Archive. The backlash forced SoundCloud to clarify (a) that it had not in fact used artist content to train models; (b) that content covered by separate rights-holder agreements (UMG, Warner) was carved out; and (c) that it maintained a "no AI" tag as a technical safeguard21. The episode is a case study in the reputational cost of ambiguous training-data terms.
4.5 Apple Music, Amazon Music, TikTok
Apple Music has adopted a comparatively quiet posture. In its 2025 disclosures the platform reported demonetizing approximately two billion fraudulent streams during 2025, representing nearly USD 17M in royalties diverted from legitimate artists, with fraud characterized as less than half of one percent of total streams. Apple has not launched a public generative product23. Amazon Music and TikTok have taken analogous stances: TikTok's ByteDance-owned Ripple is the largest in-app AI music surface by potential reach, but detailed usage figures are not publicly disclosed.
5. Monetization and Business Models
5.1 Consumer Subscription and Credits
The dominant consumer model is freemium + tiered subscription with credit-based generation quotas. Suno's Pro plan (USD 10/mo) unlocks commercial rights and a fixed monthly credit allowance; Premier (USD 30/mo) expands both. Annualized pricing offers a 20% discount. Udio's pre-settlement free tier of 600 songs per month was the most generous in the market and drove rapid seed growth; its transition to a licensed streaming model post-settlement will materially reshape unit economics. ElevenLabs sells credits across its full audio stack, with music generation drawing from a shared pool.
5.2 B2B, API, and Sync Licensing
Enterprise revenue arrives through three channels: (1) API access sold to app developers and enterprise creative teams (ElevenLabs, Mubert, Stability); (2) direct sync licensing to advertisers, game studios, and podcast networks looking for royalty-free background music (Soundraw, Beatoven, Mubert); and (3) sample-library and mastering bundles (LANDR). The CISAC study projects EUR 5B of the EUR 16B 2028 music-outputs pool sits in B2B use, the segment least exposed to the platform-side content-policy wars.
5.3 The Emerging "Licensed Platform" Model
The post-settlement templates emerging from UMG-Udio (October 2025), Warner-Suno (November 2025), and Spotify-UMG (May 2026) share three features: (a) training exclusively on authorized/licensed catalogs; (b) artist opt-in for name, image, voice, and composition; and (c) revenue sharing with rights holders on a per-generation or per-play basis12,19. This is a structurally different unit economics from the current unlicensed-training model, closer to a mechanical-license structure than to a pure software subscription, and likely defines the terms of entry for the next wave of entrants.
6. Legal and Regulatory Landscape
6.1 The RIAA-Coordinated Copyright Lawsuits
On June 24, 2024 the three major labels (UMG, Sony, and Warner) filed coordinated copyright infringement lawsuits against Suno (D. Mass. 1:24-cv-11611) and Udio/Uncharted Labs (SDNY 1:24-cv-04777). The complaints allege willful copyright infringement, DMCA violations for circumventing YouTube's "rolling cipher" (in the Suno case) to scrape training data, and Music Modernization Act violations. Initial damages were pleaded at up to USD 150,000 per infringed work, approximately USD 84M in the Suno matter on 560 recordings11,13.
The procedural picture as of July 2026 is materially different from the initial pleadings:
Warner-Suno settlement (November 25, 2025). Multi-million-dollar payment; artist opt-in rights for name, image, voice, and composition; retirement of Suno's current models; new models trained on licensed music launching in 2026. Suno additionally acquired concert-data firm Songkick from WMG. Dismissal entered January 28, 202612.
UMG-Udio settlement (October 29, 2025). Compensation plus a commitment to a licensed-only platform with opt-in artist compensation; dismissal filed November 4, 202512.
Suno case expansion motion (May 2026). UMG and Sony moved to expand the Suno complaint from the original 560 works to 61,026 recordings, identified via audio-fingerprinting service Audible Magic. If granted, theoretical statutory damages exceed USD 9.1B13. A summary-judgment hearing on Suno's fair-use defense is scheduled for July 2026 before Chief Judge F. Dennis Saylor IV; fact-discovery deadline September 30, 2026; dispositive motions due April 9, 2027.
Sony-Udio (SDNY). On April 15, 2026 the court ruled Sony's DMCA anti-circumvention claim survived dismissal. On June 30, 2026 the court denied Sony's motion to expand the case from 333 to 30,442 additional recordings, leaving Sony litigating alone on the original works after UMG and Warner settled13.
Parallel European actions. Danish collecting society Koda filed against Suno in November 2025; German society GEMA filed in January 2025. Additional independent-artist class actions are pending in Massachusetts and the Northern District of California13.
The Suno fair-use ruling, expected in H2 2026 or H1 2027, is the single most consequential legal event on the AI-music horizon. A ruling for the labels would validate the CISAC revenue-cannibalization thesis; a ruling for Suno (in whole or in part) would materially compress rights-holder leverage in future licensing negotiations.
6.2 US Statutes: NO FAKES Act and Tennessee's ELVIS Act
Tennessee enacted the ELVIS (Ensuring Likeness, Voice, and Image Security) Act on March 21, 2024, effective July 1, 2024. The Act (a) expands right-of-publicity to explicitly cover unauthorized commercial use of an individual's voice, including AI-simulated voices where no actual recording was used; (b) grants record companies standing to sue on behalf of exclusively-signed artists, giving major labels a direct litigation tool against AI music services; and (c) creates novel secondary liability for developers of software whose primary purpose is producing an identifiable individual's voice or likeness without authorization15. Because so much of the AI-music ecosystem operates through platforms with a Tennessee touch (Nashville-headquartered labels, publishers, and PROs), the Act has extraterritorial bite.
At the federal level, the NO FAKES Act (Nurture Originals, Foster Art, and Keep Entertainment Safe) was originally introduced in October 2023 and has been reintroduced and revised through 2025-2026. It would create a federal digital-replica right and secondary liability for platforms hosting unauthorized deepfakes. The parallel No AI FRAUD Act (introduced January 2024) covers similar ground on the House side. As of July 2026, neither has been enacted.
6.3 EU AI Act and CDSM Directive Article 4
The EU AI Act's general-purpose AI (GPAI) obligations came into force on August 2, 2025. Article 53(1)(c) requires GPAI providers to "put in place a policy to identify and comply with reservation of rights" under the EU's existing text-and-data-mining (TDM) exception in Article 4 of the 2019 Digital Single Market (CDSM) Directive, the copyright opt-out mechanism rights-holders use to withhold works from AI training. Article 53(1)(d) additionally requires publication of "a sufficiently detailed summary" of training content using a mandatory template issued by the European AI Office14. Non-compliance can trigger fines of up to 3% of annual worldwide turnover or EUR 15M (whichever is higher), with enforcement beginning August 2, 2026 after a grace period.
For AI music generators, the practical implication is that rights-holders (labels, publishers, PROs) can, and increasingly do, issue machine-readable opt-outs on their catalogs, which providers must honor when training. This is functionally a copyright licensing regime dressed as a transparency obligation.
6.4 International: UK, Japan, China
The UK abandoned its proposed broad TDM exception in early 2023 after industry pushback; a subsequent government consultation on AI-and-copyright ran through 2025 without a clear legislative outcome. Japan's copyright regime remains the most permissive of the major markets: Article 30-4 of the Japanese Copyright Act broadly permits training on copyrighted works "where the purpose is not to enjoy the works as such." China's Interim Measures for the Management of Generative AI Services (August 2023) require training data to respect intellectual-property rights and prohibit generating content that infringes rights of others, implemented in practice through platform-level content filtering rather than ex-ante licensing regimes.
6.5 US Copyright Office
The US Copyright Office's multi-part AI report, spanning Part 1 (Digital Replicas, July 2024), Part 2 (Copyrightability, January 2025), and Part 3 (Generative AI Training, released in pre-publication form May 2025), has three material takeaways for AI music: (a) purely AI-generated outputs are not copyrightable; only human-authored contributions to a mixed work are; (b) the Office supports federal digital-replica protection (aligning with the NO FAKES Act); and (c) on training, the Office stops short of endorsing a categorical fair-use defense and flags market-substitution as a key fair-use factor working against AI developers, directly on point for the Suno and Udio cases.
Table 2. Major Label Positions and AI Deals (as of July 2026)
Rights Holder | Litigation Posture | Confirmed AI Deals | Strategic Signal |
|---|
Universal Music Group | Sole remaining active plaintiff on Suno expansion motion (US); settled with Udio Oct 2025 | YouTube Dream Track (Nov 2023); Udio walled-garden platform (Oct 2025); Spotify AI covers & remixes (May 2026) | Setting the licensed-platform template; framing "consent, credit, compensation" |
Sony Music Entertainment | Continues Suno case (Mass.) and Udio case (SDNY) after UMG/Warner settled; June 2026 loss on Udio expansion motion | Participant in SAG-AFTRA labor AI deal (Apr 2024) | Most litigious of the three majors; slower to license |
Warner Music Group | Settled Suno Nov 2025; settled Udio in late 2025 | Suno licensing partnership (Nov 2025); SoundCloud carve-out on AI training terms | First mover on Suno settlement; sold Songkick to Suno as part of deal |
Merlin Network (indies) | Not a plaintiff | ElevenLabs Music license (2025) | Independent labels are moving faster than majors on licensed AI |
Kobalt Music Group | Not a plaintiff | ElevenLabs Music license (2025) | Publishing-side signal on licensed AI |
Sources: label press releases, RIAA court filings, MBW, Legal Clarity, Chartlex litigation tracker.
7. Industry Reactions and Disputes
7.1 The Artist Rights Alliance Open Letter
On April 2, 2024 the Artist Rights Alliance published an open letter signed by more than 200 recording artists (including Billie Eilish, Nicki Minaj, Stevie Wonder, Katy Perry, Pearl Jam, Jon Bon Jovi, Imagine Dragons, and the estates of Frank Sinatra and Bob Marley), calling on AI developers and platforms to "cease the use of artificial intelligence (AI) to infringe upon and devalue the rights of human artists." The letter names two specific misuses: unauthorized voice cloning and the use of AI-generated content to "dilute royalty pools" already squeezed by streaming economics24. Executive Director Jen Jacobsen framed the concern: "Working musicians are already struggling to make ends meet in the streaming world, and now they have the added burden of trying to compete with a deluge of AI-generated noise."
7.2 SAG-AFTRA Labor Agreement
In April 2024 SAG-AFTRA (representing ~160,000 media professionals, including recording-artist union members) announced a tentative AI-protections deal with Sony Music, Universal, Warner, and Disney Music Group, covering the period January 1, 2021 through December 31, 202625. The agreement requires clear artist consent for AI voice cloning, defines "artist" to include the individual's digital voice replica, and establishes bargaining rights over the use of AI-generated instrumentals in signed-artist recordings.
7.3 The "Heart on My Sleeve" Incident and Deepfake Precedent
The 2023 viral track "Heart on My Sleeve," which cloned the voices of Drake and The Weeknd, drew rapid takedowns from streaming platforms after Universal Music Group asserted its rights. The incident is the industry's reference case for unauthorized artist-voice cloning and is repeatedly cited in the artist-rights and legislative debates around the NO FAKES Act and ELVIS Act. It is also the event that catalyzed Spotify's later impersonation-policy overhaul.
8. Emerging Trends
8.1 Synthetic Artists and Royalty Fraud
The July 2025 "Velvet Sundown" case study crystallized the synthetic-artist risk vector. An AI-generated band operating under a plausible-sounding indie identity accumulated over one million monthly Spotify listeners in under two months, releasing two albums (a third scheduled for July 14, 2025). Deezer flagged the album as 100% AI-generated. The account initially denied AI involvement, then updated its Spotify bio to disclose that the project was "composed, voiced, and visualized with AI support." Spotify publicly stated it does not prioritize or benefit financially from AI-generated music22. Coupled with Deezer's finding that 85% of streams to AI-flagged tracks are fraudulent, the pattern points to a structural royalty-fraud channel: mass-generate AI tracks, upload at scale, drive artificial plays through click farms, and extract from the streaming royalty pool. Forbes' May 2026 investigation put the estimated size of this ecosystem in the multi-billion-dollar range23.
8.2 Watermarking and Detection
Two watermarking and detection layers are emerging as de facto industry infrastructure. Google DeepMind's SynthID applies inaudible watermarks to Lyria-generated audio, allowing downstream platforms to identify AI-generated content even after transcoding. Deezer's proprietary detector (deployed January 2025) reports 44% detection accuracy at the platform level, forming the basis for its demonetization pipeline. The DDEX AI-credit disclosure schema, backed by Spotify and other DSPs, adds a metadata layer at the point of upload. None of these systems is perfect; but together they define a rising floor of enforceability against synthetic-artist fraud.
8.3 Personalization, AI DJ, and Agent Curation
On the consumer-experience side, Spotify's AI DJ (launched 2023, expanded through 2024-2025 with AI Playlist and localized voices) is the current benchmark for personalization via generative AI applied to human catalogs. YouTube Music has parallel features. The 2026 Spotify-UMG deal extends this into user-participatory generation, letting subscribers create AI covers and remixes of licensed catalog tracks, which likely marks the beginning of an "agent-curation" era where the streaming experience is generatively assembled rather than passively played back.
8.4 The Walled-Garden Licensed Platform
The most important product-level trend is the emergence of the "walled-garden" licensed platform pioneered by the Udio-UMG settlement: a generation service trained exclusively on licensed catalogs, with artist opt-in and per-generation compensation. It resolves the training-data dispute by construction, at the cost of shrinking the training set. Whether users will accept a smaller, licensed model is the open commercial question of 2026-2027.
9. Risks and Outlook
9.1 Royalty-Pool Dilution
The central financial risk to the broader music industry is royalty-pool dilution, a scenario in which each incremental AI track competes for a share of a fixed pro-rata streaming payout. The CISAC/PMP Strategy study projects the human-creator share of the music streaming pool falls from 8.2% (2023) to 6.4% (2028), a 1.8-point loss worth ~EUR 0.9B in 2028 alone and EUR 2.3B cumulatively. The total 2028 music-streaming pool grows to EUR 57B, of which the study models 18% as AI content replacing human music and 3% as AI-driven incremental growth5. Whether platforms move to user-centric royalty models (where a listener's subscription is paid pro-rata to the artists they actually stream) is the single most important variable in whether that dilution actually materializes.
9.2 Litigation Risk
For AI-music companies, the litigation risk is bifurcated. Firms that have settled (Suno with Warner, Udio with UMG) have converted an existential legal exposure into a licensing cost line item; firms that have not are exposed to statutory damages that scale with the size of the labels' claim. The May 2026 expansion motion in the Suno case, pushing potential damages from ~USD 84M to over USD 9B, illustrates the tail risk. A successful summary-judgment fair-use ruling for Suno in H2 2026 would shift this dynamic materially; a loss would validate the current settlement path.
9.3 Regulatory Tightening
EU AI Act GPAI obligations bite in August 2026. Federal US action (NO FAKES Act) remains uncertain but well-supported. Tennessee's ELVIS Act is being copied at the state level. The direction of travel is one-way: more disclosure, more opt-out, more secondary liability. Regulatory risk is not a probability question; it is a timing question.
9.4 Platform De-Listing and Enforcement
Spotify's 75M-track removal wave and Deezer's 85% demonetization rate on AI-flagged streams are early evidence that platform-side enforcement is the practical rate-limiter on the "flood the streaming pool" business model. Firms whose economics depend on mass, low-quality AI uploads face binary platform risk; firms whose economics depend on premium licensed content do not. This split will define the winners over the next 24-36 months.
9.5 Consolidation Path
The USD 11B ElevenLabs Series D at 3× the prior round is emblematic of an unusual dynamic: the sector's best-capitalized independents are trading at premium multiples in a market whose legal foundations are actively being contested. That combination is historically the setup for a consolidation cycle in which strategic acquirers (majors, DSPs, hyperscalers) buy their way to catalog-safe AI stacks. Expect: (a) at least one major-on-AI-startup acquisition by end-2027; (b) horizontal roll-ups among vertical tools (mastering, stems, sync); and (c) hyperscaler-driven bundling of AI music into broader audio stacks (voice, dubbing, sound design).
10. Conclusion
The AI music market in mid-2026 is the highest-velocity, highest-legal-risk consumer software category since generative image models in 2022-2023. The commercial signals, from Suno's USD 300M ARR to ElevenLabs' USD 11B valuation and 75,000 AI tracks per day on Deezer, argue for a decisive, category-defining next 18 months. The legal signals, from RIAA v. Suno moving to summary judgment on fair use, to UMG and Warner banking on licensed-only platforms, to EU AI Act enforcement live from August 2026, argue that the winning business model is not "generate anything from any training data" but "generate rich licensed catalogs under enforceable rights."
For strategic investors, three signals will resolve most of the uncertainty by end-2027: (1) the fair-use ruling in RIAA v. Suno; (2) whether the Spotify-UMG AI product ships and monetizes; and (3) whether platform demonetization of AI-flagged streams meaningfully compresses the synthetic-artist fraud channel. Firms that solve for licensing, watermarking, and platform-side legitimacy early will define the sector's durable structure. Firms that continue to operate at the training-data frontier are betting on a fair-use ruling that, as of this writing, the US Copyright Office and most trial-court signaling do not support.
11. Sources and Methodology
All revenue and user figures for private companies are third-party estimates or company self-reported unless otherwise noted. Market-size figures reflect the boundary drawn by each cited research firm and are not directly comparable. Numbered citations correspond to superscripts in the body of the memo.
1. Grand View Research — Generative AI in Music Market Size, Industry Report 2030. https://www.grandviewresearch.com/industry-analysis/generative-ai-in-music-market-report
2. Goldman Sachs Research — "Music in the Air: Focus on monetisation, emerging markets and AI" (May 3, 2024). https://www.goldmansachs.com/insights/goldman-sachs-research/music-in-the-air--focus-on-monetisation--emerging-markets-and-ai
3. Market.us — AI in Music Market Report. https://market.us/report/ai-in-music-market/
4. PMP Strategy / CISAC — press release, first global study on economic impact of Generative AI in music and audiovisual (Dec 4, 2024). https://pmpstrategy.com/en/2024/12/04/pmp-strategy-and-cisac-unveil-the-first-global-study-on-the-economic-impact-of-generative-ai-in-the-music-and-audiovisual-industries/
5. CISAC / PMP Strategy — full study PDF (Dec 2024). https://app.zaiks.org.pl/media/xebmoq2r/cisac-pmps_gen-ai-impact-study-on-music-and-audiovisual-complete-report_final.pdf
6. Music Business Worldwide — "Market for Gen AI outputs to be worth over $16bn annually by 2028…" (Dec 4, 2024). https://www.musicbusinessworldwide.com/market-for-gen-ai-outputs-to-be-worth-over-16bn-annually-by-2028-but-it-could-cannibalize-24-of-music-creators-revenues-cisac-predicts/
7. TechCrunch — "Legally embattled AI music startup Suno raises at $2.45B valuation on $200M revenue" (Nov 19, 2025). https://techcrunch.com/2025/11/19/legally-embattled-ai-music-startup-suno-raises-at-2-45b-valuation-on-200m-revenue/
8. Music Business Worldwide — "Suno hits 2M paid subscribers, $300M annual revenue" (Feb 2026). https://www.musicbusinessworldwide.com/suno-hits-2m-paid-subscribers-300m-annual-revenue/
9. ElevenLabs — Series D announcement blog post (Feb 4, 2026). https://elevenlabs.io/blog/series-d
10. Wikipedia — Udio (company). https://en.wikipedia.org/wiki/Udio
11. Legal Clarity — "Suno Lawsuit Explained: From Filing to Warner Settlement". https://legalclarity.org/suno-lawsuit-explained-from-filing-to-warner-settlement/
12. Music Business Worldwide — "Why a fight over 61,000 recordings could shape the future of AI music licensing" (Jul 2026). https://www.musicbusinessworldwide.com/why-a-fight-over-61000-recordings-could-shape-the-future-of-ai-music-licensing/
13. Chartlex — Music Industry AI Lawsuits Tracker 2026. https://www.chartlex.com/blog/business/music-industry-ai-lawsuits-tracker-2026
14. Clifford Chance — "Copyright compliance under the EU AI Act for GPAI model providers" (Nov 2025). https://www.cliffordchance.com/insights/resources/blogs/ip-insights/2025/10/copyright-compliance-under-the-eu-ai-act-for-gpai-model-providers.html
15. Latham & Watkins — "The ELVIS Act: Tennessee Shakes Up Its Right of Publicity Law" (Apr 2024). https://www.lw.com/admin/upload/SiteAttachments/The-ELVIS-Act-Tennessee-Shakes-Up-Its-Right-of-Publicity-Law-and-Takes-On-Generative-AI.pdf
16. Deezer Newsroom — "AI-generated tracks represent 44% of new uploaded music" (Apr 20, 2026). https://newsroom-deezer.com/2026/04/ai-generated-tracks-represent-44-of-new-uploaded-music/
17. DJ Mag — "85% of AI-generated music streams have been demonetised — Deezer". https://djmag.com/news/85-of-ai-generated-music-streams-have-been-demonetised-deezer
18. Spotify Newsroom — "Spotify strengthens AI protections" (Sep 25, 2025). https://newsroom.spotify.com/2025-09-25/spotify-strengthens-ai-protections/
19. Forbes (Maureen Kerr) — "Spotify and Universal Just Put AI Music Behind a Paywall" (May 22, 2026). https://www.forbes.com/sites/maureenkerr/2026/05/22/spotify-and-universal-just-put-ai-music-behind-a-paywall/
20. YouTube Blog — "AI and music experiment: Dream Track" (Nov 16, 2023). https://blog.youtube/inside-youtube/ai-and-music-experiment/
21. TechCrunch — "SoundCloud changes policies to allow AI training on user content" (May 9, 2025). https://techcrunch.com/2025/05/09/soundcloud-changes-policies-to-allow-ai-training-on-user-content/
22. Euronews — "The Velvet Sundown explained: Spotify-verified AI band controversy" (Jul 8, 2025). https://www.euronews.com/culture/2025/07/08/the-velvet-sundown-explained-whats-behind-the-spotify-verified-ai-band-controversy
23. Forbes (Virginie Berger) — "How AI-Generated Music Became a $4 Billion Fraud Machine" (May 5, 2026). https://www.forbes.com/sites/virginieberger/2026/05/05/how-ai-generated-music-became-a-4-billion-fraud-machine/
24. Artist Rights Alliance — "200+ Artists Call on AI Developers, Tech Platforms Not to Devalue Music and Undermine Artists" (Apr 2, 2024). https://artistrightsnow.medium.com/200-artists-call-on-ai-developers-tech-platforms-not-to-devalue-music-and-undermine-artists-2727e17bc10a
25. Billboard — "SAG-AFTRA-Major Labels Contract with Artificial Intelligence Protections" (Apr 15, 2024). https://www.billboard.com/business/business-news/sag-aftra-major-labels-contract-artificial-intelligence-protections-1235657060/
Methodology: this memo synthesizes a verified research base of 28 primary and secondary sources, filtered through an adversarial-verification pass (majority-vote refutation, 3-voter panel per claim). Claims sourced solely from stakeholder-commissioned studies (notably CISAC/PMP Strategy) are flagged in-line; single-primary-source claims are attributed to the reporting entity. Legal, funding, and platform-policy figures move quickly; any operational figure older than six months should be re-verified before publication or reliance.