AI Content and the Creative Supply Chain

Core Conclusions

• The AI content and creative industry has moved from "model demos" into "workflow commercialization," yet the profit pool is not pooling mainly at the pure-model layer. It is flowing instead toward incumbent platforms that already own distribution, creative workflow, customer relationships, brand governance, and a copyright firewall. The clearest picture of a winner today is a platform company that combines four things (model, workflow, customer-budget access, and compliance governance), rather than a single-point generation tool: think Adobe, Canva, Figma, Meta, Google, Kuaishou, Publicis, and WPP, plus UMG, Getty, and Shutterstock with their copyright and licensing capabilities.

• The use cases already producing real revenue rank roughly as follows: image/design enhancement, ad-creative automation, marketing-delivery optimization, enterprise video and digital humans, localization and dubbing, e-commerce product imagery, and copyright-safe image generation. Adobe delivered $23.769 billion in revenue in FY2025 with Digital Media ARR reaching $19.2 billion, and Firefly-related ARR already exceeded $250 million in Q1 FY2026; Kuaishou's Kling booked RMB 340 million of revenue in 2025Q4 alone; Meitu's 2025 imaging/video/design revenue reached RMB 2.954 billion with 16.91 million paying subscribers; Meta has disclosed that over 1 million advertisers use its generative-AI tools each month to produce more than 15 million ads.

• The use cases still stuck in pilots, concepts, or subsidy-driven competition are mainly feature-length, cinema-grade video generation, consumer-facing standalone AI image/video apps, open AI-music platforms, AI-native virtual idols, and "fully automated game generation." These areas either still carry high inference costs, suffer from fragile retention, lack settled copyright and personality-rights constraints, or have no enterprise-grade budget access. OpenAI retired the original Sora product and app in April 2026, showing that video-model branding and interaction patterns are still iterating fast; Midjourney, meanwhile, has been sued repeatedly by Disney/Universal and Warner Bros. Discovery, and that copyright risk directly affects commercial sustainability.

• AI's first-stage impact on the content industry is "higher efficiency and lower cost," its second stage is "expanding content supply," and only its third stage is "reshaping copyright pricing and content consumption." What has actually shown up on the financial statements so far is more about creative-production speed, marketing-asset throughput, ad-conversion efficiency, and some subscription ARPU uplift, not the wholesale replacement of content companies. Kuaishou has disclosed that online marketing spend driven by AIGC marketing assets reached RMB 4 billion in 2025Q4; both Meta and Google tie AI directly to ad conversion and delivery optimization rather than selling a standalone "AI seat."

• The track with the highest revenue certainty is the generation closest to a budget line item, rather than the flashiest generation. Enterprise marketing budgets, advertising budgets, design-software budgets, product-shoot budgets, video-localization budgets, asset-procurement budgets, and copyright budgets are already the paying sources where AI lands for real; budgets for feature films, AI-music consumer platforms, and open-ended virtual-human livestreaming still rely more on experimental spending.

• The long-term profit pool is more likely to settle in the creative operating system than in any single model. Adobe is packaging Firefly, GenStudio, AEP, Content Credentials, and enterprise custom models into a content supply chain; Canva is turning AI into a "design OS" through Magic Studio and agentic products; Figma is embedding AI directly across the chain from brainstorm to design to prototype to build. Whoever controls brand assets, approvals, collaboration, asset management, and the delivery interface has the best chance of holding onto high margins in the AI era.

• Ad platforms and marketing groups are direct AI-content beneficiaries the market underrates. Meta, Google, Pinterest, and Kuaishou tie AI directly to conversion rate, ROAS, asset generation, automated delivery, and creative testing, while Publicis, WPP, and Omnicom build a "human-machine co-pilot agency" model across customer data, cross-media delivery, process automation, and brand governance. Platform companies capture budget migration and efficiency dividends; agency companies capture delivery-model restructuring and a lower marginal labor cost.

• Copyright and training-data licensing has not become an "irrelevant variable in the AI era." It is rising into a new revenue line and a source of differentiation. Getty and Shutterstock use "commercial-use rights plus indemnity plus licensed data" to separate themselves from unlicensed generation tools; UMG and WMG are shifting from "suing AI" to "conditionally licensing AI," trying to turn training and likeness rights into a new revenue pool. The U.S. Copyright Office, the EU, and Japan's Agency for Cultural Affairs are all steadily clarifying the boundary between AI and copyright, which means "whether you can generate it" is gradually giving way to "whether you can lawfully commercialize it at scale."

• The most worthwhile listed companies are not all in the United States. In U.S. equities, Adobe, Meta, Figma, Autodesk, Roblox, Pinterest, and Getty/Shutterstock represent different profit pools; Europe has UMG, Publicis, and WPP; among Hong Kong and China-concept names, Kuaishou, Meitu, and Tencent Music carry the most concrete evidence of AI-content commercialization. Conversely, some companies that "tell AI-video, AI-game, and AI-digital-human stories" remain at the product-launch and user-trial stage, still some distance from scaled adoption and profit contribution.

• Companies whose valuations already reflect a lot of AI-content expectation typically combine a high-growth platform narrative with direct financial delivery, for example Figma, Spotify, AppLovin, and some AI-native private firms; companies where an expectation gap may still exist between valuation and fundamentals lean more toward "AI already generating revenue, but the market still treats it as legacy software or a legacy platform," such as Adobe, Meitu, Kuaishou, and Getty/Shutterstock. This judgment rests more on relative comparison of current revenue delivery, narrative heat, and verifiable incremental business than on any external fact.

• The biggest catalysts over the next 12–24 months are not "bigger models," but four things: first, AI-video inference costs keep falling; second, copyright-licensing and likeness/voice-licensing mechanisms take shape; third, AI generation plugs directly into the ad-delivery and e-commerce-conversion loop; fourth, enterprise customers shift from feature trials to workflow-level procurement. Adobe GenStudio, Kuaishou Kling, Meta Advantage+, Google AI Max, Publicis CoreAI, and WPP Open are all advancing in this direction.

• The top risks remain three: copyright and personality-rights litigation; inference cost and price wars; and free general-purpose models displacing standalone tools. Litigation has already tested the Midjourney, Stability, and Suno/Udio paths; Shutterstock's Q1 2026 and Getty's 2025 results also show that AI does not only bring incremental revenue; it reshapes legacy businesses and triggers M&A and pricing pressure.

Supply-Chain and Profit-Pool Map

The table below breaks the AI-content supply chain down by "who creates the capability, who controls the workflow, who captures the budget, who holds the copyright, and who controls distribution." In the table, "benefit intensity" measures revenue delivery over the past two years, and "investment elasticity" measures the slope at which profit or valuation gets re-rated as AI penetration accelerates; both are this report's analytical scores.

Judgment: Over the next three years, the most likely place for excess profit to settle is "the most stable budget access + the deepest workflow integration + the clearest copyright governance," rather than "the strongest generation quality." That is why Adobe, Canva, Figma, Meta, Google, Kuaishou, Publicis, and the UMG/copyright-library companies sit closer to the center of the long-term profit pool than most single-purpose AI image/video startups.

Business Models, Value Capture, and Scenario Forecasts

The way AI-content products charge has visibly stratified. The genuinely healthy, sustainable business models fall broadly into six categories: the first is design/creative software seat upgrades, for example Adobe bundling Firefly credits, Acrobat AI, GenStudio, and enterprise editions into higher-ARPU packages; the second is SaaS subscription plus usage-based billing, for example Meitu adding token/single-feature/flat-rate hybrid pricing on top of subscriptions; the third is API/credits, better suited to Runway, voice, and digital humans; the fourth is a platform-advertising-budget take rate, where Meta, Google, Pinterest, and Kuaishou use AI to get advertisers to spend more; the fifth is copyright/asset licensing, where Getty, Shutterstock, and UMG charge through indemnity, commercial usability, and catalog licensing; the sixth is enterprise solutions and service-based delivery, where Publicis, WPP, Synthesia, and HeyGen embed AI into promotion, training, and brand-governance processes. The first four have a better chance of forming long-term repeat purchases; the model of simply "selling generation results" is more easily eroded by low-price competition and general-purpose models.

From an enterprise-budget view, the marketing department's AI-content budget flows first to ad-asset generation, product imagery, short video, localization, copy and landing-page production, and then to delivery optimization and brand governance; the ad agency's AI budget flows more toward internal workbenches, model integration, content generation, version management, approvals, and post-campaign analysis; the game company's AI budget lands more on asset prototyping, translation, localization, testing, and some NPC/UGC tools, rather than one-click generation of a complete game; the film company's AI budget lands mainly on previs, storyboards, concept art, captions, multilingual dubbing, and post-production assistance, rather than fully replacing the core creative process. Kuaishou has disclosed that AIGC marketing assets corresponded to RMB 4 billion of online marketing spend in 2025Q4; WPP Open's monthly active users have reached 33,000; Roblox uses AI primarily for creator efficiency and platform safety, not for directly selling a "generated-content subscription."

The largest value capture is shifting from a "one-time outsourcing fee" toward "recurring software fees + usage fees + ad take rate." This means three kinds of companies are best positioned: first, design software that can upgrade AI into higher-ARPU packages; second, platforms that can turn AI into higher ad spend and better ROAS; third, vendors that can promise enterprises copyright safety, auditability, and brand control. By contrast, pure point tools that charge "per image/per video" tend to fall into a price war quickly if they lack workflow and customer relationships.

The table below summarizes the pros and cons of several mainstream charging models with matching cases.

Three scenarios follow. The "software-revenue growth," "ad-ROI improvement," and "beneficiary companies" here are analytical extrapolations based on disclosed adoption data, platform ROAS cases, and each company's current commercialization stage, not company guidance.

My judgment: The best long-term investment answer is not betting on "whose model is strongest," but betting on "who can turn AI into a recurring paid product or a budget access point." Design software, ad platforms, enterprise marketing workflows, copyright licensing, and safe generation are therefore better asset shapes than open-ended video/music consumer tools.

Deep Dive Into Sub-Segments

The table below gives a compressed breakdown of the 30 sub-segments provided by the user. Scores are on a 10-point scale and evaluate revenue certainty over the next two years × margin quality × moat strength.

If I had to pick only the 5 sub-segments worth watching most closely over the next 12–24 months, I would name: AI design software, AI ad creative and delivery optimization, AI short-form/enterprise video, AI copyright licensing and safe asset generation, and AI enterprise marketing-content platforms. These five sit closest to real budgets, repeat purchases, and platform moats.

Master Target Table and Company Tiering

The table below prioritizes the listed companies with the strongest evidence of revenue landing, the clearest platform attributes, or the largest potential expectation gap. Because valuation conventions, timing, and disclosure habits differ greatly across regions, metrics that cannot be consistently verified are uniformly marked "needs further verification."

A–E Tiering:

Key Listed Companies and Private Challengers

First, the 15 listed companies I think most worth continuing to dig into. Read this as a "research priority," not a buy/sell recommendation.

Next, the 10 private companies most worth tracking. Many of these have not fully disclosed financial data; any metric that is not an official company disclosure should be treated as needing further verification.

Overall judgment on the private direction: What is truly worth watching in private markets is companies that can build a moat of brand safety, workflow, copyright, or customer relationships within a given industry, rather than "yet another image/video generator": for example Canva, Synthesia, HeyGen, Scenario, and the AI music companies that secure genuine licensing. Companies relying only on faster generation and cheaper credits usually lack a thick enough moat.

Competitive Landscape, Valuation Expectations, Risks, and Final Conclusions

On the competitive landscape, the difference among Adobe, Canva, Figma, Autodesk, and Roblox comes down to which workflow each has captured, rather than "who can or cannot do generation": Adobe captures cross-media creation and the content supply chain, Canva captures mass-market design and marketing expression, Figma captures product design and development collaboration, Autodesk captures industrial/architectural/3D professional content, and Roblox captures the UGC world and creator ecosystem. By comparison, Google Veo / AI Max, Meta Advantage+, and Kuaishou Kling sit closer to content distribution and the ad loop; Runway, Midjourney, Pika, and Luma lean more toward the model/tool layer; and Getty/Shutterstock/UMG/WMG represent the "copyright and licensing layer."

AI's impact on traditional creative jobs runs deeper than a simple "cut the creators": it automates low-end repetitive steps while elevating high-end taste, brand governance, topic selection, world-building, IP management, and cross-channel orchestration into scarcer capabilities. The jobs and companies most hit are most likely low-end design outsourcing, static-asset production, basic dubbing, caption translation, low-complexity ad execution, and low-end game-art outsourcing; the biggest beneficiaries are the companies that can integrate these steps into products and platforms. WPP, Publicis, and BlueFocus have publicly described AI's restructuring of their internal production methods; Roblox and Kuaishou also show that AI mainly amplifies supply and operating efficiency first.

From a valuation-and-expectation angle, this report sorts the key companies into four groups. Group one, "good platforms and already expensive": Figma, AppLovin, and some high-momentum AI platforms; group two, "AI revenue genuinely growing but valuation not necessarily fully reflecting it": Adobe, Meitu, Kuaishou, Publicis, and Getty; group three, "strong narrative but insufficient financial validation": Unity, many pure AI video/music startups, and some digital-human concepts; group four, "hit by AI but possibly re-rated via copyright and M&A": stock libraries like Getty/Shutterstock. This grouping carries an obvious analytical-judgment quality, but the evidence behind it is whether we already see paying users, ARR, ad budgets, licensing revenue, or margin improvement.

A simplified scoring model follows. The total score is this report's analytical result, not an external fact; but the scoring basis tries to correspond to public disclosure evidence.

On the inverse risk model, the highest copyright-litigation and training-data risk sits with Midjourney, Suno/Udio, and some open-ended image/music startups; the highest inference-cost pressure sits with video generation; the highest user-retention and commoditization risk sits with standalone consumer-facing image/video apps; the highest overvaluation risk sits with high-growth platforms and some popular private companies; and the highest general-model substitution risk sits with point tools that lack a workflow moat. Midjourney's litigation, Stability's dispute with Getty, OpenAI's Sora brand adjustment, and the U.S. Copyright Office's successive reports on copyrightability and training all show that "model performance" cannot substitute for "compliant commercial usability."

The final conclusions are as follows:

First, the AI content and creative industry is one of the application layers closest to revenue in the AI supply chain, but the real value lies in "who embeds generation into budgets and workflows," rather than in "generation capability itself."

Second, the 5 sub-segments most worth watching are: AI design software, AI ad creative/delivery optimization, AI short-form and enterprise video, AI copyright licensing and safe assets, and AI enterprise marketing-content platforms.

Third, the 10 listed companies most worth deep research are: Adobe, Meta, Figma, Kuaishou, Meitu, Publicis, Autodesk, UMG, Roblox, and Getty. They respectively represent five profit pools: the creative OS, ad-budget access, the collaborative design platform, video-model commercialization, China imaging AI, the marketing workflow, the professional design workflow, copyright licensing, the UGC ecosystem, and safe-asset licensing.

Fourth, the 10 private companies most worth tracking are: Canva, Synthesia, HeyGen, Runway, Scenario, D-ID, Midjourney, Suno, Udio, and ElevenLabs. What truly deserves watching is not just funding and generation quality, but enterprise customers, retention, copyright compliance, workflow integration, and revenue per customer.

Fifth, the 5 points the market most easily misreads are: one, the biggest beneficiary of AI content is not necessarily the model maker but often the workflow platform; two, video generation does not equal video commercialization; three, copyright is a source of pricing power, not a drag; four, ad-AI monetization is mainly a budget take rate, not selling AI seats; five, many "AI content concept stocks" are really just internal efficiency gains, still far from an AI product that can be sold externally.

Sixth, the metrics most worth tracking over the next 6–12 months are: Firefly/GenStudio ARR and enterprise custom models; Canva/Figma AI ARPU and paying-user growth; Kling revenue and inference cost; Meitu's high-tier packages and tokenized revenue; Meta/Google/Kuaishou AI-creative penetration and ad ROI; UMG/WMG licensing progress with AI companies; Getty/Shutterstock data-licensing revenue and M&A integration; and the substantive changes in copyright, personality-rights, and deepfake regulation across regions.

Seventh, "AI content platform companies" mainly include Adobe, Canva, Figma, Meta, Google, Kuaishou, Publicis, Roblox, and Spotify/TME; "AI-native content challengers" mainly include Runway, Midjourney, Synthesia, HeyGen, Scenario, and Suno/Udio; "AI content shovel-sellers" include NVIDIA, cloud providers, the C2PA/Content Credentials ecosystem, and licensing/verification infrastructure like Getty/Shutterstock; and the legacy steps at higher risk of AI replacement include low-end design outsourcing, low-end stock assets, basic dubbing/captioning, localization outsourcing, and some ad production and game-art outsourcing.

Eighth, if I had to give just one narrower follow-up research direction, I would prioritize "the intersection of AI video generation and ad-creative automation." The reason is simple: this is where real budgets, the platform loop, quantifiable ROI, enterprise procurement, video-model cost reduction, and copyright-safety needs all come together, making it the most likely place for a dual re-rating of revenue and valuation over the next 12 months. Kuaishou Kling, Adobe Firefly/GenStudio, Meta Advantage+, Google AI Max, and Publicis CoreAI are the combination most worth following along earnings reports and customer cases in this direction.

Open questions and limitations: For private companies (especially Runway, Midjourney, ElevenLabs, Suno, HeyGen, and others), ARR, paying users, and gross-margin structure are mostly not fully disclosed; some listed companies still give only qualitative descriptions of AI-related revenue without a separate line item; and valuation data and the latest market consensus across regions lack a uniform, comparable, and freely verifiable public basis. Therefore, any valuation, private-company revenue, or regional data marked "needs further verification" in this report should, in the next research round, return to company-level materials, prospectuses, stock-price terminals, and industry databases for secondary due diligence.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.