China Built Microdrama. America Pays for It.

Microdrama is the vertical serial of one to two minute episodes, sold by the episode or by the week on a phone. China built it. CNSA, China's industry body, counted 128,000 put on air in China in the first quarter of 2026 and 239,000 in the second, most of them made with AI. Its president puts the total for January to August at 430,000, which he wrote was 13 times all of 2025. Outside China the apps took about $1.27B from viewers in the first half of 2026, and Chinese apps took 93% of it, by DataEye's estimate. The largest single paying market for that export is the United States.

This paper starts in China, because that is where the machine was built, and then follows the machine west. Four things matter to a distributor or rights owner.

America pays the most. The United States supplied about 28% of overseas app revenue in September 2026 from 9.7% of the downloads. ReelShort, the largest app, is projected to take 59% of its 2026 revenue from North America. Hollywood has started making microdramas, under a union contract written for them.

Europe is next, and smaller. Its largest market, the UK, had 8.2 million monthly users in 2025. Germany's downloads grew faster than any market AppsFlyer measured. European broadcasters are experimenting, mostly with small commissions of their own, and the apps we found usage data for in Europe are mostly Chinese.

AI arrived in the West in 2026. Owl & Co counted 3% of releases outside China as AI made in the first quarter and roughly two thirds in the third. At ReelShort it was 88% by September.

We found no published price for what a catalog is worth to these apps. Peacock, Fox, A+E, Fremantle, Banijay, All3Media and BuzzFeed all did vertical deals this year. None of the announcements we found disclosed a price. That gap is the finding, and the rights questions in section seven are why it matters.

What this paper does not do is tell you who wins the audience. The published viewing figures contradict each other, and the method section shows where.

Where it started: China built the machine

The format was industrialized in China. iiMedia Research put the Chinese microdrama market at RMB 67.79B in 2025, and the China Film Administration put the national box office at RMB 51.83B. On those two numbers, which measure different things, microdrama took about 1.3 times what every cinema in the country took. QuestMobile counted 851 million deduplicated short drama users in May 2026.

Then supply exploded. China's industry body, the China Netcasting Services Association (CNSA), began publishing a quarterly microdrama guide in 2026. In the first quarter about 128,000 microdramas went on air in China and about 122,000 of them, more than 95%, were made with AI. In the second quarter 239,000 went on air, 153,000 of them AI, an AI share of 64.0%. CNSA's president, Nie Chenxi, wrote that the total for January to August was 430,000, 13 times all of 2025, and the regulator's count, quoted in COL Group's third quarter report, reached 490,000 by the end of September against 33,000 for 2025. Neither says whether the 2025 count covers the same kinds of title, and the two sets of 2026 figures do not reconcile with each other (see the method), so read the jump as the scale these bodies report, not as a measured growth rate.

The AI share fell in the second quarter as live action launches rose. The same quarter brought money for live action: the regulator, NRTA, launched a support program on 26 May, and CNSA's guide counts six platforms planning at least RMB 6B for it. The published counts cannot show how much of the rise that money caused, and the first quarter live action figure is itself in doubt (see the method). AI volume did not fall. AI lowered production costs without making them trivial: a photoreal AI drama in China costs RMB 1,500 to 8,000 a minute at market rates, by an investigation published on The Paper, and Douyin guarantees RMB 4,000 to 6,000 a minute for premium photoreal AI. Labor, not compute, is about 70% of the cost of a serious AI drama. NRTA's Order No. 16, China's first rule written for microdrama, took effect on 1 September 2026 with AI labels, and an NRTA meeting drew what Nie called a clear red line: synthesizing a person's face or voice requires the rights holder's authorization.

What the machine makes is narrow. CNSA's second quarter guide tags the hit lists of the main Chinese platforms. Among 173 live action hits with episodes under three minutes, 69.9% carried the tag "modern" and 53.8% the tag for humiliating a villain who had it coming. The longer episode hits, three to twenty minutes, were half romance, and 58.0% ran exactly 24 episodes. Of 186 AI hits, 89.8% were photoreal, made to look like live action. Abroad, the overseas titles with the heaviest ad spend shared the same top tags, emotion, comeback, revenge and CEO, and one tag belonged to AI alone: werewolves.

For a Western catalog owner the longer episode list is the one to read. Romance and suspense in 24 parts looks much more like a TV library than the three minute revenge plots do, and it is the door the library deals in section six are walking through.

Then the machine turned outward. DataEye estimates that Chinese apps took about 93% of overseas app revenue in the first half of 2026, down from 95% in 2025. The top three apps took over 40% of it and the top twenty 83%. By the same estimate, the apps that are not Chinese owned, from Ukraine's My Drama to Israel's Shortical, shared the remaining 7%, about $70M. The rest of this paper follows that money.

America pays: 28% of the money from 9.7% of the downloads

Outside China there is no official count, only trackers, and they do not agree. For 2025 the measured figures for app revenue outside China run from $2.33B, in an industry report released at a CNSA congress, to about $3.0B, in Sensor Tower's app store data and in Axios. Forecasts made during the year ran to $3.8B at Deloitte and Soochow Securities and above $4.5B in China's Economic Information Daily. For 2026 the two forecasts we found disagree by more than two to one.

Whichever tracker you pick, the United States is the largest paying market. DataEye's monthly estimates put US in app revenue at more than $76M in May 2026, more than $84M in August and nearly $80M in September, between 28% and 33% of all overseas app revenue. In August the next four markets, Japan, South Korea, the UK and Australia, came to 19.3% between them. And US revenue runs well ahead of US downloads. In September the US supplied 9.7% of downloads and about 28% of the revenue, while Indonesia, India and Brazil each supplied more of the downloads than the US did. Downloads are not people, so this says where the money is, not how many Americans watch.

The forecasts point the same way. Omdia expects the US to account for $1.5B of microdrama revenue in 2026, half of all revenue outside China, and nearly $2B in 2027. Media Partners Asia projects US revenue rising from $1.5B to $3.7B over its five year forecast. Appfigures already had US consumer spending above $1.1B for 2025.

ReelShort is the clearest case of a Chinese built app living on American money. Media Partners Asia, as relayed by Indian Television, estimates its revenue at $97M in 2023, $400M in 2024 and $785M in 2025, forecasts $1.05B for 2026, and projects $620M of that, 59%, from North America. A 36Kr report citing Sensor Tower data put the US at nearly 62% of ReelShort's cumulative revenue. The audited numbers tell the same story in yuan. Crazy Maple Studio, which runs ReelShort and reports as an associate in COL Group's accounts, grew revenue from RMB 2.91B to RMB 5.72B in 2025 and went from a RMB 7.9M profit to an RMB 85.8M loss. Media Partners Asia puts the 2025 loss at about $12M and expects a profit in 2026.

The American model is a paywall after a few free episodes. Reuters describes weekly subscriptions at around $8 to $20; TheWrap found North American subscriptions starting at $19.99 a week and $269 a year. Naavik puts day one retention at ReelShort and DramaBox at about 27%, falling to about 2% by day 30, and US revenue per daily active user down from early peaks above $2.50 to well under $0.50. Paying for new viewers is a heavy, recurring cost. ReelShort spent 55% of its 2025 revenue on user acquisition and marketing, by Media Partners Asia's estimate, and every point it cuts is worth about $10.5M of EBITDA. It is also moving billing to its own web store, away from app stores that take up to 30%.

The viewers are bought on social media, and Meta is the biggest channel. Digiday, citing eMarketer and Sensor Tower, found that 68% of US microdrama app ad spending in January to September 2025 went to social media: 25% to Facebook, 19% to TikTok, 16% to Snapchat and 8% to Instagram. Business Insider reported, in a roundup by Real Reel, that some operators put as much as 75% of their marketing budget into Meta. CNSA's second quarter guide shows the same thing from the other side: every one of the overseas titles with the heaviest ad spend ran on Facebook, its Audience Network, Instagram and Messenger.

How many Americans watch is less settled than how much they pay. Omdia counted 66 million monthly users in the US in 2025, more than double 2024, and found men are half of them, against the format's reputation. EMARKETER forecasts 5 million US microdrama viewers in 2026. These measure different things, and the method section lays them side by side rather than choosing.

Hollywood starts making them

The format now has its own union contract. SAG-AFTRA's Verticals Agreement caps budgets at $300,000 and shoots at 30 consecutive days, with a lead minimum of $250 for an eight hour day; the union's 2026 television agreement pays a day player $1,246. Most microdramas cost between $150,000 and $300,000, by TheWrap's count, and a typical production shoots in seven to ten days, by one producer quoted in the Producers Guild's magazine. The union will say only that it has signed off on dozens of productions. ReelShort and CandyJar now make some union verticals, but they are not the bulk of either app's content.

Los Angeles is trying to keep the work. In January the City Council backed a search for $5M to fund a 40% rebate, because microdrama budgets, typically under $200,000, often fall below the threshold for California's tax credits, and in a survey the council's release cites, 78% of local independent producers now shoot these projects outside the city. FilmLA publishes no microdrama count at all. CandyJar, two years old, is on track for more than $100M in annual sales, and one operator now runs three facilities of standing sets bought at liquidation prices from studios including Sony Pictures and Paramount.

The studios came in this year, mostly through the side door: licensing a few titles, testing a feed, or taking a stake in someone else's app.

Venture and growth money followed, and much of it is money for buying viewers. GammaTime raised a $14M seed in October 2025 and sold Versant a minority stake in June 2026. StoReel raised $34M, of which $25M is user growth financing from PvX Partners, and PvX has put $100M into Shortical on terms described as non dilutive and performance based. Not every app lasts: VeYou said in October it would close on 6 November, about six months after launch.

Europe: smaller, later, and mostly Chinese apps

Europe is the next market and a much smaller one. Sensor Tower put Europe at 11% of global short drama in app revenue in the first quarter of 2025 and 7% of downloads. Adjust found short drama took 4% of European video entertainment time in the fourth quarter of 2025, against 5% in North America, and its revenue per monthly user in Europe fell from $2.05 in 2025 to $0.24 in the first quarter of 2026. By Omdia's count the UK, Europe's largest market, had 8.2 million monthly users in 2025 and Germany 4.4 million, against 66 million in the US.

In money the UK is the only European country in the top five. DataEye had it fourth in August 2026, with about 4.2% of overseas app revenue, and fifth for the first half. Downloads are where Europe stands out, at least in Germany. AppsFlyer measured Germany's short drama downloads up 210% on a year earlier, the highest of any market it tracks, with France up 87%, the UK 36% and Spain 23%. Across all markets, more than 60% of short drama installs now come from paid channels.

Where European usage is measured, the apps are mostly Chinese. In Spain, GECA's OTT barometer put DramaBox at 34.3%, ReelShort at 27.5% and NetShort at 22.4% in its platform usage shares. The two European owned apps in DataEye's overseas revenue top 20, CandyJarTV, which DataEye lists as German, and My Drama, from Ukraine's Holywater, ranked 16th and 17th in August 2026, both with revenue down on a year earlier. German speakers mostly get English originals with subtitles or what OMR Daily calls sloppy AI dubbing, at up to €19.99 a week on ReelShort, or about €40 to finish an 80 episode series bought by the episode.

Europe's broadcasters are experimenting, mostly with small commissions of their own, a few recuts of their libraries, and mostly on social platforms or their own streamers rather than coin apps.

European startups are trying to build local supply, and the first one to try a library ran into the wall this paper is about. Eilin, a Munich app announced at the Berlinale in 2026, planned to buy existing content and recut it for vertical. That plan failed, and it now makes originals, on budgets of €150,000 to €250,000.

Others are trying the same route. In the UK, Tattle TV says it is exploring reworking existing UK intellectual property for vertical viewing; in Spain, Shorta launched in September combining originals with series acquired from local producers. Meanwhile Chinese catalogs are being sold into Europe the conventional way. COL Group, whose associate Crazy Maple runs ReelShort, appointed Harbour Rights in February to distribute more than 1,000 vertical titles in Europe and Latin America through output deals, and by MIP London its slate had passed 1,700 titles. COL told The Drop that the UK could be one of its most important markets and Germany "a surprisingly strong market", and that it is in talks with broadcasters and OTT platforms in several European markets.

Europe's media rules have not caught up with the apps. The AVMS Directive defines a program "irrespective of its length", so a two minute episode is not excluded by duration, and on demand services under EU jurisdiction must hold at least 30% European works in their catalogs. But none of the eight main apps we checked, ReelShort, DramaBox, ShortMax, My Drama, GoodShort, FlexTV, DramaWave and NetShort, appears on Ofcom's list of notified on demand services, and the European Audiovisual Observatory's MAVISE database returns no data for any of them. Absence from a list is not a legal finding: only services under UK jurisdiction that meet the definition notify Ofcom. The Commission's quota guidelines count one season of a series as one title and predate microdrama entirely. The only enforcement we found was an advertising ruling: the UK's ASA banned a DramaWave ad, seen in the Vinted app, as likely to cause serious and widespread offense and irresponsibly targeted.

The AI wave reaches the West

In China the AI share peaked in the first quarter. Outside China it is arriving now. Owl & Co's "What We Watched on Vertical" report, as reported by The Hollywood Reporter and relayed by Let's Data Science, counted AI made titles at 3% of vertical drama releases outside China in the first quarter of 2026, 28% in the second and roughly two thirds in the third, across 20 top platforms. At ReelShort, GASP data relayed by DataEye puts the AI share of new content at 10% in May and 88% in September, while live action output held steady at 10 to 30 titles a month.

The switch shows across the apps. The Hollywood Reporter, as relayed by Project Casting, reported that DramaBox, DramaWave and FlareFlow have increasingly used synthetic actors and settings, and that My Muse and Inkitt's Ironblood release AI verticals exclusively. FlexTV halted all traditionally shot productions early in 2026. Kunlun's half year report put the AI share of new episodes on its short drama apps, led by DramaWave, above 90% in July. My Drama's head of studio says AI is used at every stage, from script writing to localization, with human actors kept for the leads, and Holywater bought Jeynix, a face replacement and lip sync company, for "rapid localization across markets in a matter of days".

Where the switch is explained, cost is the explanation given: by Media Partners Asia writing on ReelShort, by Mega Matrix, by Holywater and by a FlexTV executive. The table keeps the published figures apart rather than averaging them, because they measure different things.

Performers report the effect. A talent manager told The Hollywood Reporter, as relayed by Project Casting, that from June the number of vertical shows casting each day fell from 15 to 20 to maybe four. Business Insider, relayed by GadgetReview on Yahoo, reported AI remakes of live action series distributed by MoboReels, GoodShort and ReelShort; one actor was paid about $5,000 for an original role and nothing for the remake that used his likeness. In Germany, hundreds of voice actors have refused Netflix contracts that would let it train AI on their voices, by the count of their association, VDS.

The door is open. The price is not on it.

Catalogs are moving into the format and out of it, mostly since late 2025, and every deal we found has the same hole in it.

The nearest thing to a statement of terms comes from Banijay Rights, which names the structures and no numbers.

RoseBerry's founders told the same outlet that rights, music, talent and residuals "don't just carry over" and have to be cleared again for the new format and its windows. A+E says it now writes repurposing rights into contracts at the start. And Eilin's plan to license and recut a library failed, by OMR Daily's account, on the complexity of licensing.

What platforms do publish is revenue share for content made for them, almost all of it in China. It is the nearest published benchmark we found for a rights owner, and it is not a license fee.

That last point is the one to hold on to. An AI series can be made from $50 a minute in the US, and a Chinese platform guarantees RMB 4,000 to 6,000 a minute for premium AI drama. That makes AI production a competing source of supply, priced in the open, which a buyer can set against a library in a negotiation. Production cost does not settle what a library is worth: recognized titles, exclusivity and an audience that already knows them are things AI cannot make to order, and neither is a clean chain of rights, which is the subject of the next section.

The rights questions a license now has to answer

Courts, regulators and unions have started answering microdrama rights questions, in the US and Europe and in China, which got there first. The table pairs each question with what the record shows and what that means for the paper you sign. It is a reading of public sources for people who negotiate licenses, not legal advice.

Every row in that table is a clause someone has to track for the life of a license, across territories, platforms and revenue share terms that, as section six shows, change by announcement. That is the work a rights system exists for.

Method, and what this data cannot tell you

What this is. A reading of public sources in English, Chinese, German, French, Italian, Spanish and Ukrainian, not a dataset. Every figure is taken from a company filing, an official guide or register, a regulator's figure as quoted by a named outlet, or a named research firm, and each was checked against the quoted sentence in the source at the time of writing. Where an outlet relays another body's figure, the outlet is named, and so is the body behind it. We did not collect app catalogs: the terms of every major app forbid automated collection.

Currency. Figures are printed in the currency the source used. We do not convert.

Conflicts we did not resolve.

• Who wins the audience. CNSA's first quarter guide, as reported by Economic Daily, says live action drew 25 times the plays of AI over the 2026 Spring Festival; Huashang Bao, as carried by The Paper, reported that AI manju took nearly 30% of 8.67 billion short drama plays over the same holiday. In the US, Holywater and Owl & Co report that its AI app converts viewers to paying subscribers at 23.6% against 23.9% for its live action app, and TheWrap reports average day seven retention 22% higher for AI titles than for comparable live action ones. In Germany the producer Red Pony concluded that AI series are "not really for the audience at the moment" (our translation). None of these publishes a common base. This paper makes no claim either way.
• How many Americans watch. Omdia counted 66 million US monthly users in 2025. EMARKETER forecasts 5 million US viewers in 2026, and puts ReelShort, via Similarweb, at 1.2 million US monthly active mobile users. They count different things.
• The US share. DataEye has the US at 28% to 33% of overseas app revenue from May to September 2026. Sensor Tower had it at 49% of global in app spending in the first quarter of 2025, and Omdia forecasts half of revenue outside China in 2026. DataEye's own US share of downloads is 4% for the first half of 2026 and 9.7% for September. Different scopes and periods; we never pool them.
• ReelShort's revenue. Media Partners Asia estimates $785M for 2025; Appfigures, via TechCrunch, counts about $1.2B of gross consumer spending; COL's filing gives RMB 5.72B for Crazy Maple Studio. Gross spending, net revenue and a company total are different measures. Media Partners Asia's target year for cutting marketing to 44% of revenue is 2028 in Indian Television's account and 2027 in Screen Daily's.
• Holywater. Forbes Ukraine reported more than $20M of revenue a year in December 2024; the company says revenue about doubled in 2025; DataEye shows My Drama's app store revenue down more than 30% on a year earlier in August 2026. DataEye counts in app purchases only, not web payments.
• Europe's size. Sensor Tower's 11% of global in app revenue for one quarter of 2025 is several times larger than the roughly $52.65M for all of 2024 that Diandian Data reported, via the payments firm Antom. Their definitions of Europe evidently differ.
• Meta's share. Up to 75% of some operators' marketing budgets, per Business Insider via Real Reel; 33% of US app ad spending for Facebook and Instagram together, per Digiday.
• US budgets. Los Angeles' council release says microdrama budgets are typically under $200,000; TheWrap puts most at $150,000 to $300,000.
• China's launch counts. Nie Chenxi wrote that over 90% of the 430,000 launches in January to August were AI. CNSA's two quarterly guides give 275,000 AI of 367,000 launches in the first half, about 74.9%, which leaves about 63,000 launches for July and August. Even if every one of those were AI, the January to August share could reach only about 78.6%, not 90%. The figures are likely different counts, with different coverage or definitions of a title. Nie does not state the 2025 base behind "13 times", and the 33,000 for 2025 quoted by COL Group may count differently again. We print them as each body published them and compute no growth rate from them.
• Live action in China's first quarter. The quarterly guides imply about 6,000 live action launches in the first quarter against 86,000 in the second, while the Hengdian studios, per the second quarter guide, hosted about 300 crews a month in January to May 2026 against 335 in 2025. We print the counts as published and make no claim about why live action rose.
• Overseas revenue for 2025. Measured figures run from $2.33B to about $3.0B; forecasts reached above $4.5B. They are shown side by side in the trackers chart, not averaged.

Gaps. We found no published license price for catalog content, which is the finding; it does not prove that none exists, since terms can sit in private contracts. No revenue figure in currency for any single European country, only ranks and shares. No count of AI made titles for Europe. Every Omdia figure here is relayed by another outlet, because Omdia's own pages refused our requests, and Owl & Co's and Media Partners Asia's reports were read through the outlets that carried them. No audited numbers for DramaBox, CandyJar or most of the largest apps, which are private.

Derived figures. Our arithmetic, labeled where it appears: the next four markets' 19.3% combined share in August 2026; Facebook and Instagram's 33% together; the 1.3 times ratio of microdrama to box office in China; the 74.9% half year AI share; the 63,000 July and August launches and the 78.6% ceiling they allow; the 6,000 live action launches in China's first quarter; the 67% at which "roughly two thirds" is drawn. Our reading of the published rankings: the UK as the only European country in the top five.