Owl & Co published its first report on the vertical video economy this month. The headline figure is $150 billion in revenue outside China in 2026, up 42% in a year. The composition is more interesting than the size. Meta, ByteDance and YouTube take 94% of it. Advertising accounts for $131 billion, user payments for $10 billion and commerce commissions for $9 billion. The micro-drama apps most people picture when they hear "vertical video" — ReelShort and the nearly 2,000 others competing for audiences worldwide and region by region — are worth about $4 billion, under 3% of the market.
What the report has sized, in other words, is mostly an advertising business in portrait orientation. Screen orientation is the only property every item in the $150 billion shares. The economics underneath differ by format, and they differ most sharply on a single line: what it costs to get someone to watch.
Media Partners Asia, cited by Owl & Co, estimates that ReelShort — the largest micro-drama app by revenue — spends about 73% of its revenue on user acquisition and app store fees. Smaller apps are probably spending a higher share. All of them spend noticeably less on making the shows. That ratio is the category in one number: these are marketing operations with a content cost attached.
Viewers have been voting on the model. The share of free, ad-supported apps in total micro-drama watch time rose from 15% to 83% in six quarters.
Paid apps behave like mobile games rather than like television. They buy an install, hold the viewer with cliffhangers, then pay again for the next cohort. Meta, ByteDance and YouTube begin every series with the audience already in the building.
Which brings up the finding in the report that should trouble anyone selling generative AI as the fix for production economics. Across leading paid apps, the number of new series in Q2 2026 rose 25% while total watch time fell 4%. Releases went from 1,659 in Q1 2025 to 5,145. Owl & Co attributes much of that acceleration to AI-generated series. On ReelShort, new live-action series released in June and July averaged 4.17 million views against 2.74 million for AI-generated ones.
Cheaper output raises supply. It does not create demand, and on this evidence it does not even hold the demand that already existed.
The streaming services entering the format are not playing the same game. Peacock has ordered original micro-dramas based on Bravo properties for its mobile app. ViX released 145 original vertical series in its first year. Owl & Co names JioHotstar among the fastest-moving players. The report says these services are chasing engagement and lower churn rather than a per-episode payment. Character.ai is testing a third variant: vertical series built around characters users can keep talking to after the episode ends. A Jamie Oliver comedy funded by Life360 runs through channels his business already operates, with the brand paying for access to an audience that exists whether or not the show does.
Different businesses, one shared advantage: the relationship with the viewer predates the promotion of the next series.
Owl & Co's own numbers show the same advantage at the top of the market. The report puts Meta's 2026 revenue at $253 billion, with vertical video at 27% of it. Do the arithmetic and that is roughly $68 billion from vertical video alone, against the $52 billion of total Netflix revenue the same report estimates. Meta's portrait-orientation slice is larger than Netflix.
Here is the figure I would want and the report, at least as presented, does not give: how much of that $131 billion advertising line is micro-drama apps buying their own installs. At a 73% acquisition ratio for the largest of them, it cannot be a rounding error, and it would mean a good deal of the same money is being counted twice — once as an app's cost of doing business, once as a platform's revenue. A market where the challengers' main expense is the incumbents' income is not a market that gets disrupted from below.
The binding constraint in this category was never the cost of making the next episode. It was the cost of finding someone to watch it. Generative AI attacks the constraint that was not binding, and by flooding the feed it makes the one that is binding worse: when 1,000 more series become possible, attention gets more expensive, not less. The scarce goods remain a reason to watch this particular story and a distribution point that does not charge you twice for the same audience — and the three companies holding 94% of the market are the ones already selling both.