Hololive operator Cover made 49.3B yen. Streaming was 18% of it

A cooking channel with 1.1 million subscribers earned 115,397 yen in a month. A gadget channel with 10,000 subscribers earned 165,945 yen. The subscriber counts are 110 times apart, and the smaller one takes home more.

The first reaction here was to distrust that ordering. One figure comes from the moment a channel was listed for sale, the other was published by the operator, so the two rulers may not match. The order of magnitude held anyway. Subscriber count explains almost nothing about what arrives in a given month.

What follows puts the bottom shelf and the top shelf in the same table. One listed company occupies the top, and it publishes its revenue breakdown every quarter.

The condition for getting on the shelf is public. The condition after that is not

The entrance to YouTube’s ad revenue share is defined numerically. A thousand subscribers plus 4,000 valid public watch hours over the past twelve months, or a thousand subscribers plus 10 million valid Shorts views over the past 90 days. Clear either one and an invitation arrives, then a channel-wide review decides whether the share begins. Review results usually take about a month.

Anyone can count up to that entrance. The divergence starts after it, because once the share begins, neither the rate per view nor the formula behind it is published anywhere. The actual amount stays invisible from outside until the person receiving it writes it down. Which leaves eight cases to line up.

Line up the eight and the middle sits at 224,686 yen a month

The median of the eight is 224,686 yen. The bottom two rows are overseas cases, and their order of magnitude runs straight off the top. Among the five rows where subscriber counts are known, the sequence 10,000, 10,000, 57,000, 1,100,000 and millions does not match the sequence of the amounts.

The firmest number belongs to an AI explainer channel. Two years to 36,500 subscribers, with total revenue across those two years in the high hundreds of thousands of yen. Convert 112 videos at 30 hours each and the hourly rate works out to 416 yen. The operator did not quit after stalling: they hit every target they had set, then acknowledged the ceiling on ad revenue.

The top shelf makes 81.5% of its revenue off the shelf

Cover Corp, which operates Hololive, reported 49.33 billion yen in revenue for the fiscal year ended March 2026. The breakdown looks like this.

Streaming itself accounts for 18.5%, and it was the only one of the four to shrink. Growth came from licensing at 25.3%, live at 18.7% and goods at 15.6%, all of which sit outside the stream.

This composition can be stated flatly. The company reached 49.3 billion yen not by raising what streaming brings in, but by moving the audience it gathered through streaming into forms that charge outside the stream. The top shelf does not earn on the shelf.

A person on 30,000 yen a month, one on $5.5M a year and a 49.3 billion yen company report the same mix

The breakdown from Shogo shows it at the smallest scale. Ad revenue at the 10,000-subscriber mark averaged about 30,000 yen a month, with nine months of actual monthly figures published. The conclusion reached was that advertising alone cannot support a living. The real engine was brand deals at 100,000 yen per video and own-product sales earning dozens of times the ad income.

At the bottom of the table, Ali Abdaal states the same thing as a ratio. Business revenue for 2023 came to over $5.5M, of which $3.5M came from courses. That division lands near 64%, and none of it is ad revenue or sponsorship. He also wrote that he had spent four years not paying proper attention to the platform hosting those courses, having judged it unimportant, and that he had been wrong.

A person on 30,000 yen a month, a person on over $5.5M a year, a company on 49.3 billion yen a year. Three rungs apart in scale, and the shape of the mix matches. The desk reads this as a point about sequence rather than about size. The bulk of revenue sits outside streaming at every rung, and the higher the rung, the larger that share. No row in this table shows the streaming share rising.

One person skipped the shelf and sold to the people standing on it

VMagicMirror is Windows software that moves a VRM avatar on a desktop screen without special hardware. Nearly every feature stays in the free version, and four years after release its cumulative sales on BOOTH passed 10 million yen. The 2,500 yen paid version works closer to a donation than a feature unlock. There is no advertising, no social media operation and no analytics, and the developer still holds a day job.

This person’s revenue has no relationship to their own view count. The structure sells more tools as more people start streaming. Leaving the shelf was never the only way to get off it.

Being off the shelf is not the safe position either

The counter-evidence comes first. Cover’s first quarter of fiscal 2027 brought revenue of 8.825 billion yen, down 8.3%, and operating profit of 651 million yen, down 33.0%. Merchandising fell 22.1%. The prior year saw net profit drop 45.7% after an impairment on metaverse-related software assets. In the same quarter live and events rose 109.6%, so the volatility shows up undisguised. Off the shelf the unit prices are higher, and inventory, facilities and the hit rate of individual projects come with them.

The figures also differ in kind. The 49.3 billion yen belongs to a company, not to the individual talents who stream. How it is divided has not been published. Three rows in the table come from the moment of listing on Rakko M&A, where the price quoted is the seller’s asking price and the closing figure stays private. And eight cases are eight out of the tens of millions of channels on the same shelf.

Nearby sit the medians by acquisition channel and 121 solo and side-project cases on the ceiling for one person. Definitions for monthly revenue and medians live in the data page.

Count the share of revenue unrelated to views, not the subscribers

The ceiling for one person is set less by the size of the market than by whether revenue could be detached from their own labour. The desk takes YouTube to be the shelf where that detachment is hardest. Views grow only alongside time, the rate is decided elsewhere, and the formula is not published.

That is why 1.1 million subscribers comes to 115,397 yen a month while 10,000 comes to 165,945. Neither operator got anything wrong. Subscriber count was simply never designed as a predictor of revenue.

Next time the counting happens, sorting by subscribers gets dropped. The sort worth running is by what percentage of revenue has nothing to do with view counts. Of the eight rows above, exactly one publishes that ratio as a figure. That one puts it at 64%.