A Licence Buys You 49.8 Million Subscribers: What's Left of Television

A city park lawn at dusk: in the near foreground a small rain puddle mirrors the whole open sky and a distant broadcast tower, an open laptop glows on the grass beside it, and the eye travels back across the lawn to a far city skyline

You cannot speak of the sea to a well-frog — it is bound by its hole.
You cannot speak of ice to a summer insect — it is bound by its season.
You cannot speak of the Way to a narrow scholar — he is bound by his doctrine.
—— Zhuangzi, “Autumn Floods” (c. 3rd century BC)

Zhuangzi never calls the frog stupid. He says it is bound by its hole. The summer insect isn’t lazy either — it simply lives in a season that has no ice.

This episode is about three such holes: the one a producer’s eighteen years of habit dug for him, the one an industry’s own economics dug for it, and the one all of us dig a little deeper every day — the algorithm.

What this episode is about

Investor’s Sunday is a Japanese finance radio programme. This is the second half of its conversation with Hiroki Takahashi — producer of the video outlet ReHacQ, CEO of Tonari Inc., eighteen years at TV Tokyo, and the man who has taken a channel from zero to a million subscribers twice. (I wrote up part one separately.)

Part one was about leaving the network. This one is about what came after: how the company actually runs, where the money comes from, what television still has that nothing else does, and what he wants to build over the next five years.

The tone is loose — it opens with him half-joking that he’d like a show on interfm playing nothing but nineties music, and the host saying, on air, fine, let’s do it. But there is hard material in here. The three minutes where he explains why broadcasters are strong is, I think, the most valuable thing in the episode.

Original episode: Investor’s Sunday, 2 August 2026 (Japanese, ~25 min)

The notes I took

He holds three roles at once and isn’t entirely sure how they fit together. Tonari is his own company, so there he is the owner. He is also a salaried employee at CyberAgent, carrying a General Producer title on the ABEMA side — roughly a specialist department-head grade. The odd on-camera appearance he takes personally. He jokes that he isn’t confident his social insurance is being filed correctly. Tonari is about fifteen people and does corporate video alongside ReHacQ — including, he notes, an onboarding film for an insurance company, which obviously could not be shot like a dating reality show.

Half of running a company is not the work you’re good at. Inside a network, short of staff, HR moves someone; short of gear, someone procures it; short of budget, you apply. On your own it is all yours: writing the hiring page, interviewing, placing orders, checking attendance, answering mail. His current headache is Wi-Fi — live streams kept failing, and it turned out the office was on a consumer connection. A business line takes two weeks of installation, so to stay online in the gap he had to sign a second contract and cancel it a week later. “All of that, you do yourself.”

He would rather not talk to people. The job is talking, so when he isn’t recording he’d like nobody knocking. The office faces Shinjuku Gyoen, and he takes remote meetings lying on the lawn. In the early evening he walks through Nichōme — by then, he says, it’s less a night out than watching the species, with every smell in the world layered on top of every other.

Radio was the medium he knew least, and it turned out to be the peaceful one. TV Tokyo had no radio arm, so he’d never been curious. Doing it, he found a different audience, a weaker response — and, crucially, no live number chasing him. Ratings are measured once every two months; most weeks nobody is counting at all. YouTube and social, by contrast, feel “like being at war every day”: you can see the comments, you can see your rivals, and the algorithm is built to be violent about it. He says it gives him stomach aches.

That you can leave a network and still make video is, to his generation, a miracle. This was the most sincere passage in the episode. There was a time when only broadcasters could fund variety or drama, so quitting the network meant quitting the craft — client work at best. Now not only is it possible, there’s a system that monetises it, and he credits that squarely to Google. “Anyone older than me would find this unbelievable.”

A television studio is extraordinary, and he thinks the cost-performance is bad. Twenty people in the studio, thirty or forty once you count everyone behind it, to produce sixty minutes. Ceiling height, lighting, sets, cameras — none of it is in the same category as web production. Genuinely impressive. But he thinks the next marginal budget will go into content rather than headcount, because some ways of spending move the numbers and some don’t.

Television’s last moat is the set that’s on while the family eats. He thinks the case for watching live has essentially gone, and content will keep drifting toward something you pull down whenever you want. But there is one thing the web can’t reproduce: on in the background, nobody watching very hard, a family half-talking and half-glancing. The decisive demographic, he argues, is households with children between three and nine or ten — before cram school starts — because adults can’t sit through what kids watch on YouTube, while a variety show or an animal programme works for both at once. That culture, he says, is worth keeping.

And one more thing the web can’t do: things you didn’t know about arrive by themselves. On YouTube you search, and then the algorithm makes you narrower. The host adds a line that landed harder than he probably meant it to — that this way, you understand less and less of the world. He’d gone all-in on YouTube for a while himself, and has drifted back toward television.

Real mass is a licence, and only broadcasters have one. This is the hardest-edged part of the episode. A Kantō-wide licence, translated into YouTube terms, means you start with roughly 49.8 million subscribers — and there are only five channels in the box. He has fought his way to nearly two million while competing with something like a million other channels. It is simply a different game. A web outlet can, he thinks, get a single programme to the impact of a network’s flagship business show; the aggregate weight of prime time is another matter. And that, he says, is worth preserving.

Regional broadcasters and local radio are under real pressure; the exits are consolidation or partnership. He notices the movement starting with the Kansai semi-key stations and the Tōkai region — Nagoya, Gifu, Mie — because the Tokyo key stations still have slack. The pattern is co-developing IP with people from the online world and running it on both. ReHacQ itself co-financed a dating reality format with ABEMA, with the main show on both sides and a companion “watch-along” strand of their own. Local radio, meanwhile, still owns the evening commute — a genuinely loyal audience and the cash cow of those stations — but the cost base is high and he expects it to shrink.

Revenue comes in three lines, and the biggest isn’t the visible one. Platform ad share; brand sponsorship (less of it, he says, than the comparable business-media outlets); and third, production commissions that arrive because someone saw ReHacQ — make us a programme, make us a corporate film. That third line, he suspects, may be the largest.

Is platform monetisation getting worse? He doesn’t especially feel it — but it depends how you play. The rules aren’t published, whether you’re judged on views or watch time is never stated, and the platform adjusts periodically; people who get adjusted against lose income. His read is that YouTube keeps steering toward the straight road, optimising for long-term quality, which makes growth-by-trick progressively harder. Book good guests, let them finish their sentences, make something that doesn’t exist elsewhere, and the changes barely touch you. That is what he does.

The five-year list: win the commissions, and do more reporting. The first he calls an urgent wish. The second is the interesting one — the company has a war photographer, working around Lebanon and Gaza, and on the ground at Henoko in Okinawa. Reporting like that does not pay for itself at web economics, and he reaches for a good analogy: the UNIQLO store in Ginza. Who knows whether that shop makes money — it’s the flagship, you do it anyway.

What’s exploding right now is science. This was my favourite passage. Physics travels unusually well: a physicist based in the US pulls enormous numbers, and even a Waseda engineering researcher’s episode ran to the point where a million views looked plausible. His explanation is clean — the lineage running from Japan’s earlier video culture into YouTube was humanities and social science, so the sciences were never mined; and parents are happy to let children watch. He asked the physicist to explain relativity in a way a child could follow, relayed it to his own kid, and got back: isn’t this just the Urashima Tarō story?

Business video is a war of all against all, but each outlet has a colour. The comparable outlets and the broadcasters’ own business channels are all in the water now. Yet what breaks out differs by house: one gets its hits from AI guests, another from football, stocks and health, and his own from science, history and politics. “Everyone does economics.” Next he wants to build IP — a format he co-financed can be licensed abroad and extended.

At the end the host jokes about investing in him. He deflects, smiling: people who put money in also put opinions in — sell the property, that sort of thing. Then the line of the episode: “That’s exactly what I quit the company to get away from.”

What I took away

1. The same growth rate means different things depending on where you’re standing.

“A licence is 49.8 million subscribers before you start, and there are only five channels in the box” is the most concrete description of structural advantage I’ve heard in a while. It isn’t better content or a stronger team. The denominator is capped by law.

Applied to companies, this is the thing a headline growth rate hides most easily. Two firms both growing thirty percent: one sits in a protected position where the number of competitors is fixed, the other is fighting a million others in an open field. The first thirty percent extrapolates; the second can be eaten by whoever shows up next. The premium belongs to the position, not the speed.

There’s a second half to that, though. Scarcity is what makes a position valuable, so the live question is whether the scarcity still holds. Five channels were worth something because five was the ceiling. He concedes himself that the case for watching live has gone — a licence caps channels, not attention.

2. The gap is where nobody has dug, not where everybody is standing.

Business and finance video is saturated; science was untouched and detonated on contact. The structure is unusually clean: demand was always there — parents actively prefer it — while supply stayed absent for reasons of historical path, not economics.

This is the same shape as hunting for a bottleneck in a supply chain. Crowds form on the demand side; the money sits where supply is scarce. On a subject everyone covers, doing it better mostly just compresses your own margin. Pricing power lives where something is genuinely wanted and nobody is making it. So the more useful question isn’t “is this theme hot” but “if demand doubled, what runs out first.”

3. Rules you can’t price are rules you shouldn’t build on.

The platform doesn’t publish its criteria and changes them. He barely notices; people optimising against the old criteria get wiped. His defence isn’t guessing the rules — it’s doing work that still counts whichever way they move: good guests, full conversations, material that exists nowhere else.

That generalises to any business leaning on a single channel: search rankings, platform revenue share, one customer’s procurement policy, one country’s subsidy. Those are dials in somebody else’s hand, and you can’t even read the current setting. So the question to ask isn’t “what’s the rate today” — it’s “if the rule flips to the version that hurts, what is left of this business?” If the answer is “nothing,” it isn’t a compounder. It’s a position.

4. If every input was searched for, your “multiple sources” may be one source counted several times.

“Things you didn’t know about arrive by themselves” against “the algorithm makes you narrower” is the well-frog line in modern dress. The frog isn’t stupid. It has simply never had the chance to meet the sea.

The damage this does to investing is specific. Once you hold a view, what you search for, what gets recommended, and who agrees in the replies all point the same way. You feel like you checked five sources; you heard one judgement repeated five times. Separating noise from structure requires that your inputs contain things you disagree with.

Two practical moves. Keep one deliberately unchosen channel open — something that arrives whether or not you asked for it. And before committing, write down the strongest version of the opposing case; if you can’t write it, you haven’t read enough yet.

5. Accumulated numbers are not a pass. Every round is judged again.

He says the difference between one million subscribers and three million is smaller than people think, because what gets judged is whether this particular video is any good. Subscribers who don’t get served, or don’t click, count for nothing.

The parallel is direct. Market share, brand, installed base, cumulative users — stock metrics get treated as proof of a moat when they are really just last period’s result. What deserves the attention is whatever has to be regenerated every period: retention, renewal rates, whether new customers are getting more expensive to acquire. A moat isn’t a noun. It’s a hypothesis you re-test every quarter.

And re-testing only works if the line is drawn first: which number, falling to what level, means I was wrong. Without that line, every piece of bad news can be filed as temporary.

6. You’re allowed positions that don’t pay for themselves — if you said so upfront and set the ceiling.

Frontline reporting doesn’t clear web economics and he does it anyway, on flagship-store logic. The framework is sound: a portfolio can hold things whose job is information, or optionality, or simply keeping you at the table where you can see what others can’t.

But there’s a fine line here. The difference between a flagship and a slow bleed is not whether it loses money — it’s whether you declared upfront that it wasn’t meant to earn, and where you stop. Saying “this is the flagship, the budget is this much” is strategy. Discovering months later that it has been bleeding and you can’t bring yourself to close it is a loss being retroactively promoted to a strategy.

The same test applies to your own time. Hours spent learning, experimenting, writing shouldn’t be judged on short-term return — but the allocation has to be stated, or it quietly eats the main job. As he puts it, half of a founder’s week goes on working out who to buy the pens from.

What this episode is about

The notes I took

What I took away

Further reading