Korean drama costs rose 27-fold. The chart concentration the argument predicts did not appear

The number doing the work in the current argument about Korean drama is a cost. Around $360,000 an episode in 2015; roughly $9.8m an episode for the second season of Squid Game in 2024. A 27-fold rise in under a decade.

Those are not our figures. We hold no budgets, no licence fees and no commissioning terms, and nothing on this page changes that. They are here because of what people build on top of them: if a Korean series now costs that much, only a few buyers on earth can pay for one, and the business should be concentrating into fewer hands.

That second part is not a claim about money. It is a claim about who ends up on the shelf — and that we can measure. So we did.

What we counted

Every Korean series that reached any country’s Netflix top 10 between July 2021 and July 2026: 493,600 chart rows across 93 markets, each title joined to the companies Wikidata credits as its producer, first broadcaster or distributor. Then, year by year, one question: how few companies does it take to cover half of that year’s titles?

If leverage concentrated, that number should be falling. The full table, every year including the part years, is at did the charts concentrate?.

It is not falling

Three companies covering half of 107 titles is an extraordinary degree of concentration — the right-hand column is what an even spread would need, and it is 28. But the extraordinary part was already there in 2022, and it is the same three-against-26 in 2025. Across the four full years the figure does not move.

That column has a defect worth saying out loud: it can only change one whole company at a time. A measure that moves in steps of one is too blunt to read a four-year trend from, and quoting 89.3% against 88.5% off the back of it would be dressing up a rounding artefact as a finding.

So here is the same question asked with a measure that moves continuously — the share of each year’s titles covered by the three largest companies, divided by what those same three would cover if every company held an equal number:

The three largest companies cover roughly ten times what an even split would give them. The Korean chart is heavily concentrated, and it always has been in the years we can see.

But the direction is down, not up. 11.56× in 2022 against 9.81× in 2025. Whatever the 27-fold cost rise did, it did not show up here as the charts closing around fewer companies.

What we are not telling you, and why

One other thing moved, and we are not reporting it as a result.

The share of charting titles that passed through a broadcaster falls from 88.8% in 2022 to 80.0% in 2025 — the shape you would expect if commissioning were shifting away from Korean networks. It is the more interesting number on this page and we are leaving it on the floor, because recorded credits per title fall across exactly the same years, from 1.85 to 1.60. In our data a title that never had a broadcaster and a title whose broadcaster nobody has typed into Wikidata yet look identical.

The test for that is obvious: repeat the measurement only on titles whose entry is demonstrably filled in. We ran it. It leaves at most 23 titles in a year, and 7 in 2025 — one title moves that figure by fourteen points, and the decline we are testing is under nine. The control is weaker than the effect, which means it settles nothing in either direction.

So the honest report is: the broadcast share may have fallen, or Wikidata may have got slower. We cannot tell which, and a number we cannot tell apart from a recording lag is not evidence.

We have been caught by this before, and that time we could prove it: a clean 69% decline in new K-pop group debuts turned out to be the speed at which Wikidata records new groups — the same decline showed up for American, Japanese and British groups. The difference is only that we could run the control there and cannot here.

What this cannot answer

It cannot tell you what anyone paid. Netflix publishes chart positions and publishes no money at all, so nothing here speaks to whether $9.8m an episode is sustainable, who absorbed it, or what terms came with it.

It also cannot tell you that concentration is absent — three companies covering ten times an even share is concentration by any reading. The finding is narrower and duller than that: over the four years we can measure, it did not increase. An argument that runs “costs rose, therefore the business concentrated” has a step in it that our charts do not support.

And a company appearing on a chart is not the same as a company holding the rights, keeping the upside, or having any say in what gets made next. Those are the questions the cost figure is really about, and they are not in this data or in any data Netflix publishes.