Production costs are reported to have risen roughly 27-fold in under a decade, and the argument that follows is that only a few buyers can now finance a Korean series. If that shifted power, the charts should show it. Did they?
Around $360,000 per episode in 2015 against roughly $9.8m per episode for Squid Game season 2 in 2024 — a 27-fold rise in under a decade.
Whose number. Reported by others. We did not measure it and do not hold production budgets of any kind. It is the claim this page tests the consequence of, not evidence this page provides. Every other figure on this page is counted from Netflix's published charts, and none of them is a cost.
| Year | Titles charting | Companies | Half covered by | If shared equally |
|---|---|---|---|---|
| 2021 — part year, 26 weeks | 59 | 32 | 3 | 16 |
| 2022 | 107 | 55 | 3 | 28 |
| 2023 | 87 | 81 | 5 | 41 |
| 2024 | 59 | 51 | 4 | 26 |
| 2025 | 65 | 52 | 3 | 26 |
| 2026 — part year, 30 weeks | 35 | 35 | 3 | 18 |
A title counts once however many companies are credited on it, so covering "half" cannot be reached by adding the same title twice. One year holds every Korean series that reached any country top 10 that year. "Half takes" is how many companies it takes to cover half of those titles, counting each title once however many companies are credited on it.
3 companies covering half of 107 titles is a great deal of concentration — an even spread would need 28. But it is the same figure at the start and the end of the period, and the column can only move a whole company at a time.
Read as a percentage that column gives 89.3% in 2022 against 88.5% in 2025, a change of -0.8 points. We do not lead with that figure, because a measure that moves in whole companies cannot support a decimal — it is here so you can see it is not being hidden, not because it means anything.
How many companies cover half the titles can only move a whole company at a time, so it is too blunt to read a trend from. The share of titles covered by the three largest companies moves continuously, and is shown against what those three would cover if every company held an equal number.
| Year | Top three cover | Even split would give | Ratio |
|---|---|---|---|
| 2021 — part year | 59.3% | 9.4% | 6.31× |
| 2022 | 63.6% | 5.5% | 11.56× |
| 2023 | 42.5% | 3.7% | 11.49× |
| 2024 | 47.5% | 5.9% | 8.05× |
| 2025 | 56.9% | 5.8% | 9.81× |
| 2026 — part year | 60% | 8.6% | 6.98× |
2022: 11.56× · 2025: 9.81× — a change of -1.75. Down, not up.
Concentration is high and did not increase. Those are two different statements and both are on this table. The lowest the multiple ever falls is 6.31×, in 2021 — so the three largest companies cover several times an even share in every year on this table, including the part years. Across the full years, that multiple fell.
"The three largest" is a count, not a list. We do not name them, because which three they are changes between years and a name here would read as a ranking.
The share of charting titles that passed through a broadcaster falls from 88.8% in 2022 to 80% in 2025, a change of -8.8 points — the shape you would expect if commissioning were moving away from Korean networks. It is the most interesting number we found, and it is not a finding.
Recorded credits per title fall across the same years, from 1.85 to 1.6. A title that never had a broadcaster and a title whose broadcaster nobody has recorded yet look identical in this data.
The control does not have the strength to settle it. A title counts as passing through a broadcaster only if Wikidata records one. If Wikidata simply fills in recent titles more slowly, that alone would produce a decline. The last column repeats the measurement on titles where both a production company and a distributor are recorded, so the entry is demonstrably filled in, and the broadcaster field is the only one in question.
We require 30 such titles in a year before we will read anything from it, so that one title cannot move the figure by more than the decline being tested. No year reaches it — the best is 23. So: The share of charting titles that passed through a broadcaster falls from 88.8% to 80%, but recorded credits per title fall over the same years (1.85 to 1.6), so a title with no recorded broadcaster and a title that never had one look identical here. Restricting to titles whose entry is demonstrably filled in leaves at most 23 titles in a year, and one title moves that figure by more than the decline being tested. We therefore cannot say whether this decline is real, and do not report it as a finding.
We have been caught by this exact thing before. 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 appeared for American, Japanese and British groups. That one we could prove, and published. This one we cannot prove either way, which is the only difference between them.
This cannot say what anyone paid, what was commissioned, or who held the rights. Netflix publishes chart positions; it publishes no money at all. A shift in who appears on a chart is consistent with a shift in power and equally consistent with a change in taste.
How many companies fill half the shelf · Why 7.7% is not one place a week · What a chart position does not say · What we can hand over · How we count
The case that Netflix gained leverage over Korean drama runs through cost: $360,000 an episode in 2015 against $9.8m in 2024. We hold no budgets. We do hold the charts, and from 2022 to 2025 concentration fell rather than rose.
9 August 2026