iScream Media (KOSDAQ: 461300): 93% of Korea's primary teachers open its software every day, 31% operating margins, and most of the market cap is cash. Last week they finally started paying it out. [DD]
I have been reading Korean micro caps straight off DART for a while now. This is the one where I went back and checked my own numbers three times. Then last week the company did the exact thing that answers the main objection to it, so here is the whole case in one place.
**The business**
iScream Media runs iScream S, the software that roughly 93% of Korea's primary school teachers open every day to run their classroom. Lesson slides, worksheets, printables, a library of about 6.5 million teaching materials built up over twenty years.
It is free. That is the entire point. It is the front door.
Because iScream already owns the teacher's daily routine, it started publishing state certified primary textbooks in 2022 and immediately took the highest adoption of the thirteen licensed publishers, somewhere around 40% to 58% share in maths, science and social studies.
**Why it is a good business**
The state pays for the textbooks. The teacher picks the publisher. iScream already owns the teacher.
Schools re-order every year, so this is recurring revenue rather than a one off sale. Switching publisher means relearning a full year of materials, so teachers do not bother. And a competitor would have to rebuild two decades of content and then prise teachers off a platform they open every single morning. That is a real moat, not a slide in a deck.
**The numbers, FY2022 to FY2025**
Revenue ₩104bn to ₩196bn, roughly 24% a year.
Operating profit ₩31bn to ₩62bn.
Operating margin up from 29% to 31.5% while it grew. Margins rising during growth is the tell of a good business.
Free cash flow about ₩48bn in FY2025, roughly a quarter of sales.
Return on the capital actually used in the operating business is comfortably above 50%. Screeners show around 30% because they treat the cash mountain as capital employed.
**The weird part**
Market cap when I first wrote this up was about ₩200bn. Cash and securities on the balance sheet: ₩182bn.
Take the 598,969 treasury shares out of the count, take the cash off, and the market was pricing a business earning ₩62bn a year at an enterprise value in the single digit billions of won. That is a few million dollars. It is not a valuation, it is a rounding error.
Two things about the data, because this is where most people get it wrong:
Screeners show roughly ₩111bn of "cash and short term investments." The real figure is ₩182bn. The missing ₩70bn odd sits in long term investments that screens do not count as cash. You only see it if you open the filing. Some of it sits in venture funds tied to the founding family, which is its own conversation.
Data providers also leave the treasury shares in the share count, so every per share number built on top of it is off.
**Why it is cheap**
Two honest reasons.
The founding Park family controls about 61% through SigongTech. No outside investor is forcing anything here.
Revenue rides on government policy. In August 2025 the ministry demoted iScream's AI digital textbook from "textbook" to "educational material" and about ₩13bn of capitalised development cost went out of the window. That risk is live and it does not go away.
**The update, and this is the new part**
Two DART disclosures in the last two weeks.
29 July: an interim dividend of ₩1,083 per share for H1 2026, about ₩13.8bn in total, payable 25 August. For scale, the entire FY2025 dividend was ₩1,554.
7 August: a ₩5bn buyback with an explicit commitment to cancel 70% of the repurchased shares. That is the part that matters. Earlier buybacks sat in treasury or got pledged against an exchangeable bond, which does nothing for a minority holder. Cancellation is permanent. Net effect on the share count is roughly 1.7%.
I would not annualise that dividend and call it a 12% yield, even though the arithmetic gets you there. Korean interim dividends are lumpy and the stated policy is a 40% payout ratio, not a fixed amount per share. What the two filings do give you is direction. The Value-Up commitment is now producing cash actually leaving the company and shares actually being destroyed, and that was the open question the whole thesis hung on.
The stock is up roughly 13.5% since I published and about 8% on the buyback announcement alone, so the cash to market cap ratio is not quite as extreme as it was. It is still absurd.
**What would break it**
The elementary curriculum rollout that drove textbook adoption from 2022 is finished. There is no new grade cohort in FY2026, so growth flattens from here and I would not underwrite that 24% CAGR going forward.
And Korea will have fewer than 300,000 first graders this year. The demographic decline is real and it is slow. My argument is not that this business grows forever. It is that a slowly declining business and a business priced below its own cash pile are two different things, and the second one keeps paying you while you wait.
Not investment advice, I may hold a position. All figures from DART filings, FY2022 through Q1 2026, plus the July and August disclosures.
Full teardown with the charts, the segment breakdown, the seasonality trap (textbooks ship in Q2 and Q4 only, which makes Q1 look like a collapse) and the demographics pushback is here: [https://www.numbersnotnarrative.com/p/i-scream-media-93-of-koreas-teachers](https://www.numbersnotnarrative.com/p/i-scream-media-93-of-koreas-teachers)
Curious whether anyone here has a view on whether this keeps coming, or whether those two filings are a one off.
