The other side of the trade: KRW 55 trillion, and 19 November

Every figure below traces to a DART filing or to Korea Exchange investor data. Nothing is estimated. Dollar conversions use the 23 September snapshot of 1,365.1 won, which is a snapshot and not a closing rate.

The sentence that was missing

On 23 September, foreign investors sold KRW 1.18 trillion (~$864M) of SK hynix. Every desk in Seoul wrote that down. Almost nobody wrote down who bought it.

SK hynix did. On the same session the "other corporations" account bought KRW 1.14 trillion (~$835M) of SK hynix stock, and the company had told the exchange a month earlier that it would.

That single pairing is the story, and it repeats every trading day. It is not an inference from flows. It is two disclosure filings and the exchange's own investor data agreeing with each other, day after day, since 20 August.

What the filings say

Two Korean companies filed a 주요사항보고서 (자기주식취득결정), a major-matter report of a decision to acquire treasury shares, within three days of each other.

Combined: KRW 55.00 trillion (~$40.3B) of announced on-market buying, inside a three-month window that both companies put in writing. SK hynix's filing states the amount to the won: 40,004,340,000,000. Where this piece rounds, it rounds only the display.

Each company's buying began on the first trading day after its own filing. SK hynix filed on Wednesday 19 August and the "other corporations" bucket bought KRW 1.04 trillion of SK hynix on Thursday 20 August. Samsung filed on Friday 21 August and the same bucket bought KRW 527 billion of Samsung on Monday 24 August. Before those dates the bucket behaved as it always had: on 18 August it was a net sale of KRW 38 billion spread across 418 stocks.

What the market data shows

Measured on the exchange's investor data, through 23 September:

Two things are worth saying plainly about that table.

First, the daily figures move. SK hynix ranged from KRW 988 billion to KRW 1.18 trillion, Samsung from KRW 431 billion to KRW 560 billion. A band of roughly ten per cent is what a fixed daily share count looks like when it is priced at a moving market. It is the signature of a programme executing to a schedule, not of discretionary buying.

Second, the pace. At the rate of the past month, both programmes reach their announced amounts well before their stated deadlines. Samsung's 15 trillion is about thirty trading days of buying at its current pace; it has used twenty-three. SK hynix's 40 trillion is about thirty-six sessions; it has used twenty-five.

A limit we will state rather than hide: the cumulative figures above are our measurement of the "other corporations" net buying in each stock, not a company-reported execution total. Korean issuers report execution in a 자기주식취득결과보고서 after a programme completes. Until those land, the two columns are the best available proxy and nothing more. We have not seen a single filing that reports these totals, and we do not present them as if we had.

Why the two programmes are not the same event

Both are called a buyback. They do different things to the value of a share.

SK hynix states the purpose as cancellation. Shares that are cancelled are gone. The denominator of every per-share figure falls permanently: earnings per share, book value per share, dividend per share on an unchanged payout. Nothing can put those shares back.

Samsung states the purpose as employee share compensation. Shares bought for that purpose are held and then delivered to employees. They come back to the market's float on a schedule the company controls. The per-share effect is a timing effect, not a permanent one.

This is the distinction our Follow-Through Score is built around, and it is the one that most English-language coverage collapses. A reader who is told "Samsung and SK hynix are both buying back KRW 55.00 trillion of stock" has been given a number and taken away the only part of it that determines per-share value.

The scope of this comparison, stated plainly

The table above compares two on-market purchase programmes running in the same window. It is not a comparison of the two companies' capital returns, and it should not be read as one. Both boards announced considerably more in the same week, and both announcements sit in a different evidence class from the filings above: they are plans with ranges and pending board decisions, not major-matter reports with fixed amounts and dates.

• SK hynix, 19 August: the 40 trillion is 24.07 million shares, about 3.3% of shares outstanding, to be cancelled in full after acquisition. The company also raised its shareholder-return target to more than 50% of cumulative free cash flow for 2025 to 2027, from a previous ceiling of 50%, and said cash dividends would run alongside, with expanded fixed and special dividends under review.
• Samsung Electronics, 21 August, the same day as its filing: the board resolved a 2026 shareholder return of roughly 90 to 110 trillion won, including about 30 trillion in cash dividends in the third quarter. The 15 trillion buyback for employee compensation is one line inside that plan, not the plan. Details are to be fixed at the October board, with the remainder decided at the January board. For reference, the company executed 8.4 trillion of buy-and-cancel last year.

Two things follow, and they point in opposite directions.

Judged on total announced capital return for 2026, Samsung's number is the larger one, and a comparison that stops at 40 versus 15 gets that backwards. Judged on what puts a bid under the stock between now and November, the dividend does not participate: a cash dividend is money leaving the company, not an order in the book. Only the two on-market programmes absorb selling, and only they have an end date that is already written down.

That is why the rest of this piece stays with the two programmes. The broader plans change how you rank the two companies. They do not change who is standing on the other side of the trade this quarter.

What 23 September actually looked like

The index closed up 0.90% at 7,080.92. It opened at 7,153.99, which was also the high of the day, and traded down to 7,014.98 before closing. Non-arbitrage programme selling of KRW 788 billion sat on the index all session.

Underneath, two stocks accounted for almost everything that mattered.

SK hynix. Foreign investors sold KRW 1.18 trillion (~$864M). Institutions bought KRW 230 billion, retail sold KRW 190 billion, and the disclosed corporate buyer took KRW 1.14 trillion (~$835M). The foreign sale and the company's own purchase were within four per cent of each other.

Samsung Electronics. Retail sold KRW 2.17 trillion (~$1.6B), the largest single-stock retail sale of the session. Foreign investors bought KRW 1.29 trillion (~$944M), institutions KRW 383 billion, and the company's programme KRW 502 billion (~$368M).

The common reading of that day was a rotation inside semiconductors: long Samsung, short SK hynix. That reading describes the buy side of one stock and leaves the sell side of the other unexplained. What the data shows is that the largest single counterparty to the day's foreign selling in SK hynix was SK hynix, and that nearly a quarter of the retail exit from Samsung was absorbed by Samsung.

The date that is already in the calendar

Disclosed programmes end. These two end on 19 November and 21 November, and they may exhaust their announced amounts before then.

The announced amounts run out before the dates do. Samsung has KRW 3.32 trillion left of its 15 trillion and has been buying about KRW 508 billion a session: six or seven more sessions. SK hynix has KRW 12.54 trillion left of its 40 trillion and has been buying about KRW 1.10 trillion a session: roughly eleven more. Counting from 23 September and allowing for the Chuseok closure, that is the first half of October for both, not the third week of November.

That calculation assumes the pace holds, and the pace is the one term here that is not a filing. It is also the one to watch. If the daily figures fall while a large part of the announced amount is still unspent, the buying is being paced rather than completed, and the end moves back toward the filed date.

That matters more than it sounds. A buyer who is price-insensitive, publicly scheduled, and buying between eight and twenty-four per cent of a stock's daily turnover is not a marginal participant. When the buying stops, it stops completely, on a date that is already known to everyone who reads the filings.

We are not going to tell you what happens next, because we cannot measure it. What we can say is what changes and what does not:

• What is known: the end dates, the announced amounts, the purposes, and the pace to date.
• What is not known: whether foreign selling in SK hynix persists past the programme, who the marginal buyer is after it, and whether either company files a further programme. None of that is in any document we have read.

The question we will be carrying into November is narrow enough to answer with data when the time comes: when the disclosed buyer leaves, who takes the other side?

What we will be watching, and why

1. The progress ratio, weekly. At 68.6% and 77.8%, the announced amounts are the binding constraint before the calendar is. If the daily figures fall sharply, the amount left is what tells you why: with little left the programme is finishing, with much left it is being paced.

2. The gap between the two stocks. SK hynix cancels, Samsung compensates. The per-share arithmetic diverges at the end of each programme, not during it.

3. Foreign selling in SK hynix. Not whether it turns positive on any one session, which is a coin flip, but whether it is still running at trillion-won scale when the programme's bid is gone.

4. The result filings. 자기주식취득결과보고서 for both companies will give the executed totals. That is the moment our measurement can be checked against the companies' own numbers, and we will publish the comparison either way.

Data sources: DART filings 20260819000254 and 20260821000616; Korea Exchange investor trading data [12009] and per-stock net purchases; KRX index data. Board announcements of 19 and 21 August 2026 for the wider shareholder-return plans, which are plans and not filings of fixed amount.

Research only. Not investment advice. KoreaAlpha does not provide personalised recommendations.