Kwak Dong-shin has been chairman of Hanmi Semiconductor Co., Ltd. (KRX:042700) since March 2007 and is identified in its filings as the effective controlling shareholder. He holds 32,018,809 of the company's 95,312,200 shares — 33.59%.
In the six months to August 2026 he bought more, three times, on the open market:
5 January: 47,715 shares at ₩129,910, for ₩6,198,681,300. 29 April: 9,576 shares at ₩315,407, for ₩3,020,342,000. 18 June: 23,657 shares at ₩338,917, for ₩8,017,769,000.
Total: 80,948 shares, ₩17,236,792,300.
Then on 2 July he filed a plan to buy 20,290 more on 3 August, at an expected ₩246,500 a share, for ₩5,001,485,000. That would take him to 32,039,099 shares, or 33.61%.
The stated purpose, verbatim from the filing: acquiring company shares directly for the purpose of enhancing corporate value.
The mechanism deserves explaining, because it has no direct American equivalent.
Since July 2024, Korean law requires company officers and holders of 10% or more to file a trading plan before dealing in their own company's shares — disclosing the intended direction, quantity, expected price and the window. Execution must fall within 70% to 130% of the planned transaction amount. The filing also requires a table of every transaction in the preceding six months, at actual prices.
The effect is that a Korean controlling shareholder cannot quietly accumulate or quietly sell. The market sees the intention a month ahead and can price it.
The closest US comparison is a Rule 10b5-1 plan, but the difference matters. A 10b5-1 plan is adopted privately and provides an affirmative defence; the details are not published in advance. Korea's version is a public announcement of what an insider is about to do, and it is mandatory rather than elective.
For an investor, this makes Korean insider activity unusually legible. The record above is not reconstructed from post-hoc filings. It was announced.
Set the purchase prices against the share price range. Hanmi traded between ₩81,500 and ₩409,500 over the past fifty-two weeks, and closed at ₩217,500 on 25 August 2026, giving a market value of ₩20,730,403,500,000.
The chairman bought at ₩129,910 in January — well timed, roughly 60% above the year's low but far below what followed. Then at ₩315,407 in April and ₩338,917 in June, both of which are above where the stock trades today. The August plan was priced off the ₩246,500 close of 1 July, also above the current level.
So of four disclosed purchases, one looks well timed and three were made at prices the market has since marked down.
That is the useful part. An insider who buys only at the bottom is buying opportunistically. An insider who keeps buying as the price rises, and then keeps buying as it falls back, is either extremely confident about something or is making a deliberate statement about confidence. The amounts — ₩22.2bn across four transactions, equal to about 0.11% of shares outstanding — are large in absolute money and small as a percentage, which fits the second reading more than the first.
It also fits the company's other disclosures. Hanmi filed a corporate value enhancement plan on 20 July, eighteen days after the trading plan, and has held an unusual number of investor events — fifteen investor relations announcements filed in three months.
The business behind the purchases is doing well. Second-quarter 2026 revenue was ₩251.1bn with operating income of ₩130.3bn, up 39.5% and 51.0% year on year, giving a record 51.9% operating margin. The company is the leading supplier of thermocompression bonders, the machines that stack and press the DRAM dies that make high bandwidth memory, and memory makers have been ramping HBM4 production during 2026.
It is also volatile in a way few large Korean companies are. In the fourth quarter of 2025, revenue was ₩83.0bn — a third of the second quarter of 2026 — and the operating margin fell to 33.3%. Equipment orders arrive in waves.
The balance sheet carries no risk at all: total liabilities of ₩123.0bn against ₩690.3bn of equity at the end of FY2025, with ₩276.2bn of cash.
Be sceptical properly.
₩22.2bn is a great deal of money to most people and a small amount relative to a 33.59% stake in a ₩20.7tn company. Kwak's holding is worth roughly ₩7tn at the current price. Adding ₩22bn to that is a 0.3% increase in his own position — closer to a gesture than a bet.
Controlling shareholders also buy for reasons other than valuation. Raising a stake from 33.59% toward 34% strengthens control at the margin, and in a market where the amended Commercial Act has increased shareholder rights and made hostile action easier to contemplate, incremental accumulation by a founder-controller has a defensive logic independent of price.
And insiders are not reliably right about their own stock. The April and June purchases are underwater. Somebody with complete knowledge of Hanmi's order book bought at ₩338,917 in June, and the market has since taken 36% off that price. Whatever he knows, it did not include what the stock would do.
The next trading plan filing. Under the disclosure rule, any further purchase requires an advance announcement, so continuation or cessation is visible. A chairman who keeps buying below ₩220,000 after buying at ₩338,917 is doubling down. One who stops has learned something.
Second, whether the 3 August plan executed as filed. The rule permits execution between 70% and 130% of the planned amount, and a subsequent ownership report shows the actual. Ownership reports were filed on 16 June and 30 July; the August transaction would appear in a later one.
Third, and most important for anyone deciding whether the insider signal is worth following: the company's own capital return. A controlling shareholder buying stock personally and a company buying stock corporately are very different acts. The first concentrates ownership; the second raises everyone's per-share claim. Hanmi filed a corporate value enhancement plan in July. What that plan commits the company itself to do is worth more than what the chairman does with his own money.
kstock reads DART every morning and writes up what moved — the contract, the buyback, the number that does not add up. The daily post and a Saturday roundup, by email.