Mirae Asset Securities Co., Ltd. (KRX:006800) made ₩1,905.2bn of net profit in the second quarter of 2026. That is more than any of Korea's five largest commercial banks earned in the same three months, roughly ₩600bn ahead of Shinhan Bank. It was the second consecutive quarter above ₩1tn, a first for a Korean brokerage, and it brought first-half net profit to ₩2,907.2bn — 1.8 times what the company earned in the whole of FY2025. Return on equity, per the company's own second-quarter presentation, was 39.5%.
The shares closed at ₩34,900 on September 1. Equity at the end of June was ₩16,249.9bn against 543,021,051 shares, so book value is about ₩29,924 a share. The market is paying 1.17 times book, and about 3.3 times first-half earnings annualised, for a business earning a 39.5% return on that book.
Those two paragraphs cannot both be describing a rational market unless the market has decided the earnings are not the earnings. That's the argument worth having.
Before anything else, a warning about the numbers a screen will show you. Second-quarter revenue was ₩21,631.4bn. That's ₩21.6tn of "revenue" at a company with ₩16.2tn of equity, and it means nothing resembling what revenue means at a manufacturer.
A Korean securities firm's 영업수익, its operating revenue, is a gross figure. It sweeps in interest income, the proceeds of securities sold from the trading book, gains on derivatives before the offsetting losses, and commission income, all added together. The matching costs sit in operating expenses. So revenue tripling year on year, from ₩7,694.9bn in Q2 2025, mostly tells you the trading book turned over more, not that the business tripled. The line to use is operating income, which was ₩2,489.9bn, or net profit.
The same applies to cash flow. Operating activities consumed ₩7,734.0bn in the first half, and financing activities produced ₩9,394.8bn. At an industrial company that pattern would be an alarm. Here it is what growth looks like: margin loans to clients, bigger trading inventory and repo-funded positions all sit in operating cash flow, and they're financed with borrowings. Judging a brokerage on free cash flow produces nonsense in both directions.
And the balance sheet grew hard. Total assets went from ₩140,348.0bn in June 2025 to ₩195,605.1bn in June 2026, up 39%, and from ₩150,283.9bn at December, up 30% in six months.
The headline driver reported everywhere was brokerage fee income of ₩625.6bn, up 36% on the quarter and an all-time high, on the back of a strong domestic market. That is real and it is the part investors most want to see, because commission income scales with account activity and is the closest thing a brokerage has to recurring revenue.
It is also about a quarter of the quarter's ₩2,489.9bn of operating income, and a third of net profit.
The rest came from elsewhere. Net interest is not the answer: interest income of ₩1,601.4bn against interest expense of ₩1,418.0bn leaves a ₩183.4bn spread, which has been roughly flat for two years. What's left is trading and principal investment, investment banking fees, wealth management and the overseas business, and the summary statements don't split them. The quarterly report does, and anyone underwriting this stock should read that page before anything else.
The reason it matters is that those lines have the widest distribution. Commission income falls when volumes fall. Principal investment gains can go negative. Q3 2025 shows the shape of a bad quarter under the same roof: ₩222.8bn of operating income, less than a tenth of what Q2 2026 produced, in a company of roughly the same size.
So the market's 3.3 times multiple isn't obviously stupid. It's a statement that a 39.5% ROE is a peak, not a level.
Assets of ₩195,605.1bn against ₩16,249.9bn of equity is 12.0 times leverage, up from 11.3 times a year earlier. That is normal for the industry and it's also the whole risk: a 12-times-levered book needs only a few percent of asset impairment to consume a year's earnings.
Two things suggest the cycle is late rather than early. Credit loss provisions were ₩11.1bn in the quarter, against ₩149.7bn taken across FY2023. Provisions that low usually mean the credit environment is benign, which is a description of the present, not a forecast. And in the last two weeks of August alone the company filed a stream of shelf-registration supplements for equity-linked securities and equity-linked derivative bonds.
That last item deserves explaining to an American reader, because it has no clean US retail analogue at this scale. Korean brokerages sell structured products — ELS — to ordinary savers in enormous volume. They typically pay an above-deposit coupon unless an underlying index falls through a knock-in barrier, at which point the buyer takes equity-like losses. In 2024 a wave of these tied to the Hang Seng China Enterprises Index breached, and Korean financial firms ended up paying out large compensation to retail buyers under regulatory pressure over mis-selling. The product is profitable to issue and it carries a tail that is part market risk and part conduct risk. A US reader looking for the risk in a Korean brokerage should look here before looking at the loan book.
On August 27 the Korea Economic Daily reported that Mirae Asset and Toss were acquiring a Japanese brokerage. The company filed a response the same day. It said the Toss part is wrong — the two are unrelated — and then confirmed the substance: it is reviewing an acquisition of a Japanese brokerage, with nothing specifically decided. It committed to disclose again when the matter is settled or within one month, and put a date on it: September 23.
This mechanism is worth knowing. Korea has a compulsory clarification regime, 풍문 또는 보도에 대한 해명, under which a listed company must answer a market rumour or media report rather than decline to comment. When the answer is 미확정 — undetermined — the company doesn't get to leave it there. It has to name a re-disclosure date, and the exchange holds it to that date.
A US company in the same position says it does not comment on market speculation, and no calendar entry exists. Here, an investor knows that within three weeks Mirae Asset either announces a Japanese acquisition, says it isn't happening, or files another undetermined answer with a further date — and each of those three is informative.
That matters for the thesis because it goes to what the company does with the money. A firm earning ₩2.9tn in six months and trading at 1.17 times book has two obvious uses for capital: buy back stock at that multiple, or expand. The Japanese review says which one management is thinking about.
The bear case doesn't need much work. Korean brokerage earnings have always been cyclical and this cycle has been violent: the shares have a fifty-two-week range of ₩18,440 to ₩83,800 and sit at ₩34,900, 58% below the high. A market that has already watched this stock round-trip once inside a year is entitled to discount a record quarter.
Add that the profit mix is weighted toward the volatile lines, that leverage is rising, that provisions are at a low, and that a large cross-border acquisition would consume capital rather than return it. On those facts, 1.17 times book for a business whose through-cycle ROE is far below 39.5% is defensible arithmetic rather than mispricing.
The bull answer is that book value itself is compounding fast — equity went from ₩13,478.2bn in December to ₩16,249.9bn in June, and retained earnings from ₩6,718.7bn to ₩9,169.1bn — so a flat multiple on a rising book still produces a return. That is the quieter case, and it doesn't require the peak to repeat.
September 23, for the Japanese answer. Then the Q3 report in November, and inside it the fee-and-commission breakdown rather than the headline. Two numbers settle the argument: whether brokerage fee income holds anywhere near ₩625.6bn once the domestic market cools, and what the trading and principal investment lines do when they aren't being helped.
The third thing is capital return. At 1.17 times book with ₩2.9tn earned in six months, a buyback is the cheapest thing this company could buy. Whether the board reaches for it or for Tokyo tells you how management reads its own multiple.
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.