KB Financial Group Inc. (KRX:105560, NYSE:KB) earned ₩1,992.2bn attributable to its shareholders in the second quarter of 2026, the biggest quarterly profit in the group's history, and ₩3,884.6bn across the first half. That's 13.1% more than the same period last year. The group's own release puts return on equity at 14.09% and the common equity tier 1 ratio at 13.74%.
All of that is accurate and none of it is a lending story. Net interest income, which is what a deposit-funded universal bank is supposed to earn its living from, grew 1.7% over the half and went backwards in the quarter that set the record. The profit came from fees, from a smaller provisioning charge, and from trading. Whether that's a problem depends entirely on how repeatable you think each of those three is, and they're not equally repeatable.
The half-year report gives both the six-month cumulative figure and the three-month column, which lets you back out the first quarter. Net interest income was ₩3,334.8bn in Q1 and ₩3,143.4bn in Q2. That's a 5.7% sequential decline, in the quarter KB described as its best ever.
Year on year the picture is flatter than it is bad: ₩6,478.3bn for the half against ₩6,368.7bn, up 1.7%. Both sides of the spread contracted. Interest income fell 1.5% to ₩14,511.8bn while interest expense fell 4.0% to ₩8,033.6bn, which is the signature of a falling policy rate reaching deposit costs faster than it reaches loan yields. For now that arithmetic works in KB's favour. It stops working when the loan book finishes repricing.
What makes the flat NII awkward is that the balance sheet did not stand still. Loans measured at amortised cost went from ₩491,978.0bn at the end of 2025 to ₩505,399.0bn at the end of June, up 2.7% in six months. Customer deposits grew faster, 5.2%, to ₩486,343.5bn. Cash and due from banks rose 28.3%. A bank taking in deposits at twice the rate it can lend them is a bank whose margin is being squeezed from the funding side, and the ₩13.4tn of new lending produced essentially no additional net interest income.
Total assets grew ₩69.0tn in six months, to ₩866.9tn. Only ₩13.4tn of that was lending. Other assets nearly doubled to ₩51.9tn and other liabilities rose ₩33.3tn to ₩77.0tn, the two moving together in a way that usually means unsettled securities positions at a brokerage subsidiary rather than anything about the bank. The half-year summary statements don't break it out, so I'm inferring from the pairing rather than reading it.
Net fee and commission income was ₩2,961.2bn for the half against ₩1,966.0bn, up 50.6%. In the second quarter alone it was ₩1,601.9bn against ₩1,031.9bn, so the acceleration is getting faster, not fading. Gross fee income rose 40.6% while fee expense rose only 16.2%, meaning the mix improved as well as the volume.
Here's the number that frames the whole half. Operating profit rose ₩1,014.1bn year on year. Net fee income rose ₩995.3bn. Those two figures are within ₩20bn of each other.
That's arithmetic coincidence, not causation. Several other lines moved by hundreds of billions and partly cancelled each other out: gains on financial instruments at fair value through profit or loss added ₩473.4bn, other operating losses shrank by ₩1,114.8bn, general and administrative expenses rose ₩299.1bn, the insurance service result fell ₩94.2bn, and other insurance finance charges swung ₩1,583.3bn against the group. But when the single fastest-growing line and the change in operating profit are that close, the sentence to remember is that without the fee growth there is no record.
Credit costs did the rest. The provision for credit losses was ₩1,013.0bn for the half against ₩1,310.7bn, a ₩297.7bn reduction that flows straight to the bottom line. That's about 29% of the increase in operating profit. Provisions did tick up sequentially, from ₩493.3bn in Q1 to ₩519.7bn in Q2, so the improvement is a year-on-year effect that has already stopped compounding.
Other insurance finance charges came in at negative ₩1,996.6bn for the half, against negative ₩413.3bn a year earlier. Taken alone that looks like a subsidiary blowing up. It isn't, or at least the statements argue it isn't: in other comprehensive income, insurance contract-related gains ran to positive ₩4,245.2bn against negative ₩1,071.7bn last year. The liability side and the asset side of an insurance book move against each other when discount rates move, and IFRS 17 splits the two across the income statement and OCI. The net effect on equity is far smaller than the P&L line implies.
This cuts both ways for the bear case. It means the ₩2.0tn charge shouldn't be read as economic damage. It also means anyone treating KB's insurance arm as a stable earnings contributor is reading a line that will keep swinging by trillions of won as Korean rates move.
At ₩166,600 the shares carry a market capitalisation of ₩59.1tn. First-half basic earnings per share were ₩10,670. Annualise that and you get roughly 7.8x. Measured against last year instead, FY2025 profit attributable to the parent was ₩5,833.2bn, of which ₩202.4bn went to hybrid capital holders as coupons, leaving about ₩5,630.8bn for common shareholders and a multiple near 10.5x. Both are defensible, and the gap between them is the whole argument: 7.8x assumes the first half repeats, 10.5x assumes it doesn't.
The bull case takes the fee growth as structural. Korean retail investors have been moving money into securities accounts and overseas equities at a pace that shows up directly in brokerage and wealth management fees, and KB Securities is a large enough franchise to capture it. If that's the driver, a 50% year is not a one-off and the low multiple is simply wrong.
The bear case is that fee income of this kind is the most cyclical thing a Korean financial group owns. It rises with market turnover and falls with it, and it fell hard in 2022. Pair that with net interest income that is flat at best and provisions that have stopped falling, and a bad market quarter takes the group's earnings down more than a loan-loss cycle would have a decade ago. The group is more sensitive to the KOSPI than its balance sheet suggests, and less sensitive to the loan book.
There's a third possibility I can't rule out from the summary statements, which is that part of the fee jump is a consolidation or classification change rather than underlying volume. The half-year report's summary income statement doesn't segment fees by subsidiary. That would be in the segment note, which I haven't read.
Two lines in the third-quarter report, which is due in mid-November on the usual Korean filing calendar.
First, net interest income. It has now printed ₩3,334.8bn and ₩3,143.4bn in consecutive quarters. A third quarter below ₩3.1tn would confirm that the sequential decline is a margin trend rather than a timing quirk, and it would mean the loan growth of the first half bought nothing.
Second, net fee income against the ₩1,601.9bn Q2 mark. Holding above ₩1.5tn would make the case that the step-up is a new level. Falling back toward ₩1.2tn would put the group's operating profit growth close to zero and reprice the multiple without a single thing changing at the bank.
The provisioning charge is the tiebreaker. It has already stopped helping year on year. If it starts rising while fees fall, the record half of 2026 will read in hindsight as the top.
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.