On July 24 NAVER Corporation's (KRX:035420) board resolved to cancel 4,901,094 of its own shares. Four outside directors attended and none were absent. The filing values the block at ₩1,016,977,005,000, computed from the July 24 close of ₩207,500, and the cancellation took effect on August 3, taking the share count from 156,977,585 to 152,076,491.
That is 3.12% of the company retired permanently, out of treasury stock the company already held, under the article of the Commercial Act that lets a board cancel shares bought within distributable profit without reducing share capital.
Three days later, on July 27, NAVER filed an amendment to its Nvidia disclosure. Among the changes: Nvidia will participate in a third-party allotment of newly issued NAVER common shares worth about $1bn.
At the ₩1,402.50 exchange rate used in Korean filings in late August, $1bn is roughly ₩1.40tn. At the August 27 closing price of ₩216,500, that buys about 6.5 million shares.
So the sequence is: retire 4.9 million shares, then issue roughly 6.5 million. Net, the company ends up with about 1.6 million more shares outstanding than it had before either decision, or roughly 1% more.
Nobody has been cheated here, and it is worth being clear about that before going further. The treasury shares were bought in earlier periods, so the cancellation consumed no cash in July. The allotment brings in roughly ₩1.40tn that NAVER did not previously have. In economic terms this is an equity raise from a strategic partner, at a scale of about 4% of the company, dressed on the same page as a shareholder return.
What it is not is what a cancellation announcement usually signals. When a company retires 3% of itself, the ordinary reading is that management thinks the shares are cheap and there is nothing better to do with the money. Announcing three days later that you will issue more shares than you just cancelled changes that reading substantially.
The obvious alternative was available and NAVER did not take it. It held 4.9 million treasury shares. Selling or transferring them to Nvidia would have delivered a comparable stake, raised comparable cash, and left the share count unchanged rather than higher.
There are reasonable explanations. Disposing of treasury stock to a chosen third party is politically fraught in Korea, because the same mechanism has historically been used to place friendly blocks with allies and entrench control, and it has been a live subject of governance reform debate. A newly issued third-party allotment goes through a defined process with a price formula and a regulatory filing.
The other explanation is commitment. The cancellation traces back to a fair disclosure NAVER made on February 6, 2026, which the filing cites as the related disclosure. If the company had publicly promised to cancel a specified quantity of treasury shares, redirecting them to Nvidia instead would have been a broken promise, and a visible one.
So the sequencing may be the honest outcome of two separate obligations rather than an attempt to have it both ways. It is still the case that continuing shareholders end up owning slightly less of the company than they did in July, and that the ₩1.02tn headline on the cancellation reads better than the combined effect.
None of this means NAVER is a poor steward of capital. The evidence points the other way.
Dividends paid were ₩62.4bn in FY2023, ₩119.0bn in FY2024 and ₩168.4bn in FY2025. In the first half of 2026 alone the figure was ₩393.6bn, more than double the whole of last year. Add the ₩1.02tn cancellation and the capital returned this year is a material fraction of a company capitalised at ₩32.92tn.
The balance sheet supports more. Cash stood at ₩6.38tn at June 30, total equity at ₩31.03tn, and retained earnings at ₩28.26tn. The shares trade at about 1.06 times book value, which for a business with a durable search monopoly in its home market and a 16% operating margin is not a demanding multiple.
A company at book value, retiring stock, raising its dividend and simultaneously bringing in a strategic investor at a scale that funds an entirely new business line, is not obviously doing anything wrong.
Take the other side seriously. What NAVER has done is fund an AI infrastructure venture with partner equity rather than with debt or with its own cash, while continuing to return capital from the existing business.
That separation is arguably ideal. The search and commerce operation generates cash and gives it back. The AI factory, which is speculative, capital-hungry and outside NAVER's demonstrated competence, gets funded by an investor with a strategic reason to want it to exist and the technical knowledge to judge whether it can work. If the venture fails, the loss is shared with a partner who paid to be there.
Compare it to the alternatives. Funding ₩1.40tn from the balance sheet would have consumed a fifth of the cash. Borrowing it would have put leverage against an unproven business. Neither is better than selling 4% to Nvidia.
And 1% net dilution is small. Investors who object should ask themselves what they would pay for a 4% stake in a NAVER that has Nvidia as a shareholder and a partner, versus one that does not.
The allotment price. Korean third-party allotments are priced off a formula based on trading over a reference period, usually with a permitted discount, and the separate capital increase filing of July 27 contains the mechanics. A price near the market means Nvidia paid up. A price at a meaningful discount means existing holders subsidised the strategic relationship, and the size of that subsidy is calculable once the number is public.
The second thing is whether the subscription completes at all. A capital increase decision is not a receipt. The NAVER filing says the company will report the status of the wider Nvidia project by December 31, 2026, and the closing of a $1bn share subscription would qualify as material progress well before then.
The third is whether the cancellations continue. NAVER has now retired 3.12% of itself in one resolution. If a further cancellation follows in the first half of 2027, on top of the shares issued to Nvidia, the return programme is genuinely running alongside the dilution rather than being interrupted by it.
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.