On 21 August 2026, Kakao Corp.'s (KRX:035720) board approved a plan to divide the company in two. Trading in the shares was halted and resumed the same day. Alongside the split decision, the company filed a merger decision, a record-date notice, and a corporate value enhancement plan that sets out the whole thing in unusual detail.
The two resulting companies have provisional names: Kakao X and Kakao AI.
The stated reason, from the filing's own diagnosis section, is worth quoting closely. The group's diverse business portfolio makes a conglomerate discount unavoidable, and market demand for a re-rating of its medium-to-long-term growth businesses has been increasing.
A Korean company writing "conglomerate discount" into a regulatory filing as the problem it is solving is not common.
The split is an 인적분할 — a horizontal split, in which existing shareholders receive shares in both resulting companies in proportion to what they held. No cash changes hands. Nobody is diluted. If you own 1% of Kakao today, you own 1% of Kakao X and 1% of Kakao AI afterwards.
The alternative structure, a 물적분할 or vertical split, hands the new subsidiary entirely to the parent and gives existing shareholders nothing directly. That is what LG Chem did with its battery business, and the resulting anger from retail investors was one of the main political drivers behind Korea's corporate governance reforms of the last three years.
Kakao choosing the horizontal route is therefore a deliberate signal, and a costly one — it gives up the option of raising capital at the new entity without the parent's shareholders participating.
The filing sets 2030 financial goals for each entity.
Kakao X: revenue above ₩10tn, growing at roughly 13% a year compounded from FY2025. Its stated role is supporting innovative growth by business area and securing growth engines.
Kakao AI: revenue above ₩6tn, growing at roughly 20% a year, with EBITDA above ₩2tn, an operating margin above 30%, and return on equity above 25%. Its stated role is advancing service completeness and revenue models on the strength of AI technology.
Work backwards from those and the implied FY2025 starting bases are about ₩5.4tn for Kakao X and ₩2.4tn for Kakao AI — which sum to roughly ₩7.8tn against the ₩8,099,147,815,086 of revenue Kakao actually reported in FY2025. The targets are coherent with the current business rather than aspirational arithmetic bolted on afterwards.
Together they imply group revenue of ₩16tn by 2030, roughly double FY2025.
The margin target is the striking one. Kakao's whole-company operating margin was 9.0% in FY2025 and 12.1% across the first half of 2026. Kakao AI is meant to run above 30%. That is a different kind of business — closer to a software company than to the advertising, commerce, mobility and content mix the group runs today.
The value-up plan commits to three things beyond the split.
Thirty percent of after-tax dividends received from subsidiaries will be returned as cash dividends or buybacks and cancellations. Up to 30% of investment gains, measured after cost of capital and tax, will fund additional returns. And ₩300bn of buyback and cancellation over the three years following the split.
Be honest about the scale. ₩300bn over three years is real money and it is modest against a company of this size. Kakao's dividend history is thinner still: dividends paid were ₩38.9bn in FY2025 against net income of ₩517.96bn — about 7.5% — and the value-up filing itself records no figure at all for the prior year's payout ratio, noting the company does not qualify as a high-dividend company under Korean tax law.
So the return commitments are a start rather than a transformation. The value creation, if it comes, has to come from the re-rating the split is designed to produce.
The governance limb is vaguer: improving board expertise and establishing an optimal decision-making structure. That is the section to press management on, because a split that leaves the same board making the same decisions at two companies achieves less than the filing implies.
Conglomerate discounts are usually real, and Kakao's case for one is stronger than most.
An investor who wants exposure to Korean AI adoption currently has to buy a company that also runs a taxi-hailing app, a web-comics publisher, a game studio, and stakes in a listed internet bank and a listed payments company. An investor who wants the messaging platform's advertising business has to take the AI investment spending too. Neither gets what they want, and both discount the parts they did not ask for.
Separating them lets each be valued on its own metrics, and lets each raise capital on its own terms. Kakao AI at 20% growth and a 30% margin would attract a completely different shareholder base from Kakao X at 13% growth.
The company has also been shrinking the sprawl for two years. In the second quarter of 2026 it recognised a loss of ₩180.8bn from discontinued operations arising from deconsolidations and stake sales. Non-current assets fell from ₩15,409.3bn at the end of December to ₩12,251.2bn at 30 June, with intangible assets — mostly goodwill from a decade of acquisitions — dropping from ₩5,227.8bn to ₩3,642.9bn. The split follows a genuine simplification rather than substituting for one.
And the operating business is performing. Second-quarter revenue of ₩2,098.5bn and operating income of ₩277.0bn were both quarterly records, up 9% and 36% respectively.
Three objections deserve stating.
First, splits do not create value by themselves. Two sets of listed company costs, two boards, two finance functions, and shared services that have to be contracted between the entities. If the discount was 20% and the friction costs 5%, the gain is smaller than the announcement suggests.
Second, the split ratio is everything and it is not yet in front of shareholders. How the assets, cash and debt divide between Kakao X and Kakao AI determines what each holder actually receives, and Kakao's ₩8,178.1bn of cash at 30 June is a large prize to allocate.
Third, the AI revenue does not exist yet. Kakao has said it targets more than 10m monthly active users of AI services inside KakaoTalk by the end of 2026, with meaningful AI revenue starting in 2027 and expanding faster through 2028. So Kakao AI is being set a ₩6tn 2030 revenue target and a 30% margin on a business whose monetisation has not started. Most of its ₩2.4tn implied starting revenue must therefore be existing businesses reclassified into the AI entity — which is fine, and which means the label is doing some work.
The split ratio and the asset allocation, in the securities registration statement that must precede the shareholder vote. Cash, debt, the KakaoBank and KakaoPay stakes, and which operating businesses land where. Everything else in this story is downstream of that document.
Second, whether the AI monetisation milestones hold. Ten million monthly active users by December is a checkable, dated commitment, and the first one due.
Third, the buyback. ₩300bn over three years is the floor. Whether the boards of the two successor companies go beyond it — particularly Kakao AI, if it hits a 25% return on equity — is what turns a structural fix into a shareholder return.
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.