SEOUL, 22 AUG 2026 — Samsung Electronics said it expects shareholder returns of between 90 trillion and 110 trillion won this year, or roughly US$65bn to US$80bn, which it described as the largest ever by a Korean company. The announcement followed a board meeting on 21 August.
Days earlier, SK Hynix approved the repurchase and cancellation of 40 trillion won of its own shares, about US$28.6bn and the largest treasury share cancellation in the history of Korean listed companies. Two firms, one boom, and the same conclusion within a single week.
What each company committed
Samsung's package is staged. Around 30 trillion won goes out as cash dividends in the third quarter, including the regular quarterly dividend, and roughly 15 trillion won is earmarked for share buybacks tied to employee compensation. The size and structure of the remainder will be decided at a board meeting in late January 2027, weighing further dividends against buybacks and cancellations.
SK Hynix's is not staged, and the distinction matters. Repurchase and cancellation permanently retires the shares rather than parking them in treasury, which lifts earnings per share for everyone remaining and cannot be quietly reversed.
The two decisions are not the same decision
Both companies are returning capital, and the mechanisms carry different messages.
A dividend is a distribution that can be raised, held, or cut next year. A cancellation is structural: the share count falls permanently. A board that cancels forty trillion won of stock is making a much firmer claim about the future than a board that raises a dividend, stating that it does not expect to need that equity again.
Samsung's larger headline number is also the softer one. Roughly 45 trillion won of the 90 to 110 trillion range is specified now; the balance is a January decision presented as an expectation. That is a reasonable way to run a capital plan in a volatile industry, and it is not the same as money committed.
The awkward part is the capex running alongside it
The moves are striking because both companies are in the middle of the heaviest construction programmes in their history.
SK Hynix has committed 54 trillion won to its Yongin M15X and Cheongju M17 fabs, which we covered when the figure was announced. It is now cancelling 40 trillion won of equity in the same year. Those numbers sit within a third of each other, and the company is doing both.
Conventionally, a large buyback signals that management has run out of attractive things to build. That reading is hard to square with the facts here. Neither firm is short of projects, and the high-bandwidth memory required for AI accelerators is capacity-constrained by any account, including the pricing data showing HBM4 estimated at around US$550 a stack against US$200 for HBM3.
The boom is simply generating cash faster than even a historically aggressive capital programme can absorb it. For a company like Samsung, whose second-quarter operating profit rose nineteen-fold year on year, returning the surplus is preferable to holding it. A business with that profile can fund a fab and retire stock in the same year without either decision constraining the other.
Korean governance is the other half of the explanation
The timing suggests a second, parallel explanation: Korean corporate governance.
Korean listed companies have traded for years at a persistent discount to international peers, attributed to conglomerate ownership structures, cross-shareholdings and a historically low tolerance for returning capital to minority holders. Shareholder returns of this scale, framed explicitly as a national record and announced within days of each other by the two largest firms in the sector, read as a response to that pressure as much as to the cash position.
The two motives are hard to separate, which is why SK Hynix’s cancellation matters more than Samsung’s dividend. A company under pressure to look generous can raise a dividend and reverse it. Retiring shares removes the option.
What this signals about the cycle
Memory is the most cyclical business in semiconductors, and the industry's recurring error is treating the top of a cycle as a permanent level. Capital returns announced at a peak have historically preceded the write-downs.
This cycle is at least differently shaped. Demand is contracted, not speculative, with HBM supply largely sold forward to a handful of accelerator makers. And with only three major suppliers, the price collapse typical of a fragmented market is less likely if demand softens.
The counter-argument is the one the industry always ignores at this point in the cycle: contracted demand reflects buyers' current expectations of their own demand, and a slowdown in AI capital expenditure propagates upstream with a lag rather than not at all. Cancelling equity at the top of a cycle is only obviously correct if the cycle does not turn.
Why buyers should read this as a supply signal
For anyone procuring accelerators, the shareholder payout is less important than what it signals about future supply.
A supplier in a constrained market would normally use surplus cash to add capacity and take share while prices are high. Choosing to return cash instead implies that expansion is limited by something other than money — perhaps cleanroom construction, equipment lead times, packaging capacity, or qualification cycles measured in months.
If that is right, the practical consequence is that HBM scarcity is not a funding problem and will not be solved by anyone spending more. It resolves on a construction schedule, which is slower and more predictable than a price signal.
That matters directly for the data centre programmes being built across this region, because memory is a rising share of accelerator cost and none of it is substitutable. A buyer in Johor, Batam or Da Nang negotiating 2027 capacity is pricing against a supply curve these two boards have just told the market they do not intend to bend with capital.
What remains unconfirmed
The composition of the remaining Samsung returns beyond the third-quarter dividend and the employee-compensation buyback is undecided until the January 2027 board meeting, and the 90 to 110 trillion won figure is an expectation rather than an authorised total. Whether any portion of Samsung's buyback involves cancellation rather than treasury holding is not established in the available reporting.
For SK Hynix, the execution timetable for the 40 trillion won repurchase is not described. Neither company has published a statement connecting these decisions to its capital expenditure plans, and no revised capex guidance accompanies either announcement.
What to watch for
The January board meeting is the first real test. If the remainder of Samsung's package is weighted towards cancellation rather than dividends, the two companies have converged on the same structural view. If it is weighted towards dividends, Samsung has kept its options open in a way SK Hynix has not.
The second signal is capex guidance. Returns of this size are only conservative if construction budgets hold, and any reduction announced alongside them would change the interpretation entirely.
The third is whether other Korean listed companies follow. If they do, this is a governance shift with a semiconductor trigger rather than a semiconductor event.
View author profile → · Editorial policy
