Everything About AI
NextFin News - Samsung Electronics shares slid as much as 6.6% on Wednesday after the world's largest memory-chip maker posted record quarterly profit yet handed investors the one thing they were waiting for only in January: a concrete shareholder-return plan. The sell-off captures a widening rift in the AI chip trade - record earnings no longer move the stock; capital discipline does. Samsung said it would announce the details of its return program at the end of January 2027, a deferral that stood in sharp relief against rival SK Hynix, which reported blockbuster results a day earlier alongside a commitment to return roughly $45 billion more to shareholders this year.
The Situation: Record Profit, Muted Payout
Samsung Electronics reported third-quarter consolidated sales of approximately 195 trillion won and operating profit of about 107.4 trillion won, figures that clear the 106.1 trillion won consensus tracked by LSEG and mark the first time a technology company has posted quarterly operating profit above 100 trillion won. On the surface, the numbers are extraordinary: operating profit up roughly 783% from a year earlier and 20% from the previous quarter, extending the company's streak of quarterly profit records to four.
Yet the stock reversed earlier gains and slid as much as 6.6% in Seoul. Three things weighed on sentiment. First, profit from the company's chip operations missed estimates. Second, capital expenditure expanded, signaling that the windfall will be reinvested rather than returned. Third, the shareholder-return plan that investors had hoped would be fleshed out on earnings day was deferred to a board meeting at the end of January.
The contrast with SK Hynix was immediate and punishing. SK Hynix posted record third-quarter revenue of 104.6 trillion won and operating profit of 83.3 trillion won, then confirmed a full-year 2026 shareholder-return program of approximately 100 trillion won, including at least 30 trillion won in cash dividends and more than 60 trillion won in share repurchases and cancellations. That leaves roughly $45 billion in new returns beyond the buyback already executed earlier this year. Samsung, by comparison, announced in August a 2026 return estimate of 90 trillion to 110 trillion won - the largest ever by a Korean company and about five times its previous record of 20.3 trillion won set in 2020 - but the composition remains unresolved, with only about 30 trillion won in third-quarter dividends and a 15 trillion won buyback for employee compensation approved so far.
The central question this sell-off poses is not whether Samsung earned enough. It did. The question is whether the market is now pricing AI chipmakers on earnings power or on capital-return credibility - and whether Samsung's decision to hold cash for the coming capacity war is a defensible strategic call or a governance discount waiting to widen.
Layer 2 - Analysis
Why a Record Quarter Was Not Enough
The mechanics of the disappointment are specific. Samsung's consolidated operating profit of 107.4 trillion won beat the headline consensus of 106.1 trillion won, but the semiconductor division - which investors treat as the actual company within the conglomerate - fell short of estimates. The market had extrapolated from the preliminary guidance and from SK Hynix's 557% year-on-year jump in operating profit; Samsung's chip unit did not clear that bar.
At the same time, capital expenditure rose. In the second quarter alone, Samsung spent 16.8 trillion won on capex, up 5.5 trillion won sequentially, with 15.4 trillion won directed to the Device Solutions division as it expanded the new fab in Pyeongtaek and related infrastructure. The third-quarter earnings confirmed that spending is still climbing. For investors who bought the stock on the promise that the AI memory shortage would convert directly into free cash flow, expanding capex is a double negative: it reduces the cash available for returns today while betting that the demand cycle will still be strong when that capacity comes online in 2027 and 2028.
This is where the cyclical-versus-structural call matters. Samsung's management is treating the AI memory boom as a structural shift that justifies multi-year capacity investment - a regime in which HBM4 and advanced packaging become the mainstream product and long-term agreements lock in pricing. The evidence for that view is real: the company says it has secured long-term agreements covering more than half of its 2027-2028 production capacity, and both Samsung and Micron have signaled that supply tightness will persist through 2027. SK Hynix goes further, forecasting memory shortages through 2030. Under that reading, reinvesting the windfall is not hoarding; it is buying market share in a market that will not clear on its own.
But the stock market is voting on a different timeline. The shares have already done the heavy lifting: Samsung's common stock climbed from 94,800 won in late November 2025 to a record 362,500 won on June 18, 2026, a near-fourfold rise, before giving back roughly 40% of that peak by mid-October. At that price level, the marginal buyer is no longer underwriting the 2026 earnings print - those were priced in months ago. The marginal buyer is underwriting 2027 and 2028, and wants proof that the cash generated in between will find its way back to shareholders rather than into another fab.
The Buyback Gap: Dividends Are Not Cancellations
The heart of the investor revolt is compositional. Samsung's approved returns so far are weighted toward dividends and employee-compensation buybacks - neither of which shrinks the share count. SK Hynix's plan, by contrast, is weighted toward repurchases and cancellations, which directly reduce the denominator and lift earnings per share. In a market obsessed with per-share metrics, a dividend is income; a cancellation is a rerating mechanism.
The distinction is not academic. Samsung's 15 trillion won buyback is earmarked for employee compensation, meaning those shares will be reissued as stock bonuses rather than retired. The 30 trillion won in third-quarter dividends is real cash to shareholders, but it carries no commitment beyond the current quarter. The remaining 60 trillion to 80 trillion won - the bulk of the 90 trillion to 110 trillion won package - will be decided at the January 2027 board meeting, once full-year results are confirmed. That structure hands management maximum optionality and investors minimum certainty.
"Unlike SK Hynix, Samsung has not indicated a possible increase in its current shareholder payout policy, nor has it announced a plan to cancel treasury shares, which could have a more direct and significant impact on the share price, and this is disappointing."
Son In-joon, an analyst at Eugene Securities, put the frustration plainly. The line that matters is "cancel treasury shares." Until Samsung commits to retirements, the so-called Samsung discount - the persistent valuation gap between Samsung and its peers despite comparable or superior scale - has a fundamental reason to persist.
There is also a governance subtext. Samsung's ownership structure, with significant stakes held by its affiliated insurers, complicates large-scale cancellations; retiring shares would dilute the controlling family's indirect hold unless the affiliates participate proportionately. SK Hynix faces no such constraint, which is one reason it could move faster and harder. The market is now pricing that governance friction into Samsung's multiple, and no amount of operating-profit growth offsets a multiple that refuses to expand.
Second-Order Effect: The AI Trade Is Being Repriced on Cash Conversion
The first-order reading of Wednesday's move is simple: no buyback, stock falls. The second-order implication is more consequential. The AI chip trade, which rewarded capacity announcements and supply shortages for the past eighteen months, is beginning to reward cash conversion and capital-return credibility instead. That is a regime shift within the regime shift.
Consider the sequence. In 2025 and early 2026, memory stocks rallied on the scarcity premium - tight HBM supply, rising contract prices, multi-year data-center buildouts. Samsung's stock rose nearly fourfold. But as the cycle matured, the market's question changed from "how much profit will you make?" to "what will you do with the cash?" SK Hynix answered with cancellations; Samsung answered with a deferral. The 6.6% slide is the market repricing Samsung from a pure AI-supply beneficiary into a company whose capital-allocation credibility now carries a discount.
The transmission channel runs through the discount rate applied to future earnings. A company that returns cash credibly trades at a lower equity risk premium; a company that retains cash for reinvestment must prove that reinvestment earns above its cost of capital. Samsung's bet - that 2027-2028 capacity will be absorbed at favorable prices under long-term agreements - is a bet on continued above-cost returns. If that bet is right, today's selloff is a cyclical overreaction. If it is wrong, the retained cash becomes stranded capacity, and the shares reprice again, lower.
The Counter-Thesis: Samsung Is Right to Wait
The strongest case against the market's verdict is that Samsung's caution is rational, even necessary. The memory industry is entering a capital-intensity phase that resembles the foundry arms race more than the commodity cycle of the past. HBM4 requires new packaging lines, advanced DRAM nodes demand extreme ultraviolet lithography at scale, and customers are signing long-term agreements that effectively pre-sell capacity. In that environment, the company that invests fastest and largest wins the next cycle's share - and the company that distributes its cash today may find itself capacity-constrained tomorrow.
There is also a timing argument. Samsung's August announcement already committed to 90 trillion to 110 trillion won for 2026; the January board meeting is not a retreat, it is the scheduled decision point for the portion that depends on full-year results. Deferring the exact split between dividends and cancellations until the full-year numbers are known is prudent financial management, not a broken promise. And the company has delivered: 19.6 trillion won in regular dividends over 2024-2025, an 8.4 trillion won buyback and cancellation in 2025, and a three-year policy committing 50% of free cash flow to shareholders.
This counter-thesis is coherent. Its weakness is that it asks investors to trust management's capital allocation at the exact moment management has declined to show its hand. Prudence is not a catalyst. And in a market where SK Hynix has set a new benchmark for shareholder returns, Samsung's "we will decide in January" reads less like discipline and more like hesitation.
What to Watch
The falsifying signal is concrete: if Samsung's January 2027 board meeting does not announce a meaningful share-cancellation component - specifically, retirements that reduce the share count by at least 2% - the governance-discount thesis is confirmed and the multiple will not expand regardless of earnings. Conversely, a cancellation commitment of that magnitude would likely re-rate the stock sharply, because it would prove that the cash is coming back rather than going into the ground.
Near term, watch three things. First, whether the semiconductor division's margin holds as capex ramps - a memory operating margin below 70% would signal that pricing power is already peaking, against an expected level near 76% this quarter. Second, whether foreign investors continue selling; the October selloff was driven in part by persistent foreign outflows from the KOSPI. Third, whether the KOSPI itself stabilizes; Samsung's 40% drawdown from its June peak has tracked the broader index's de-rating as much as company-specific disappointment.
Scenarios split cleanly by horizon. In the short term, sentiment is negative: the deferral has created an overhang that will not lift until January, and any softness in the December quarter or in HBM pricing would extend the pressure. Over the medium term, the path depends on the January decision - a credible cancellation plan restores the rerating case; a dividend-only outcome locks in the discount. Over the long term, the structural question dominates: if AI memory demand remains tight through 2027 and beyond as management expects, Samsung's reinvestment will look prescient and today's selloff will read as a cyclical entry point. If the AI capex cycle peaks sooner, the retained cash becomes the story, and the shares underperform regardless of how generous the dividends are.
Samsung's problem is not that it earned too little - it is that in a market now paying for cash returned, not cash earned, a record quarter without a buyback is a quarter the market knows how to discount.
Explore more exclusive insights at nextfin.ai.
