Samsung Electronics reported operating profit of 107.4 trillion won (about $80bn) for July to September on Thursday, up from 12.2 trillion won a year earlier. Almost all of it — 106.5 trillion won, a fifteenfold jump — came from semiconductors. It was the first quarter in which any technology company cleared 100 trillion won in operating profit.

Markets shrugged. The stock opened higher, reversed, and fell as much as 6.2% in morning trade, while rival SK Hynix rose 0.7%.

The reason is that Samsung's record was already known and its forward-looking commentary was not what investors wanted. The company had guided to 107.4 trillion won earlier this month; the chip division's 106.5 trillion came in just below the 107.3 trillion analysts polled by LSEG expected. What moved the stock was everything around the number.

A shorter shortage than the rival. SK Hynix said on Wednesday that its DRAM and high-bandwidth memory output is sold out through 2029 and that the shortage will persist to 2030, while detailing roughly $75bn in shareholder returns for this year. Samsung said its HBM is sold out for next year and that the market will stay tight through 2028 — a year or two less bullish, depending on how you read it. "Samsung's decline was likely attributable to its outlook on how long the memory shortage would last, which came across as marginally less bullish than SK Hynix's commentary yesterday," Morningstar analyst Jing Jie Yu told Reuters.

The returns decision moved to January. Samsung had already committed to 90–110 trillion won of shareholder returns this year, including 30 trillion won of third-quarter dividends. It declined to extend or refresh the policy, saying it will finalise the next framework in January. "The market appeared to expect an additional buyback and/or dividend announcement," said CLSA's Sanjeev Rana.

Capex is enormous and rising. Chip capital spending rose 62% to 76.8 trillion won this year, and Samsung said 2027 will bring another sharp increase, with AI infrastructure capex reaching "a historic level" as demand "will far exceed the industry supply" compared with 2026. It also said new fabs take time, which is why it expects tightness through 2028.

The bill is landing elsewhere in the company. Higher memory prices pushed the mobile division to a loss of 800 billion won. That is the uncomfortable shape of a memory supercycle: the same price surge that produces a record at the chip division is a cost line at the handset business.

The foundry unit remains loss-making, though Samsung reported better revenue on sales of logic dies for HBM and 2nm orders from hyperscalers. Mass production at its Taylor, Texas plant starts next year for Tesla, and the company is weighing another advanced-node fab.

Samsung also said it has long-term supply contracts with ten major customers and is negotiating with ten more, targeting 60–70% of its output under such deals by 2030 — extending beyond hyperscalers into mobile and automotive buyers. That is a hedge against the moment when memory pricing turns, and it is the part of the strategy that outlasts any single quarter.

For now the arithmetic is simple: a company earning $80bn in three months trades at 3.3 times forecast earnings, because investors are not paying for this quarter. They are pricing the length of the shortage, and Samsung just told them it may end sooner than its rival did.

Sources