Samsung Electronics Co. crossed a line on Oct. 8 that no quarter of its history had ever reached. In its preliminary guidance for the third quarter, the company said operating profit rose 782.5% from a year earlier to 107.4 trillion won, the first time a single quarter’s profit has topped 100 trillion won. Sales climbed 126.6% to 195 trillion won.
The engine was semiconductors. The division’s revenue is estimated to have surpassed 100 trillion won for the quarter, with the company’s newest high-bandwidth memory, HBM4, ramping in step with Nvidia’s Vera Rubin platform. Samsung said its HBM shipments rose nearly 50% from the previous quarter, the clearest sign yet that the newest memory has moved from qualification to volume.
The contrast inside the company is sharp. Its device business, which groups phones and home appliances into what Samsung calls the DX division, lost more than 1 trillion won for the quarter, according to the guidance. Samsung is, in effect, two companies at once: one riding an AI chip boom, the other fighting for share in consumer markets that have stopped growing.
Cumulative operating profit for the first nine months reached 254.13 trillion won. Reuters described the quarter as the highest single-quarter profit by a technology company on record. Samsung said it would report divisional detail on Oct. 29, when the guidance becomes audited results.
The numbers are a product of a market in which demand for AI infrastructure has outrun the supply of memory chips. That shortage is expected to persist into 2027, according to analysts and regulators, which means the companies that can make the most advanced memory are pricing from a position of unusual strength.
Samsung has spent years trying to prove it can keep pace in high-bandwidth memory. It trailed SK Hynix in the earlier generations of the technology and has been racing to catch up as the industry moves to HBM4. Winning a place in Nvidia’s next platform was treated inside the company as a matter of survival, and this quarter’s shipment jump suggests that effort is beginning to pay.
The memory business is the reason Samsung exists as the company the world knows. Its DRAM and NAND chips are the standard parts inside servers, phones and data centers, and the firm’s fortunes have tracked the memory cycle for decades. What makes this cycle different is the customer driving it: the AI accelerator makers whose appetite for memory is growing faster than the industry’s ability to add capacity.
That concentration is the risk. A quarter this large rests heavily on a single customer ecosystem and a single product class. Memory has always been cyclical, and the cycle has rarely run this hot. If AI spending cools, the boom that lifted Samsung’s profit could reverse just as quickly, a pattern the company’s own history has demonstrated more than once.
The memory cycle has a long memory of its own. In past booms, Samsung and its rivals overbuilt, and the oversupply that followed crushed prices for years. The company’s executives have argued that this time is different, because AI demand is more durable than the PC and smartphone cycles that drove earlier expansions. Whether that argument holds is the central question behind the company’s guidance, and behind the market’s willingness to pay for it.
The device division is the counterweight. Samsung’s phones and home appliances once carried the company through weak chip years; now the relationship is reversed, and the DX division’s losses are a drag the chip boom must overcome. Fixing that business, in a market where smartphones have stopped growing and Chinese rivals keep cutting prices, will take more than a strong quarter in memory.
HBM4 itself is the near-term test. The company has staked its recovery on supplying the newest memory to Nvidia and its rivals, and the near-50% jump in shipments suggests the product is now shipping in volume. But qualification and volume are different things, and the market will be watching whether Samsung can keep that ramp going into next year without giving back the share it has won.
The guidance is preliminary, and the gap between guidance and final results has occasionally been wide enough to matter. Samsung has learned to be cautious with early numbers, and investors will wait for the divisional detail at the end of the month before treating the figures as final.
For now, the direction is not in doubt. Samsung’s decade has been defined by its memory business, and memory has rarely been this valuable. The question is what the company does with the cash while the boom lasts, and whether its device division can stop bleeding while the chip division carries the whole enterprise. A company this dependent on one product cycle is only as durable as that cycle, and the market is now priced for it to keep going.
