Samsung's memory profit surges as mobile losses deepen

South Korean conglomerate Samsung Electronics reported Q3 operating profit of KRW107.4 trillion (US$80 billion), with 99% coming from memory as AI infrastructure demand absorbed nearly all the high-bandwidth memory chips it could produce. The result was broadly in line with analyst expectations, yet Samsung shares fell 4.1% in Seoul trading on Thursday.

Device Solutions, which covers the memory and foundry businesses, generated sales of KRW147.1 trillion ($110 billion), up 15% from the previous quarter and 344% from a year earlier. The scale of that increase against the group's wider performance is hard to blur: memory supplied 99% of operating profit, while the mobile and networks division lost money.

CFO Sooncheol Park said on Thursday's earnings call that memory demand next year will "far exceed supply due to robust AI investment." He sees no pause in demand for advanced chips despite rising concern about the debt created by the AI rollout, putting Samsung's forecast on the opposite side of that debate.

A company statement projected that AI capital expenditure will "reach a historic level as the AI ecosystem expands" in 2027. It also expects AI-driven demand to "far exceed the industry supply in comparison with 2026." Those are management expectations, not independent market findings, and they depend on AI investment staying robust.

How much is Samsung investing?

Samsung spent KRW57 trillion ($42.6 billion) on capital expenditure during the first nine months of this year, 77% more than in 2025. For the full year, the company expects total investment likely to reach KRW84.4 trillion ($63 billion), a 60% increase.

The nine-month outlay already represents about two-thirds of that annual forecast. Still, "likely" is doing useful work here: Samsung has set an aggressive investment path, not a guaranteed spending schedule. The investment case will stand or fall on whether demand holds up against the debt burden surrounding the AI rollout.

Why did mobile devices and networks lose money?

The same memory demand is now squeezing Samsung's handset and network businesses. Higher component costs, caused mostly by the memory boom, left the mobile device and networks division with an operating loss of KRW800 billion ($598 million) in its second consecutive quarterly loss.

Revenue in that division grew 11%, so costs rather than a sales contraction drove the loss. A year earlier, Samsung reported an operating profit of KRW3.6 billion ($2.7 million) in the unit, making the latest result a sharp reversal.

Network-equipment sales fell by a third year on year to KRW400 billion ($300 million), diverging from the wider division's revenue growth. Daniel Araujo, head of the mobile experience division, expects revenue to decline sequentially in Q4 even as smartphone shipments rise. More phones, apparently, do not automatically mean more money for their manufacturer.

For comparison, memory produced 99% of Samsung's operating profit, while network equipment generated sales of only KRW400 billion. The comparison captures the quarter: one business generated nearly all operating profit, while the other lost KRW800 billion.

What did currencies do to the bottom line?

Park also said "record exchange rates" for the US dollar, euro and other currencies reduced Samsung's bottom line by KRW10 trillion ($7.48 billion). The phrase is unusual, but the figure is substantial: currency effects can move the bottom line even when the operating business posts a bumper profit.

What is VMO2 testing in London?

Separately, Virgin Media O2 has installed an Nvidia processor at a busy Nokia site in central London to see whether it offers genuine AI benefits. The operator is using the deployment as it judges Ericsson and Nokia's AI-RAN offerings, giving the vendor comparison a live network test.

The next checkpoints are Q4 revenue, smartphone shipments and Samsung's progress toward its likely KRW84.4 trillion full-year investment. Management also expects next year's memory demand to "far exceed supply," while its 2027 AI-spending projection rests on AI investment remaining robust. That is the forecast; the chip cycle will supply the verdict.