SEOUL – Samsung Electronics Co. posted operating profit of 89.4 trillion won ($58.4 billion) in the second quarter of 2026, a 19-fold increase over the same period a year earlier and the largest single-quarter operating profit any technology company has ever reported.
Revenue reached 171 trillion won, more than double the 74.57 trillion won Samsung generated in Q2 2025. The result beat analyst consensus of 84.4 trillion won by roughly 16 trillion won, according to the preliminary earnings guidance Samsung published Tuesday. Full results, including segment-level breakdowns, are scheduled for July 30.
No divisional breakdown accompanied the preliminary release. The driver is not in question. Samsung manufactures the high-bandwidth memory chips that power every major AI accelerator deployed at scale, and demand for those chips has outrun available supply for three consecutive quarters. Tuesday’s result reflects how completely that demand has repriced the company’s business.
Twelve months ago, Samsung posted operating profit of 4.68 trillion won for the quarter, roughly a tenth of what Nvidia Corp. was earning from GPU sales at the time. Tuesday’s figure inverts that relationship. Samsung’s Q2 2026 operating profit exceeds both Nvidia and Apple Inc. on a single-quarter basis, placing the South Korean chipmaker at the top of the technology industry by this metric for the period.
The pricing data makes the structure of the quarter visible. DRAM average selling prices rose 44 percent quarter-on-quarter in Q2, while NAND flash climbed 53 percent, according to Citi Research. These are not incremental adjustments. They are the kind of price movement that happens when a discrete category of buyer is absorbing supply faster than production can expand. AI data center operators are that category. Samsung did not produce this outcome by cutting prices. The AI buildout requires high-bandwidth memory at a scale that did not previously exist, and Samsung is one of three companies capable of making it in volume.
The second is SK Hynix Co. (000660.KS), which has been Samsung’s most direct competitive pressure throughout the AI memory cycle. Hynix captured early orders from major AI customers by achieving higher yield rates on HBM3E stacks, the advanced packaging format Nvidia’s H100 and H200 accelerators require, before Samsung had reached equivalent production consistency. The Q2 result suggests Samsung has closed that yield gap at commercial volumes. What the full July 30 report will clarify is the HBM generation mix in Samsung’s Q2 output, since HBM3E commands significantly higher margins than HBM2E.
Samsung’s Tuesday announcement noted that the company plans to construct additional fabrication facilities to meet rising demand, without specifying location, timeline, or capital commitment. The near-term bottleneck in memory supply is not wafer starts. It is the yield qualification process for advanced HBM stacks, which is inherently time-limited regardless of available fab capacity. More facilities address the medium-term gap analysts expect to emerge as AI inference extends from data centers into edge and consumer deployment.
What Tuesday’s figure also establishes is the degree to which the memory business has detached from the broader semiconductor cycle. In prior downturns, memory pricing followed the inventory correction that tends to accompany oversupply, with Samsung’s earnings contracting sharply. The current cycle has produced no such correction because AI data centers do not behave like consumer electronics buyers. They are purchasing at the maximum rate their construction timelines permit, not at the maximum rate their current workloads require. That distinction matters for forecasting: it means Samsung’s pricing environment could hold even as overall memory supply expands, so long as AI capital spending remains at the levels currently budgeted by major cloud providers.
Samsung’s quarter comes two weeks after Micron Technology’s record quarterly figures showed the same dynamic across the full memory supply chain, with Micron management stating it had no visibility on when demand would ease. Earlier this year, Samsung narrowly avoided a production disruption when its largest union suspended a planned strike involving nearly 48,000 workers. The resolution kept South Korean fabs at full utilization through the quarter that produced Tuesday’s result.
The one division unlikely to have contributed meaningfully to the headline figure is foundry. Samsung’s contract chipmaking business, which competes with Taiwan Semiconductor Manufacturing Co. on leading-edge process nodes, has been ceding advanced orders and faces cost pressure from bonus expense allocations across the broader semiconductor segment. The preliminary guidance provides no basis for separating foundry drag from memory margin. That calculation waits for July 30.
Full-year projections built on Q2’s result and the trajectory of AI infrastructure spending suggest Samsung could generate cumulative 2026 profit that surpasses its total semiconductor earnings from the past four decades, according to executives who addressed investor meetings earlier this year. Nikkei Asia reported that Samsung plans to accelerate capacity expansion as the memory shortage shows no resolution before 2027. The demand that produced $58.4 billion in operating income over three months did not emerge from a speculative cycle. It emerged from AI inference deployment at scale, and that deployment is still accelerating.
