Samsung’s Quarterly Profit Jumps Nearly Ninefold on AI Memory
Preliminary operating profit beat expectations as constrained DRAM, NAND and high-bandwidth memory supply lifted pricing.
Samsung Electronics projected third-quarter operating profit of 107.4 trillion won, about $80.2 billion, nearly nine times the 12.17 trillion won earned a year earlier. The preliminary result exceeded an LSEG SmartEstimate of 106.1 trillion won and would mark a fourth consecutive quarterly profit record.
Revenue is expected to rise 127% year over year to 195 trillion won for the July–September period. Samsung will publish full results, including divisional detail, later in October, so the current release establishes the scale of the rebound but not yet the precise contribution of memory, foundry, mobile devices and consumer electronics.
The core driver is the AI infrastructure cycle. Demand for high-bandwidth memory has grown alongside accelerator deployments, while conventional DRAM and NAND supply has remained tight. Those markets are connected: manufacturers allocate capital and production capacity toward higher-value products, which can restrict output elsewhere and lift pricing across the memory complex.
Samsung’s result shows how quickly semiconductor operating leverage can reverse. Memory manufacturing has high fixed costs, so falling prices can erase profit even when factories keep running. When supply tightens and pricing improves, incremental revenue can flow rapidly to earnings. That dynamic explains why the profit increase is far larger than the revenue increase.
The figures also matter for the competitive balance with SK Hynix, Micron and advanced packaging partners. High-bandwidth memory requires more than commodity wafer capacity: yields, stacking, thermal performance and qualification with major accelerator customers determine who captures the highest margins. Samsung’s preliminary total does not reveal whether market-share gains or broad pricing did most of the work.
For AI developers and cloud groups, the result is evidence that component scarcity remains expensive. Memory is a growing share of system cost and a possible constraint on data-centre deployment. Higher supplier profits should finance capacity additions, but new fabrication and packaging capability takes time, and aggressive expansion can eventually recreate the industry’s familiar oversupply cycle.
Investors should treat the release as guidance rather than audited final accounts. The company may revise figures, and the absence of divisional data makes it difficult to separate sustainable operating improvement from inventory effects, currency movements and short-term price spikes. The reported expectation that tightness could persist into 2027 is an industry view, not a guarantee.
The extraordinary year-on-year comparison also reflects a depressed base. Memory earnings were weak a year ago, so the 783% increase should not be extrapolated. A more useful test will be sequential pricing, shipment growth and margins once the detailed accounts arrive.
Capital allocation now becomes a key question. Record profit gives Samsung room to invest in leading-edge memory and packaging, reward shareholders and absorb foundry losses. Spending too slowly could surrender AI share; spending too quickly could recreate excess supply just as competitors add capacity.
Customers will also seek longer supply agreements to reduce volatility. Those contracts can improve visibility for Samsung while locking buyers into volumes and prices that may look expensive if the cycle turns. The negotiation balance currently favours suppliers, but semiconductor cycles rarely stay one-sided indefinitely.
Why it matters
Samsung’s profit surge converts the abstract AI-capacity boom into supplier cash flow. It strengthens the case for continued semiconductor investment while warning customers that memory—not just processors and electricity—may remain a critical bottleneck.