Samsung Electronics said second-quarter operating profit rose 1,813.8% from a year earlier to a record 89.49 trillion won, driven almost entirely by its semiconductor division as demand for high-performance memory used in AI data centers accelerated. Revenue reached 171.5 trillion won, up 130% year on year, while net profit climbed 1,299.9% to 71.62 trillion won, exceeding the 68.36 trillion won estimate cited in the company’s earnings materials.
The results place Samsung’s memory business at the center of an intensifying contest for data-center hardware. Device Solutions, the division housing Samsung’s semiconductor operations, generated 89.2 trillion won in operating profit—more than 99% of the group total—while record memory sales and higher-value HBM products lifted profitability far beyond that of its consumer electronics businesses.
Samsung reported an operating margin of 52.2% for the quarter and a net margin of 41.8%, unusually high levels for a company with major businesses spanning chips, smartphones, displays, appliances and automotive technology. Earnings per share rose to 10,849 won from 737 won a year earlier and increased by 3,726 won from the first quarter.
Semiconductor profits eclipse other Samsung businesses
Device Solutions reported quarterly revenue of 127.5 trillion won, up 56% from 81.7 trillion won in the preceding quarter. Its operating profit rose 66% quarter on quarter from 53.7 trillion won and was more than 250 times higher than a year earlier, according to Samsung.
Memory sales reached a record 120.8 trillion won, with Samsung reporting record shipments of both DRAM and NAND flash memory. DRAM is widely used as working memory in servers, while NAND is commonly used in enterprise solid-state drives that store the vast datasets required for cloud computing and AI applications.
Samsung said it had completed mass production of HBM4, the latest generation of high-bandwidth memory, and started volume deliveries to major customers. HBM stacks memory chips vertically to increase data-transfer speeds while reducing the physical space required on advanced processors. It is a critical component for AI accelerators, which need rapid access to large pools of data during training and inference workloads.
The company also said it shipped HBM4E samples described as an industry first, alongside increased output of DDR5 memory, SOCAMM2 modules and enterprise SSDs. Samsung did not identify the customers receiving HBM4 deliveries.
The numbers show how sharply the economics of Samsung’s chip operations have diverged from its consumer-facing divisions. Device Experience, which includes smartphones, televisions and home appliances, posted an operating loss of 0.8 trillion won. Its mobile business produced 32.3 trillion won in revenue, up 14% from a year earlier, but recorded an operating loss of 0.7 trillion won.
Samsung’s Visual Display and Digital Appliances operations generated 14.5 trillion won in revenue and an operating loss of 0.01 trillion won, effectively near break-even. Samsung Display recorded 7.5 trillion won in revenue, up 17% year on year, and 0.7 trillion won in operating profit, compared with 0.4 trillion won in the previous quarter. Harman, Samsung’s automotive technology subsidiary, reported revenue of 4.6 trillion won and operating profit of 0.4 trillion won.
AI hardware demand reshapes the supply chain
Samsung’s results reflect a supply chain in which leading memory capacity is increasingly directed toward server and AI customers. The company raised quarterly research and development spending to 16 trillion won, up 41% from the first quarter, while capital expenditure reached 14.1 trillion won.
That spending suggests Samsung is using the current earnings surge to reinforce production and technology capacity rather than simply treating the quarter as a short-lived upcycle. HBM and advanced server-memory manufacturing are difficult to expand quickly: production requires advanced packaging, close coordination with processor designers and stringent performance testing. Those constraints can keep prices elevated even as manufacturers increase output.
The pressure on data-center supply is relevant to digital-asset infrastructure, though the connection is more indirect than it may initially appear. Bitcoin miners generally rely on application-specific integrated circuits, or ASICs, rather than HBM4, DDR5 or enterprise SSDs. A boom in AI memory does not mean that miners are directly competing for Samsung’s newest memory modules.
Competition can emerge elsewhere in the infrastructure stack. Large AI facilities require land, substations, transmission access, cooling equipment and long-term electricity contracts—the same categories of physical infrastructure that shape mining economics. In regions with constrained power availability, data-center expansion could make cheap industrial electricity harder to secure for operations that depend on low and predictable energy costs.
The supplied Kapuściński data projects that new facilities will consume more than 500 terawatt hours of global electricity by the end of the year. The figure points to the scale of power demand associated with expanding compute infrastructure, although electricity-market effects will vary widely by grid, country and local generation mix.
Mining operators face a separate efficiency test
Bitcoin network processing power recently approached 1,000 exahashes per second, according to the supplied network data, marking a new high for the network’s aggregate hash rate. A rising hash rate generally reflects more machines competing to produce blocks, and the protocol’s difficulty adjustments reduce the share of block rewards available to less efficient equipment.
Older mining fleets are therefore likely to face pressure from two directions: greater network competition and potentially more contested power markets in selected locations. Operators with modern ASICs, efficient cooling systems and fixed-rate electricity agreements would be better positioned than those exposed to volatile wholesale power prices.
The risk of forced asset sales by highly indebted mining operators depends on Bitcoin’s market price, the timing of difficulty adjustments, equipment efficiency and individual financing terms. Samsung’s earnings do not establish that such sales are imminent. They do illustrate the scale of capital moving toward AI-oriented computing infrastructure, where memory suppliers are capturing an unusually large share of the current spending cycle.
For Samsung, first-half consolidated revenue reached 305.37 trillion won and operating profit totaled 146.73 trillion won, increases of 98.7% and 1,191.5% respectively from a year earlier. Operating cash flow rose to a reported record of 105.1 trillion won, while cash and cash equivalents stood at 190 trillion won at the end of June.
Samsung spent 5.6 trillion won on treasury-share buybacks and paid 6.2 trillion won in dividends during the quarter. Its net cash position reached 167.6 trillion won, nearly double the 86.7 trillion won reported a year earlier, giving the company substantial capacity to fund the costly race for advanced memory production.
Riding Samsung’s AI-chip boom? Deepen your strategy with our focused guide on top 2026 utility altcoins poised for tech-driven growth.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
