The regulatory filing, published Aug. 14, is a snapshot of a company reinvesting every advantage it has: Samsung Electronics spent 27.3 trillion won, about $19 billion, on research and development in the first half of the year, up 51.5 percent from a year earlier and the highest half-year figure in the company’s history. Capital expenditure of 28 trillion won also set a record.
The numbers are remarkable even by Samsung’s standards, and they explain the strategy behind them. Memory chip prices have tripled this year, driven by the AI boom’s appetite for DRAM and high-bandwidth memory, and Samsung is pouring the windfall back into capacity and technology. The company that dominates the memory market is betting that the current cycle lasts long enough to pay for the next one.
The R&D spending reflects a company under competitive pressure from two directions at once. SK Hynix has taken the lead in high-bandwidth memory, the most profitable segment of the AI memory market, and Samsung has been working to close the technology gap with its own stacked architectures. At the same time, the company is trying to hold its foundry business together against TSMC, which dominates contract chip manufacturing. Both fights require research money, and the filing shows Samsung is not holding back.
The capex figure of 28 trillion won funds a familiar pattern: new fabrication capacity, packaging lines and the expansion of the company’s most advanced nodes. Samsung has said it will keep building through the cycle rather than hoarding cash, a strategy that has defined its approach to memory for decades. The company built aggressively during past booms and emerged stronger each time, though each boom also produced excess capacity that dragged prices down afterward.
The memory cycle that is funding all of this is one of the strongest in the industry’s history. Prices for DRAM and NAND have more than tripled from their lows, according to industry trackers, as AI servers consume memory in unprecedented volumes and suppliers kept capacity tight. The price recovery has been so complete that memory has returned to being Samsung’s profit engine, and the first-half results would have been far smaller without it.
Samsung’s response to the boom has been to spend like it expects it to continue. The company’s executives have said publicly that AI demand for memory will outlast the current product cycle, pointing to the number of data centers under construction and the memory content per server, which has multiplied as AI models grow. The bet is that the industry’s historical boom-and-bust pattern has changed, because the demand is coming from infrastructure rather than from consumer devices.
The R&D number also carries a message about Samsung’s ambitions beyond memory. The company is spending heavily on artificial intelligence features for its own products, from smartphones to appliances, and on the semiconductor technologies that will define the next decade, including advanced packaging and the materials that will replace today’s chipmaking processes. The 51.5 percent increase shows the company treating the current windfall as a once-in-a-generation opportunity to invest its way into the future.
Investors have generally supported the spending, judging by the stock’s performance this year, though some analysts have cautioned that the investment pace assumes the memory boom holds. A downturn would leave Samsung with capacity and research programs built at the top of the cycle, the same risk the company has run in every previous boom. Samsung’s own history suggests it would survive such a downturn; its balance sheet has carried it through every memory crash since the 1980s.
The competitive context makes the spending partly defensive. SK Hynix is expanding at a record pace of its own, and Micron is adding capacity with government support in the United States. Samsung cannot afford to let its technology lead slip in high-bandwidth memory while its rivals build, and the research spending is the way it holds its position. The company’s engineers are said to be working on the next generation of stacked memory, with volume production expected in coming years.
The price recovery has a supply-side explanation that supports Samsung’s confidence. Memory makers spent the downturn cutting capacity and idling fabs, and the industry has been slower to restart than in past cycles, in part because the most advanced capacity requires long lead times and enormous investment. Samsung’s own construction program is part of that slow response, which means the current pricing environment could persist longer than previous booms. The company’s spending is both a bet on that persistence and a hedge against it.
The filing’s two numbers, 27.3 trillion won in R&D and 28 trillion won in capex, together describe a company that has chosen its strategy: spend the cycle’s profits on the next cycle. Samsung has done this before, and it has usually worked, because the memory industry rewards scale and technology more than it rewards restraint. The risk is that this cycle, with its AI-driven demand and its government-backed competitors, behaves differently from the ones before it. Samsung is betting that it does not.
