Yonhap News
Samsung Electronics (005930.KS) and SK hynix (000660.KS) both weakened in the wake of a plunge in U.S. semiconductor stocks and concerns over rising interest rates. Yet brokerages maintained their optimism, raising price targets or holding existing bullish forecasts. Contrary to worries about the end of the artificial intelligence (AI) investment cycle that emerged after Broadcom’s earnings release, the prevailing view emphasizes the possibility of continued memory demand growth and supply shortages driven by the spread of agentic AI.
On Monday, NH Investment & Securities raised its price target for Samsung Electronics to 530,000 won from 490,000 won, and for SK hynix to 3.2 million won from 3.1 million won. Mirae Asset Securities and SK Securities also maintained their existing price targets in reports the same day, assessing the recent share price correction as an opportunity to increase positions and buy. They did not lower their expectations despite an anticipated plunge following the 10.26% crash in the Philadelphia Semiconductor Index on Friday, which came amid the possibility of rate hikes triggered by strong U.S. employment data.
That day, Samsung Electronics closed at 295,500 won, down 10.18% from the previous session, giving up the 300,000 won level on a closing basis for the first time in six trading sessions. SK hynix also finished at 1.911 million won, down 7.68%. Samsung Electronics has fallen 18% since the 4th, while SK hynix has dropped 19% since the 2nd.
Nevertheless, NH Investment & Securities forecast that as the spread of agentic AI gains momentum, the memory capacity installed in server and edge devices will increase, and inference demand for processing actual tasks will also rise accordingly. “In the agentic AI era, the importance of the central processing unit (CPU) is being highlighted even more, and the DRAM capacity required for a single AI inference server will increase about twofold compared to existing general-purpose servers,” said Ryu Young-ho, a researcher at NH Investment & Securities. “As the price gap between general-purpose memory and high-bandwidth memory (HBM) has recently widened, justification for further HBM price increases in 2027 has also been secured.”
Regarding the reduced memory installation in Nvidia’s next-generation platform “Vera Rubin,” which recently shook the market, the dominant view interpreted it instead as evidence of a supply shortage. The assessment is that it is not a simple slowdown in demand, but the result of reallocating product lineups amid limited memory supply. “Semiconductor stocks fell sharply due to anxiety that Nvidia’s SOCAMM2 adoption would be cut roughly in half, but this is not a decline in demand but a strategic choice to secure sales volume,” stressed Han Dong-hee, a researcher at SK Securities. “The structural memory bottleneck and earnings strength in the AI era are not values that change in the short term.”
Brokerages attribute the cause of this plunge to profit-taking following an overheated share price rally and to interest rate and exchange rate variables, rather than to a slowdown in the AI investment cycle. As a result, the prevailing view is that it should be treated as a bargain-buying opportunity rather than a panic sell. “The share prices of Samsung Electronics and SK hynix are undergoing a severe correction, but the spot price of DRAM, which represents the memory market conditions, has been calm,” said Kim Young-gun, a researcher at Mirae Asset Securities. “With long-term agreements (LTAs) linked from Big Tech to chipmakers, it is reasonable to treat this noise as an occasion to increase positions.”