SK Hynix Price Targets Diverge by 2.5 Times as AI Spending Slowdown Fears Split Views

The gap in brokerage target prices for SK Hynix and Samsung Electronics has widened to more than twofold. The split reflects diverging views on whether the memory-chip boom can continue as concern grows over a slowdown in artificial intelligence spending.

As of Nov. 11, Morgan Stanley's target price for SK Hynix stood at 1.6 million won ($1,150), while Kyobo Securities and SK Securities set theirs at 4 million won ($2,880), according to industry data. That amounts to a gap of as much as 2.5 times. For Samsung Electronics, the range ran from Morgan Stanley's 309,000 won ($222) to Yuanta Securities' 690,000 won ($497), a spread of as much as 2.2 times.

Morgan Stanley recently slashed its target prices. It cut SK Hynix to 1.6 million won from 2.6 million won, a reduction of about 38.5%, and lowered Samsung Electronics to 309,000 won from 381,000 won, down about 18.9%. The revisions reflected rising uncertainty over AI data-center regulation after the U.S. midterm elections and growing concern about semiconductor profitability.

Morgan Stanley projects memory supply will increase once new chip production facilities begin operating by the end of next year. It also expects the pace of AI data-center construction to slow. The bank further factored in the possibility that a 5-gigawatt data-center project slated for 2028 could be pushed back to 2029.

"The most important debate in the memory-chip market right now is whether the industry can avoid a '2028 earnings cliff,'" Morgan Stanley analyst Shawn Kim said. Concern is growing less around next year's earnings themselves and more around whether growth can be achieved and sustained after 2028.

Domestic brokerages argue that the imbalance in memory supply and demand will not be easily resolved even if supply increases. Shinhan Securities projects D-RAM supply growth of 21% next year and NAND flash growth of 23%, which it says would still fall short of meeting demand. It also forecasts demand for high-bandwidth memory, or HBM, will rise 69% to 59 billion gigabits next year from 35 billion gigabits this year.

Some analysts also say expanding long-term supply agreements, or LTAs, could prevent the kind of sharp industry downturn seen in the past. KB Securities said the likelihood of a sharp drop in AI infrastructure investment by 2028 remains low, given the data-center projects currently under way.