[Edaily Reporter Hyera Lee ] On the 15th, KB Securities maintained its “Buy” rating and target price of 4.2 million won for SK hynix(000660), forecasting that the memory supply shortage resulting from expanded investment in artificial intelligence (AI) data centers will continue until at least 2028.

Kim Dong-won, Head of Research at KB Securities, stated, “Considering the future investment plans of Big Tech companies and the explosive demand for AI, it is highly likely that the memory supply shortage will persist until at least 2028,” adding, “The recent decline in stock prices is due to psychological concerns, and excessive worry could actually present a buying opportunity.”
Kim explained, “Next year, as capital investment focuses on high-bandwidth memory (HBM), the expansion of new production capacity for general-purpose memory is expected to be effectively limited.” He added, “Long-term supply agreements (LTAs) between Big Tech companies and memory manufacturers, which will begin in earnest next year, will have the effect of securing production capacity on a first-come, first-served basis. He explained, “The memory supply shortage that general consumers will experience in 2027 will effectively approach zero.”
He also dismissed recent concerns about Meta scaling back its AI investments. Director Kim assessed, “The recent concerns about Meta scaling back its investments are merely short-term noise.” He noted, “Following the construction of a 7-gigawatt (GW) AI data center this year, Meta plans to add another 7 GW next year, securing a total of 14 GW of AI computing infrastructure by then,” and “The Hyperion data center currently under construction in Richland Parish, Louisiana, is also expected to be expanded to at least 5 GW, and the investment scale is projected to increase from the initial $10 billion (approximately 15 trillion won) to over $50 billion (approximately 75 trillion won),” he said.
He also predicted that the pace of investment in AI data centers would accelerate further. He analyzed, “With the U.S. Federal Energy Regulatory Commission (FERC) approving a fast-track process to streamline grid connection procedures—which had been the biggest bottleneck in AI data center construction—the time required for grid connection, which previously took more than five years, is expected to be reduced to one to two years,” adding, “As U.S. Big Tech companies expand their investments in their own power generation facilities, they will be able to bring forward the timeline for building AI data centers by more than double compared to before.”
Kim added, “Although SK hynix’s stock price recently fell 36% from its previous high in just three weeks due to concerns over a slowdown in AI investment, the long-term growth trajectory of the AI industry and the memory supply-demand environment have not fundamentally changed compared to a month ago,” noting, “The recent stock price decline could actually present a buying opportunity.”