▲ The SK Hynix headquarters located in Icheon, Gyeonggi Province
SK Hynix has dismissed concerns over a potential oversupply of memory chips, forecasting that global artificial intelligence (AI) infrastructure investment momentum will remain robust moving forward.
During its third-quarter earnings conference call today (October 28), SK Hynix addressed market concerns regarding the sustainability of AI demand, stating, "We view this as a process where customers are examining the feasibility and profitability of individual projects more closely while solidifying the timing and pace of their investment execution."
The company added that AI infrastructure investments are translating into actual business results and that the pool of investment entities is structurally diversifying, leading it to believe that AI investment and demand momentum will continue.
"Customers' AI investments are actually leading to the expanded use of AI services, which is reflecting in business performance," SK Hynix stated. "Major hyperscalers announced in their recent earnings releases that cloud division revenue grew by roughly 30% to 80% year-on-year, and we believe a significant portion of this is driven by the expansion of AI businesses."
SK Hynix also emphasized that major customers' AI-related revenue growth and financial performance are being confirmed through various indicators, such as a rise in long-term agreements (LTAs) for large-scale enterprise AI solutions, a roughly two- to five-fold increase in cloud order backlogs year-on-year, and a steady rise in paid AI service users.
Furthermore, SK Hynix stated, "Even if a specific entity adjusts its investment pace, the structure is shifting so that demand expansion in other areas can offset it."
While initial AI investments were concentrated on a small number of big tech companies—creating a structure where an investment adjustment by a specific firm directly tied to overall demand volatility—the company explained that the investment base is now rapidly broadening beyond cloud service providers to include the establishment of independent AI infrastructure by general enterprises, sovereign AI investments by various countries, and on-premise demand.
"Accordingly, the current structural supply-demand imbalance in the memory market, where demand outpaces supply, will persist for the time being," SK Hynix added.
In the case of SK Hynix, combining already secured LTA volumes and demand currently under additional discussion, DRAM and high-bandwidth memory (HBM) supplies are virtually sold out for the three-year period through 2029.
(Photo: Yonhap News)
※
Copying, redistribution, and unauthorized use in AI training are strictly prohibited.