Counterpoint Research projects the global memory market to grow to 2,100 trillion won by 2027, with server products expanding to 57% of the total. /Courtesy of Counterpoint Research

With expanded investment in artificial intelligence (AI) infrastructure, an outlook has emerged that the global memory market will continue its growth through next year. As demand rises for server products such as high bandwidth memory (HBM), the center of the memory market is rapidly shifting to AI servers.

According to the Memory Tracker of market research firm Counterpoint Research on the 23rd, the global memory market is projected to expand from 360 trillion won in 2025 to 1,500 trillion won this year. It was expected to grow to 2,100 trillion won in 2027.

Counterpoint Research saw server memory leading market growth. The share of server products is projected to rise from 37% in 2025 to 56% this year and 57% in 2027. With AI data center investment expanding, demand is increasing in tandem for HBM, DRAM, and enterprise solid-state drives (SSD).

Counterpoint Research viewed memory as the key variable in AI infrastructure investment. As AI models grow and inference services expand, data processing speed, power efficiency, and storage performance have become factors that determine total infrastructure expense.

However, the possibility of price adjustments was raised starting in the second half of 2027. If supply increases in earnest due to new facility expansions, memory prices could correct quickly. Counterpoint Research explained that precise supply-demand management is needed to ensure the current growth does not remain a temporary boom.

Long-term agreements (LTA), custom HBM strategies, and the speed of transition to next-generation processes were cited as key variables for future competition. An LTA is a contract in which the customer and supplier set volumes and conditions for a certain period to ensure stable product supply. Custom HBM refers to tailored memory whose performance and power characteristics are adjusted to match a customer's AI chip architecture.

※ This article has been translated by AI. Share your feedback here.