AI Memory Chip Shortage Likely to Persist Despite Korea Investment Boom, Nomura Says

Concerns that a new wave of investment announcements by South Korean memory chipmakers could trigger another memory chip oversupply cycle may be exaggerated, according to Nomura analysts, who believe strong artificial intelligence (AI) demand will continue to keep the market tight for years.

Recent announcements from South Korean memory manufacturers and related companies outlined long-term investment plans totaling approximately 4.8 quadrillion won ($3.5 trillion), with about 3.7 quadrillion won dedicated to memory semiconductor projects. The scale of the planned spending has fueled investor concerns that aggressive capacity expansion could eventually flood the market with memory chips.

However, Nomura said those fears overlook the industry's current supply constraints and the long development timeline required for semiconductor manufacturing projects.

The brokerage noted that AI-driven demand remains exceptionally strong, particularly for high-bandwidth memory (HBM), which is essential for AI accelerators and advanced data center infrastructure. Memory manufacturers continue to prioritize the production of higher-margin HBM products, limiting available capacity for conventional DRAM and NAND chips and keeping supply tight across the broader memory market.

Nomura also emphasized that the newly announced investment projects are unlikely to have a meaningful impact on global memory supply in the near future. Building semiconductor clusters, fabrication facilities, and supporting infrastructure requires several years before production can begin.

As an example, the firm pointed to South Korea's Yongin Semiconductor Cluster, which was first announced nine years ago but is only expected to begin limited production in late 2027. That timeline highlights how major semiconductor investments typically take more than a decade to translate into significant manufacturing output.

Separately, Nomura downplayed investor concerns surrounding Meta Platforms' (NASDAQ: META) reported plan to commercialize excess computing capacity. Some investors interpreted the move as a sign that AI infrastructure demand could be slowing.

Instead, Nomura argued that selling unused data center capacity is a logical strategy for large technology companies seeking to improve returns on invested capital. The firm said the initiative could also benefit AI developers such as OpenAI and Anthropic by providing access to additional computing resources.

Rather than signaling weaker AI hardware demand, Nomura believes Meta's strategy could encourage broader AI adoption by lowering computing costs, potentially increasing overall demand for AI workloads and reinforcing the long-term growth outlook for the memory semiconductor industry.