Nomura Sees AI Memory Market Remaining Tight Despite New Investment Plans

Analysts Dismiss Oversupply Concerns

Nomura believes recent concerns about a slowdown in the artificial intelligence memory market are misplaced, arguing that major investment announcements by South Korean chipmakers are unlikely to create excess supply in the foreseeable future.

The brokerage noted that memory manufacturers and related businesses have unveiled investment plans totaling around 4.8 quadrillion won ($3.5 trillion), with approximately 3.7 quadrillion won allocated to memory-focused projects.

While the announcements sparked concerns about future oversupply, Nomura expects market conditions to remain tight.

High-Bandwidth Memory Continues to Lead Demand

According to the brokerage, demand generated by artificial intelligence continues to outstrip supply, particularly for high-bandwidth memory (HBM), where manufacturers are concentrating production because of stronger profitability.

As a result, supply of conventional DRAM and NAND products is expected to remain constrained.

Nomura also noted that semiconductor manufacturing hubs require years to complete, meaning recently announced projects will have little impact on production over the medium term.

Long Development Timeline Limits Near-Term Supply Growth

The firm pointed to the Yongin Semiconductor Cluster, estimating that the project will not begin limited production until late 2027 despite being launched nine years ago.

That timeline illustrates how major semiconductor investments typically require more than a decade before generating meaningful commercial output.

Meta’s Computing Strategy Could Expand AI Adoption

Nomura also challenged concerns surrounding Meta’s (NASDAQ:META) plans to sell excess computing capacity.

Rather than indicating softer demand, the brokerage believes the move should improve asset utilisation while providing additional computing resources for companies including OpenAI and Anthropic.

Lower computing costs could encourage broader AI adoption, supporting continued investment in memory and other semiconductor technologies.