Citi Sees Long-Term Opportunity in Memory Chip Stocks Despite Recent Pullback

Citi believes the recent decline in memory semiconductor stocks represents a potential buying opportunity, arguing that the artificial intelligence-driven memory cycle remains in its early stages and could ultimately surpass the industry's previous expansion during the early 2000s.

According to the investment bank, today's market shares several characteristics with the NAND memory boom between 2001 and 2007, when the rapid adoption of products such as digital cameras and MP3 players fueled sustained demand for flash memory while gradually replacing older storage technologies.

AI Demand Continues to Support the Sector

Shares of major memory chip manufacturers, including Micron Technology, Samsung Electronics, and SK Hynix, have retreated more than 20% from recent highs after a strong rally earlier this year. The pullback has been driven largely by investor concerns over elevated valuations and questions surrounding the pace of artificial intelligence-related capital spending.

Despite the recent weakness, Citi maintains that the long-term outlook for the industry remains favorable.

The bank believes the current cycle has the potential to outperform the memory market expansion experienced between 2001 and 2007, largely because AI applications are increasing demand for both DRAM and NAND memory technologies.

Long-Term Agreements Point to Continued Growth

One factor supporting Citi's positive outlook is the growing number of long-term supply agreements being signed across the industry.

According to the bank, customers are increasingly committing to contracts lasting three to five years, a trend that suggests confidence in sustained demand and provides greater visibility into future revenue and earnings for memory manufacturers.

High-Bandwidth Memory Remains a Key Focus

Citi also highlighted the continuing shortage of high-bandwidth memory (HBM), a critical component used in advanced artificial intelligence systems.

The bank expects supply constraints to influence how AI infrastructure is deployed over the coming years. Rather than relying exclusively on a smaller number of graphics processors equipped with larger amounts of HBM, developers may increasingly expand computing capacity by deploying a greater number of GPUs with more moderate memory configurations.

Even under that scenario, Citi expects overall HBM capacity within AI systems to grow significantly.

Its projections estimate total HBM capacity per AI system could increase by approximately 434%, rising from 20.7 terabytes to 110.6 terabytes, while the number of GPUs within those systems expands from 72 to 576.

Positive Outlook for SK Hynix

Citi also reiterated its constructive view on SK Hynix, noting comments from the company's second-quarter earnings conference call indicating management is evaluating additional initiatives aimed at enhancing shareholder returns.

Overall, the bank believes that although memory semiconductor stocks have recently experienced increased volatility, the structural growth driven by artificial intelligence continues to provide a favorable long-term investment outlook for the sector.