AI-driven storage demand growth is irreversible; storage chip concepts rebound with volatility Demingli approaches the daily limit

(Source: Finance News)

          The bottleneck in the AI industry chain is shifting from GPUs to storage, packaging, and network bandwidth, with the rigid growth in storage demand expected to persist for at least 3 to 5 years.            

On March 24, the concept of storage chips rebounded in the afternoon, with Demingli (001309.SZ) approaching the daily limit, and Gu’ao Technology (rights protection) (300551.SZ), Shangluo Electronics (300975.SZ), Saiteng Co., Ltd. (603283.SH), Yingxin Development (rights protection) (000620.SZ), and Qipai Technology (688216.SH) also rising.

On the news front, Wedbush analysts reported on Monday that DRAM and NAND storage prices are rising rapidly, predicting that by the fourth quarter of 2025, prices will increase by “triple digits” in the first half of 2026. Among them, DRAM prices are expected to rise by 130% to 150%, and NAND’s increase is close to this level.

The underlying logic of this round of storage price increases is the geometric pull of AI computing power explosion on storage demand, with strong growth on the demand side changing the supply-demand balance in the storage industry. The demand for storage from AI servers is 8 to 10 times that of traditional servers. As global AI large models transition from the training phase to large-scale scenario implementation, the shipment volume of AI servers is experiencing explosive growth, directly consuming a large amount of storage capacity.

Market data confirms the strong momentum of AI demand. According to Gartner’s forecast, global AI server shipments are expected to grow by over 180% year-on-year in 2026, with total shipments exceeding 1.5 million units, corresponding to a total demand for AI storage chips of over 12 million pieces. However, global total capacity in 2025 is only 8 million pieces, leading to a supply-demand gap of over 30%. SEMI data indicates that global sales of chips related to AI servers will reach $169 billion in 2026, a year-on-year increase of 55%, with HBM demand growing particularly significantly.

Tight supply is also an important reason for the current rise in storage chip prices. Currently, the global storage industry’s inventory cycle has fallen to historical lows, with SK Hynix disclosing that overall inventory is only about 4 weeks, at a historically low level. In terms of capacity, leading manufacturers prioritize allocating advanced capacity to high-profit products required by AI data centers, squeezing consumer-grade storage capacity. Kioxia previously stated that its NAND flash memory capacity for the entire year of 2026 has been sold out, and the tight supply situation is expected to persist until 2027. SK Hynix also indicated that its HBM capacity for 2026 has been sold out in advance.

Institutions generally believe that the storage growth driven by AI demand is irreversible. A research report from CITIC Securities stated that due to the strong demand from AI, storage demand will maintain a high prosperity level, and the price increase of storage chips is expected to continue throughout 2026 and continuously catalyze the industry. Morgan Stanley pointed out that the bottleneck in the AI industry chain is shifting from GPUs to storage, packaging, and network bandwidth, with the rigid growth in storage demand expected to last for at least 3 to 5 years.

Massive information, precise interpretation, all available on the Sina Finance APP.