Orient Securities has released a research report stating that robust AI demand is supporting a continuous rise in storage prices. The firm forecasts DRAM contract prices to increase by 58-63% quarter-over-quarter in Q2 2026, while NAND flash prices are expected to rise by 70-75%. To ensure supply stability, downstream Cloud Service Provider (CSP) customers are showing high acceptance of price hikes and long-term agreements, which is likely to extend the profitability cycle for storage manufacturers. Concurrently, Compute Express Link (CXL) memory pooling solutions are optimizing AI storage efficiency. With accelerated industry adoption, these solutions are poised to reshape the composition of memory hardware and drive rapid growth in the related chip market.
The primary views of Orient Securities are as follows:
Strong AI storage demand is expected to keep storage prices firm. Some investors are concerned that potential pressure on consumer electronics demand could impact overall storage demand and future price increases. However, the firm believes that robust demand from AI applications will sustain strong storage price performance. From a demand perspective, AI inference processes are set to significantly alter data center storage architectures, continuously boosting the need for active data storage. On the supply side, major memory manufacturers are prioritizing bit output for server-related storage products, which is expected to maintain tight supply conditions in the consumer electronics storage segment as well. According to TrendForce, general-purpose DRAM contract prices are projected to grow 58-63% quarter-over-quarter in Q2 2026. The chain reaction of price increases across the entire NAND Flash product line remains strong, with overall contract prices anticipated to rise 70-75% quarter-over-quarter in Q2 2026.
High acceptance of price increases and LTAs from CSPs may extend the profitability cycle for storage manufacturers. Some investors worry that the current high rate of storage contract price increases might lead to slower future growth, potentially affecting storage manufacturers' performance. Orient Securities contends that, given the persistently strong AI storage demand, downstream customers like CSPs have a high tolerance for price hikes and long-term agreements (LTAs), which could prolong the profitable cycle for storage makers. On the supply side, leading international memory giants such as Samsung, Micron, and SK Hynix are prioritizing capital expenditure towards the High Bandwidth Memory (HBM) market. To avoid an oversupply-induced downturn in the general-purpose memory market, their expansion plans for commodity memory are likely to remain limited. Consequently, original equipment manufacturers are expected to continue pushing for LTAs with major clients like large CSPs as a basis for future capacity expansion. Currently, Samsung is engaged in in-depth LTA negotiations with major CSPs including Google and Microsoft, while Micron has already signed its first five-year LTA with a customer. From the customer perspective, with new storage capacity not expected to come online on a large scale until late 2027 or 2028, CSPs are willing to accept higher prices and enter into LTAs to secure stable supply. Looking ahead, these long-term agreements are anticipated to extend the profitability cycle for storage manufacturers.
CXL memory pooling solutions can optimize AI storage system efficiency, and industry adoption progress warrants ongoing attention. CXL memory pooling technology enhances storage system efficiency and expands memory capacity, potentially reshaping the memory hardware composition within AI computing infrastructure. In terms of industry progress, related hardware and software for CXL memory pooling are gradually maturing, with leading companies accelerating their deployments. Among international players, NVIDIA's Vera CPU supports the CXL protocol, and Astera Labs is collaborating with Microsoft to develop CXL-attached memory for Azure M-series virtual machines. Domestically, at the 2025 Yunqi Conference, Alibaba Cloud announced the world's first PolarDB database-dedicated server based on CXL 2.0 Switch technology. Alibaba's Panjiu general-purpose computing hyper-node server can utilize CXL interconnects to provide CPU nodes with a high-bandwidth, high-capacity elastic memory pool. Looking forward, CXL technology penetration is expected to rise rapidly. According to TechInsights, CXL's share of the server DRAM market is projected to grow from nearly zero in 2024 to approximately 15% by 2030. The adoption of CXL solutions is also expected to fuel high growth in the CXL interconnect chip industry. Based on Montage Technology's 2025 annual report, the market size for CXL interconnect chips is forecast to reach $1.7 billion by 2030, with a compound annual growth rate of 170% from 2025 to 2030.
Investment recommendations and targets focus on the strong AI storage demand and the progress of CXL memory pooling solutions. Relevant investment targets include: domestic storage chip design firms such as Gigadevice, Puya Semiconductor, Giantec Semiconductor, Dosin, and Ingenic Semiconductor; companies benefiting from storage technology iterations like Montage Technology, Union Memory, and ASR Microelectronics; domestic storage module manufacturers including Longsys, Netac Technology, and Biwin Storage; semiconductor equipment companies such as AMEC, Jingzhida, Jingyi Equipment, Leadmicro, Piotech, and Naura Technology Group; domestic packaging and testing firms like Kaifa Technology, Huicheng, and Tongfu Microelectronics; and supporting logic chip manufacturers such as Nexchip.
Potential risks include slower-than-expected AI adoption, slower-than-anticipated technological iteration speeds, and delays in domestic supply chain development.
Comments