"Storage "super cycle" logic strengthened again! As DRAM demand soars, Samsung DDR5 prices surge by 60%"

Media reports citing informed sources revealed that Samsung Electronics, the world's largest storage chip manufacturer based in South Korea, will significantly raise prices for some crucial storage chips globally this month compared to September. According to sources, Samsung will increase prices for certain storage chips by as much as 60%, due to a severe shortage caused by the exponential demand expansion for these chips brought about by tech giants aggressively constructing AI data centers worldwide. Reported by media citing informed sources, contract prices for 32GB DDR5 storage chip modules in November have surged from $149 in September to $239, and prices for other high-performance DDR5 storage modules have also risen significantly, with increases of up to 60%. By significantly raising prices for certain storage chips, the world's largest manufacturer of DRAM and NAND storage chips, Samsung, is making a bold move to strengthen the importance of the "super cycle" in storage, which is crucial for the global storage chip/product domain. In Google, Microsoft, and Meta-led construction of these super-sized AI data centers, in addition to HBM storage systems built using 3D-stacked DRAM, server-level DDR5 is also a "must-have" and core storage resource that must be procured on a large scale. They complement each other rather than serving as substitutes. The capacity of DRAM clusters supporting sky-high AI training/inference power demands is typically 4-8 times that of traditional CPU servers, with many single machines already exceeding 1TB of DRAM and clearly transitioning to DDR5, mainly because DDR5 offers about a 50% increase in bandwidth, making it more suitable for massive AI workloads. It is worth noting that it is SK Hynix, Samsung, and Micron that have mostly concentrated capacity on HBM storage systems the production capacity and the complexity of advanced process manufacturing and packaging testing required for such storage products are much more complex compared to lower-level storage devices such as the DDR series and HDD series. As a result, the three major leaders in storage chip manufacturing are continually shifting capacity, leading to a situation where DDR5 and even the previous generation DDR4, as well as HDD and enterprise-grade SSD products, are collectively facing a supply-demand imbalance. Recent incredibly strong performances announced by the two global storage giants Samsung Electronics and SK Hynix, as well as Western Digital and Seagate among other storage giants, have led Wall Street giants like Morgan Stanley to proclaim that the "storage super cycle" has arrived, highlighting the continued exponential expansion of demand for DRAM/NAND series storage products driven by the surging demand for AI training/inference power as well as the recovery cycle driven by the consumer electronics demand for edge AI. Samsung raises prices for DDR5 storage chips by up to 60% amid severe supply shortages According to media reports citing informed sources, this move occurred after Samsung Electronics decided to delay the formal announcement of pricing for storage chip supply contracts in October and is in line with the understanding that pricing details are usually announced monthly. Fusion Worldwide's president Tobey Gonnerman stated in media reports that the contract price for the widely used 32GB DDR5 storage chip module quoted by Samsung Electronics in November has surged to $239, far exceeding the $149 in September. These DDR5 storage chips are widely used in AI server clusters, as well as in some high-end personal computing products and other high-performance edge devices. Sources said that the South Korean tech giant has also significantly raised prices for 16GB and 128GB DDR5 storage chip modules by about 50%, to $135 and $1,194 respectively. Additionally, the prices for larger storage capacity 64GB and 96GB DDR5 have risen by over 30%. The media also reported that another group of sources briefed on the pricing information within Samsung's internal circle confirmed the significant price increase. Sources pointed out that the severity of the DRAM storage chip shortage has led to panic buying on a large scale by most enterprise-level customers. Jeff Kim, a senior research director at KB Securities, stated that since Samsung Electronics' transition to HBM storage systems is relatively slower compared to Micron and HBM leader SK Hynix, the world's largest DRAM chip manufacturer, Samsung Electronics, is more optimistic than its competitors in the broader storage chip domain outside of HBM storage systems. This is mainly due to Samsung's larger inventory compared to peers, providing a stronger bargaining power. The media added that Ellie Wang, a senior analyst at the semiconductor consulting firm TrendForce, suggested that Samsung Electronics is likely to raise quarterly contract prices by 40% to 50% between October and December, higher than the average 30% increase expected in the broader storage industry. Analyst Ellie Wang pointed out that the company is confident in the continued significant price increases, mainly due to the incredibly strong demand for storage brought about by the AI data center construction boom, with various enterprises signing long-term storage product supply agreements with storage suppliers. She added that these agreements either cover the entire year of 2026 or span from 2026 to 2027. All AI infrastructure projects involve high-performance storage products in data centers For projects as large as the $500 billion "Interstellar Gate" AI infrastructure project and OpenAI's nearly trillion-dollar AI computing infrastructure agreements, these super AI infrastructure projects are heavily reliant on NVIDIA's AI GPU clusters and enterprise-level high-performance storage products in data centers (including HBM storage systems, enterprise-level SSD/HDDs, server-level DDR5, among others). In this unprecedented AI investment cycle centered around AI model updates and the expansion/construction of AI data centers, core AI power component manufacturers like NVIDIA are undoubtedly the biggest winners; following closely are high-end memory suppliers represented by HBM (SK Hynix, Samsung, and Micron), and enterprise-level high-performance storage manufacturers serving AI data centers (nearline HDDs and data center SSDs). These two chains are driving the dual-wheel AI investment cycle of "AI power x storage," where HBM storage systems are the first storage product tier closely following AI GPU/AI ASIC clusters, and enterprise-level HDD/SSDs are the other major force supporting the AI data storage infrastructure construction frenzy. In the unprecedented "AI power competition" associated with the rapid global expansion and AI training/inference infrastructure, Morgan Stanley and other Wall Street giants have acclaimed the arrival of the "storage super cycle," where enterprise-level hard drive demand surge drives up the stock prices of data storage product giants such as Seagate, SanDisk, and Western Digital by triple-digit percentages this year, significantly outperforming US stock indexes and even global stock markets. Morgan Stanley stated that SK Hynix's "sold-out" signal indicates a tighter supply of storage chip capacity, leading the entire storage chip industry into a so-called "seller's market," where a continuous rise in storage chip prices will persist throughout 2026 and likely extend into 2027. In its latest research report, Morgan Stanley noted that with massive investments by global large-scale enterprises and various government agencies in AI, the demand for core storage chips closely associated with artificial intelligence training/inference systems remains incredibly hot, driving substantial revenue growth in data center storage business, including the HBM storage systems, server-level DDR5, and enterprise-grade SSDs. The duration of this round of the storage "super cycle" is expected to far exceed historical peaks. Morgan Stanley pointed out that the core driver of this round of the storage super cycle has undergone a qualitative change. The demand side is no longer price-sensitive traditional consumer electronics customers, but rather AI data centers and cloud computing service giants engaged in an arms race to build massive AI computing infrastructure, as well as some sovereign government institutions. For them, obtaining storage products including DRAM and NAND series is a strategic "necessity," with price sensitivity reduced to a minimum, while the huge capacity required for HBM continues to structurally encroach on the DDR4 and DDR5 capacity of the three storage giants.