DDR4 Prices Surge 1,845%, While DDR5 Jumps 465%

Memory prices appear to have no upper limit — only further upside. While contract prices set by original manufacturers still follow relatively defined ranges, spot market prices downstream have become highly volatile, often changing on a daily basis with a wide variety of increase patterns. Among all products, DDR4 has seen the most dramatic price surge.

Under normal industry evolution, DDR4 should have gradually ceded ground to newer standards. DDR5 has been on the market for over four years, yet the past year has seen highly abnormal price behavior. DDR4 prices have risen far beyond expectations, significantly outperforming DDR5.

According to data cited by overseas media, DDR5 spot prices have risen more than threefold since September last year. However, as of January 26, 2026, DDR4 prices have surged by approximately 1,845% year-on-year, nearly 19 times higher than a year ago.

DDR4 began its upward trend early last year, with a clear rise in the first half and accelerated gains in the second half. In contrast, despite its more advanced technology and higher positioning, DDR5 recorded an annual increase of only about 465% (roughly fivefold), significantly lagging behind DDR4.

Driven by the rapid expansion of artificial intelligence applications, global demand for memory chips has reached historic highs. At the same time, due to substantial differences in profit margins across memory categories, manufacturers have increasingly prioritized the production of high-margin products such as HBM, leading to capacity compression for conventional DDR4 and DDR5. Reduced supply has created ample room for market speculation, resulting in stockpiling and price arbitrage in the DDR4 spot market.

When will the memory market return to normal? Based on current signals from multiple sources, price increases are likely to continue. This cycle is not merely a short-term imbalance between supply and demand or an inventory adjustment, but rather a structural transformation driven by artificial intelligence, fundamentally reshaping the role of DRAM and NAND.

At the core of this transformation is NVIDIA’s next-generation AI accelerator, Vera Rubin. Its data processing unit (DPU), BlueField-4, is equipped with 128 GB of low-power DRAM (LPDDR5X) supplied by Samsung Electronics and SK hynix. The extensive adoption of LPDDR — traditionally used in mobile devices — is driven by the need for context memory. To seamlessly retain user intent and historical conversation context, AI systems must store massive volumes of data in real time.

In addition, the rapid growth of the robotics market, including humanoid robots, represents another major tailwind for the memory industry. Due to strict power efficiency requirements, battery-powered robots inevitably rely on mobile DRAM and LPDDR solutions.

TrendForce forecasts that in the first quarter of this year, DRAM and NAND flash prices will rise by more than 60% and 38% quarter-on-quarter, respectively. The entire system stack — spanning both hardware and software — is undergoing structural reorganization around memory. The global memory semiconductor market is projected to reach USD 551.6 billion by 2026, with momentum continuing into 2027, when the market size is expected to expand to USD 842.7 billion, representing 53% year-on-year growth.

Morgan Stanley notes that supply-demand gaps in traditional memory products continue to widen, with the industry entering a new super-cycle in 2026. Products such as DDR4, DDR3, NOR Flash, and SLC/MLC NAND are expected to face increasing supply constraints. As of January 2026, leading enterprises remain aggressive in DDR4 procurement, and due to limited supply, first-quarter DDR4 prices could rise by as much as 50%, with momentum extending into the second quarter. Meanwhile, capacity shifts toward DDR4 have resulted in severe shortages of high-density DDR3, boosting performance among related suppliers. In the flash memory segment, NOR Flash prices are projected to rise 20–30% in Q1, with upward trends likely to persist into the second half of 2026.

NVIDIA CEO Jensen Huang recently stated that AI memory will soon become the largest storage market globally. NVIDIA’s powerful processors rely heavily on high-bandwidth memory supplied by Micron, SK hynix, and Samsung for training and operating large AI models. As AI models grow increasingly complex, the volume of data consumed and generated continues to surge, fueling unprecedented demand for memory chips and benefiting even traditional storage system manufacturers.

Currently, memory and flash shortages — along with rising prices — have spread beyond traditional enterprise users. Computers and smartphones have borne the initial impact, while smart vehicles are emerging as the next pressure point. High-end vehicles are increasingly adopting memory specifications comparable to consumer electronics. If price increases persist, memory costs per vehicle will continue to rise, squeezing automakers’ profit margins and intensifying operational pressure across the industry.

That said, the memory market has long exhibited pronounced cyclical characteristics, oscillating between supply shortages and oversupply every few years. This inherent volatility has discouraged manufacturers from aggressively expanding capacity, particularly given the high costs and long lead times associated with building new fabrication facilities.