Recently, friends planning to upgrade their computers were likely shocked by the prices of memory modules: Acer Predator DDR5 kits were around 6200 yuan at the end of September, skyrocketing to 8000 yuan by late November; even more astonishing is the Samsung 32GB DDR5 server memory, which saw a contract price increase from $149 in September to $239 in November, a staggering rise of 60%.
Not only computer memory, but the cost of mobile storage is also soaring. Industry insiders have revealed that the prices of new devices could generally rise by 3%-8% in 2026, with the price difference between the 512GB and 256GB versions widening from 300-400 yuan to 500-600 yuan. Many are puzzled: hasn’t China already achieved mass production of storage chips? Logically, an increase in production capacity should lead to lower prices, so why is it that prices seem to change daily? Today, we will clarify the reasons behind this price surge.
Core Reason One: AI Drives “Insatiable” Demand, Computational Power Fuels Storage Needs
The key driver behind this price increase is the explosive growth of artificial intelligence. Some may ask: Isn’t AI reliant on graphics cards? What does it have to do with storage chips? The answer is: the more powerful the “brain” of AI, the larger the “memory bank” it requires.
The data is compelling: a single AI server uses 8 times more DRAM (memory chips) than traditional servers, and the usage of NAND (flash memory chips) is even 3 times higher. For instance, OpenAI’s “Stargate” project requires 900,000 DRAM wafers per month, accounting for nearly 40% of global total capacity. The demand for high-bandwidth memory (HBM) at the core of AI is even more staggering, increasing from 300,000 units in 2024 to 1.2 million units in 2025, a fourfold increase in just one year.
In addition to AI servers, demand in consumer electronics and the automotive sector is also rebounding. In the third quarter of 2025, global smartphone shipments are expected to grow by 12% year-on-year, with new devices generally upgrading storage capacity from 256GB to 512GB; L3 autonomous vehicles are expected to use over 10GB of DRAM and more than 200GB of NAND, with multiple demands compounding to continuously widen the supply-demand gap for storage chips.
Core Reason Two: International Giants Actively “Limit Production and Shift Production”, Intentionally Shrinking Supply
If demand is the “pulling force”, then the capacity adjustments by international giants are the more critical “pushing force”. The global storage chip market has long been dominated by a few companies such as Samsung, SK Hynix, and Micron, which together control 93% of DRAM capacity and over 85% of NAND capacity. In pursuit of higher profits, they have collectively adopted a strategy of “production cuts + production shifts”.
In terms of production cuts, Samsung’s NAND wafer production is expected to decrease by about 7% in 2025, while SK Hynix’s NAND production will drop by 10%, and Kioxia has also reduced its capacity by 2.3%. Even more drastic is the shift in production: these giants are moving their capacity from low-profit traditional products like DDR4 to high-value products like HBM and DDR5. Samsung has even announced that it will stop accepting DDR4 orders by June 2025 and complete its final shipments by December, while SK Hynix plans to completely cease DDR4 production by April 2026.
This structural adjustment has directly led to market imbalance: there is insufficient high-end storage capacity needed for AI, while ordinary consumer-grade storage is sharply reduced due to production cuts. To make matters worse, suppliers seeing the trend of rising prices have begun to “hold back sales”; companies like GigaDevice have even paused shipments to reassess pricing, further exacerbating the “hard-to-get” situation.
Core Reason Three: Low Inventory Meets Panic Buying, Amplifying Price Effects
This price surge is also fueled by the “inventory cycle”. From 2022 to 2023, the global storage industry underwent a prolonged period of destocking, with inventory days dropping from 120 days to below 60 days, reaching historical lows. The industry was originally in a normal phase of inventory recovery but suddenly faced an explosion in AI demand and production cuts from giants, instantly breaking the supply-demand balance.
Starting in the third quarter of 2025, foundries received a large number of urgent orders, and some customers had to turn to the spot market to secure supplies, creating a cycle of “low inventory + urgent orders”. Downstream manufacturers, fearing tighter future supplies, began panic buying and stockpiling, leading to large-scale purchasing phenomena in Huaqiangbei, while the reduced circulation of spot goods further pushed prices up, falling into a vicious cycle of “price increase – stockpiling – further price increase”.
A senior executive from a well-established storage company candidly stated: “The shortage of DDR5 and DDR4 is unlikely to change until 2027.” This expectation of long-term shortages has led more people to join the stockpiling ranks, continuously amplifying the price effects.
Key Question: Domestic Storage Chips Have Achieved Mass Production, So Why Can’t Prices Be Controlled?
Many are concerned: haven’t Yangtze Memory’s 3D NAND and Changxin Memory’s DRAM already achieved mass production? Why haven’t they managed to stabilize prices? The answer is “significant progress, but still gaps remain”.
From the achievements, domestic storage is indeed commendable: Yangtze Memory’s 232-layer 3D NAND technology is globally leading, with a market share expected to reach 9% by 2025; Changxin Memory’s DDR5 yield rate has reached 80%, with costs 15%-20% lower than those of Korean manufacturers, and a projected global market share of 5% by 2025. In the domestic market, brands like Huawei and Xiaomi have begun to adopt domestic storage in bulk, with some server manufacturers increasing their procurement of domestic chips by 80% year-on-year.
However, the shortcomings are also evident: first, the production capacity share remains low, with domestic storage accounting for only about 10% of the global market, making it difficult to shake the dominance of international giants; second, high-end products lag behind; in the core AI HBM sector, Samsung and SK Hynix have already mass-produced HBM3E and even HBM4, while Changxin Memory will not achieve mass production of HBM3 until 2026, with a technological gap of about 2-3 years. Faced with a global supply-demand gap, domestic capacity can only alleviate some pressure but cannot fundamentally change the landscape.
Impact on Ordinary Consumers: Should You Buy Digital Products Early?
From a consumer’s perspective, the most pressing concern is “how long will this price increase last?” Institutions predict that DRAM prices are expected to rise by another 13%-18% in the fourth quarter of 2025, with the increase potentially narrowing to 5%-10% in the first half of 2026, but the overall high price trend is expected to continue until 2026 or even longer.
The advice for purchasing digital products is clear: if you have a need to build a system or change your phone soon, it is advisable to act quickly. Most flagship models currently on sale are still priced based on costs from the end of 2024, while new devices in 2026 will fully reflect the pressure of rising storage prices. The price reductions by brands like Redmi are actually a “window period” behavior to clear inventory. In simple terms, the logic of “buy early to save money, buy late to spend more” will hold true in the short term.
In the long run, as companies like Yangtze Memory and Changxin Memory continue to expand production, the market share of domestic storage will gradually increase, and the price fluctuations caused by geopolitical factors and technological monopolies will slowly diminish. However, for now, this surge in storage chip prices triggered by AI is far from over.
Have you recently encountered price increases related to storage? Feel free to share your experiences in the comments!