Storage Chip Prices Surge by 50%: Entering a Super Cycle

“NAND flash memory prices have surged by 50% this November!” SanDisk’s price increase notice has sent shockwaves through the tech community. Since September, storage chip prices have skyrocketed, from DDR4 memory to SSD flash storage, with industry insiders declaring that the dormant storage market has finally entered a “super cycle.”

This price increase is no coincidence. As Samsung and SK Group shift their production lines towards AI servers, and 245TB ultra-large capacity SSDs replace mechanical hard drives, a storage “earthquake” triggered by a computing power revolution has been quietly brewing.

1. Korean Giants Dominate 70% of the Market: Is Price Increase Inevitable?

When discussing storage price increases, one cannot overlook the two “top players”: Samsung and SK Group. In the DRAM market for Q2 2025, SK Hynix holds a solid 38.7% market share, followed closely by Samsung at 32.7%, together capturing over 70% of the market; the situation is even more pronounced in the NAND sector, with Samsung at 32.9% and SK Group at 21.1%, controlling nearly half of the market.

“It’s like two vendors in a market occupying seven out of ten stalls; if they want to adjust prices, the entire market has to follow suit,” an industry analyst likened. Since September, Samsung has taken the lead in reducing DDR4 production capacity, with SK Hynix quickly following suit, redirecting their production lines towards server storage—after all, the profit margin for the latter is 30% higher than for consumer-grade products; it would be foolish to ignore the money.

Even more aggressive is SK Group’s HBM (High Bandwidth Memory) capacity, which has already been fully booked by giants like NVIDIA and AMD for the next two years. “It’s not that we want to raise prices; it’s that our capacity cannot be adjusted quickly enough,” a Samsung representative responded at the Import Expo, revealing a hint of “Versailles” attitude: DDR4 and DDR5 share production lines, and with the explosive demand from AI, who would still want to produce low-margin consumer-grade products?

2. AI as the “Behind-the-Scenes Driver”: HBM is Insufficient, ESSD to the Rescue

If the storage price increase is a play, AI is the director. Many people think AI only requires GPUs and HBMs, forgetting that “after computing, data must be stored”—HBM acts like a computer’s “temporary notepad,” responsible for data exchange with the GPU; ESSD (Enterprise Solid State Drive) is the “permanent filing cabinet,” where all training data and inference results must be stored.

“In the past, training an AI model required dozens of TB of data; now, large models often require PB-level data, equivalent to 1000 TB,” a representative from SK Group pointed out. Data from TrendForce is even more illustrative: the demand for AI inference has tripled the need for high-speed storage in data centers, while HDDs (mechanical hard drives) cannot keep up due to their slow speed and high power consumption, leaving a massive supply gap that can only be filled by SSDs.

It’s like giving Bolt a pair of flip-flops—no matter how fast the GPU runs, if the data can’t be stored, it’s useless. Thus, Samsung and SK Group are collectively “betting” on ESSDs, with ultra-large capacity products of 122TB and 245TB being rapidly developed. Although SSDs are more expensive than HDDs, TrendForce has calculated that large data centers using SSDs can offset the price difference within three years through savings on electricity, cooling, and rack space.

3. Can Domestic Manufacturers Take Over? Changxin and Yangtze River Need to “Accelerate”

In the face of the capacity migration of Korean manufacturers, domestic consumers are most concerned: can our own storage chips step up? The answer is “yes, but we need to wait.”

There are two “seed players” in the domestic market: Yangtze Memory Technologies focuses on NAND, with an estimated global market share of about 9% in Q2 2025; Changxin Memory is deeply involved in DRAM, with a projected share of 4.1% in Q1 2025. Changxin plans to increase its monthly production capacity from 200,000 wafers to 300,000, and Yangtze Memory is also expanding production, but a Samsung representative poured cold water on the optimism: “Building production lines requires purchasing equipment, and debugging takes at least six months; it won’t be quick to fill the gap.”

It’s also important to distinguish between “real players” and “middlemen.” The popular stocks in the A-share market, such as Jiangbolong and Baiwei Storage, are actually “packaging factories”—they buy chips from others and process them into modules, just like a bakery buys flour to make cakes; if the flour (chips) is insufficient, naturally, not many cakes can be made. The real breakthrough will come from original manufacturers like Changxin and Yangtze River that can produce their own “flour.”

However, there is good news: niche storage types like Nor Flash and SLC NAND have not been affected. A Samsung representative explained: “Their processes are mature, with small capacities, and they do not share production lines with the storage used for AI, so capacity will not be squeezed.” The storage used in home routers and smartwatches falls into this category, so there is no need to worry about price increases for now.

4. How Long Will the Price Increase Last? At Least Until Mid-2026

Will this wave of price increases be a “flash in the pan”? A Samsung representative provided a clear answer: “At least until mid-2026.” There are three solid reasons behind this:

First, AI demand is a “necessity”; large models, autonomous driving, and smart cities are all competing for storage, and the demand side will not decrease in the short term; second, capacity adjustments have a “time lag”; it takes at least one year from building production lines to mass production, and the supply-demand gap will not be filled immediately; finally, the inventory cycle is “assisting”; storage manufacturers have been too aggressive in destocking over the past two years, and now inventory is at a low point, with replenishment demand further driving up prices.

For ordinary consumers, the impact is actually limited. The consumer-grade storage used in phones and computers may see slight increases in the short term, but it won’t be as dramatic as for enterprise-grade products; the real impact will be on tech companies, especially cloud service providers and AI enterprises that rely on storage, as cost pressures will be passed on to service prices.

5. Behind the Price Increase: The “Storage Revolution” of the Computing Power Era

This wave of storage price increases is essentially a forced response of “storage capacity” to the “computing power revolution.” Just as the industrial revolution gave rise to railways, the information revolution birthed the internet, the AI revolution inevitably requires stronger storage support.

We used to say, “Data is oil”; now we realize that storage chips are the “oil depot”—without sufficient oil depots, no matter how much oil there is, it cannot be utilized. The capacity migration of Korean manufacturers is not merely a “price speculation” but a strategic choice betting on the future; the pursuit of domestic manufacturers is not just to fill the gap but also to master the “data storage rights” in the computing power era.

From DDR4 to DDR5, from HDD to SSD, every upgrade of storage chips is paving the way for future technologies like AI and the metaverse. This wave of price increases may bring short-term pain, but in the long run, it signifies that a new era has arrived—a new storage era characterized by “capacity to store and speed to read.”

As for when the price increases will stop? Perhaps we will have to wait for the new production lines of Changxin and Yangtze River to ramp up, for the demand for AI storage to stabilize, and for our own “storage chips” to truly stand up. Until then, this “super cycle” will continue.