An employee of Chinese NAND maker YMTC has reportedly said the global memory shortage will last three more years. The remark comes from "The Survivor", published by The Wire China on October 4, 2026. The anonymous employee also reportedly mentioned price increases and production allocation that prioritizes higher-margin products.
The comment projects a fairly long period of tight supply, against expectations that increased output from Chinese manufacturers will ease the shortage. However, outlooks for NAND, the memory used in SSDs, differ even between research firms and major manufacturers. What matters is not only how much production grows, but also whether the added supply goes to AI and data center customers or to PCs and consumer products. That allocation will also influence where prices go.
What does "three more years of shortage" actually refer to?
The "three years" figure is not an official supply-demand forecast published by YMTC. It is based on anonymous reporting by The Wire China, and in the article summaries we could verify, it is not limited to NAND but refers to "memory" in general. We could not directly check the paywalled portion of the original article, so the employee's position, the basis for the forecast, and any breakdown by product are unknown.
Three years from autumn 2026 would point to around 2029. But YMTC has not officially said the shortage will end in 2029. The broad outlook for the memory market as a whole needs to be considered separately from when specific NAND products can be bought in the quantities needed.
NAND is a type of memory that retains data when power is off and is used in SSDs and similar devices. DRAM, by contrast, mainly holds data temporarily while it is being processed. High-bandwidth memory (HBM), used in AI accelerators, is also a type of DRAM.
AI demand is pushing up both NAND and DRAM, but the manufacturing constraints are not the same.
In its July 30 outlook, TrendForce explained that shifting production capacity toward HBM is squeezing the supply of conventional DRAM. Because HBM requires more wafer input than standard DRAM, the same number of wafers does not yield a proportional increase in usable memory capacity.
TrendForce expects DRAM supply to remain tight in 2027, while projecting that NAND supply and demand could ease in the second half of 2027. The single phrase "memory shortage" covers several markets in different situations.
Even within the industry, the 2027 NAND outlook is split
In its July 21 NAND forecast, TrendForce estimated that the 2026 NAND market will see supply fall 4–5% short of demand, but predicted supply will exceed demand in the second half of 2027, easing the squeeze.
The assumption is that the shift to high-layer-count NAND, which stacks more layers, and the start-up of new equipment will increase the memory capacity produced from the same wafers, while demand for consumer electronics stays weak.
Micron, by contrast, said in its September 30 earnings prepared remarks that supply constraints will persist in both 2027 and 2028, even if industry-wide NAND bit shipments grow in the mid-20% range in each year. The market outlook on page 5 of the document discusses NAND and DRAM separately.
In other words, for the 2027 NAND market, TrendForce sees tightness easing in the second half, while Micron sees supply constraints continuing.
| Source and date | Scope and period | Supply-demand outlook | Assumptions on capacity additions and demand |
|---|---|---|---|
| TrendForce, July 21 and 30, 2026 | Global NAND market, 2027 | Shortage eases in the second half | Supply capacity grows through the shift to high-layer products and new equipment, while consumer demand is assumed to be weak |
| Micron, September 30, 2026 | NAND industry, 2027 and 2028 | Supply constraints continue in both years | Even with industry-wide bit shipments growing in the mid-20% range each year, supply is expected to fall short of demand |
The table compares the two outlooks for 2027, the year their forecast periods overlap for the same global NAND market. However, they were published about two months apart, and their assumptions about demand and production capacity are not the same.
The difference in views therefore cannot be explained merely by confusion between NAND and DRAM. At the same time, these two forecasts alone do not allow us to judge which is correct at this point.
TrendForce's easing forecast already factors in increased output from Chinese manufacturers. It projects that as Chinese makers bring new equipment online, their share of global NAND production capacity will rise to nearly 19% in 2027.
This is a forecast for Chinese manufacturers as a whole, not YMTC's share alone. When comparing it with the anonymous employee's "three more years" remark, it is worth considering not only that the comment may cover memory broadly, but also that the assumptions may differ about how far demand will keep growing even after output increases.
Even with higher output, supply prioritized for AI SSDs won't reach PCs
Demand for enterprise SSDs is growing not only in the training of AI models but also in the operation of actual inference services.
TrendForce's September 30 research forecasts that bit demand for enterprise SSDs will grow more than 80% year on year in 2026. Data volumes handled in real-time search and caching are increasing, and high-capacity QLC SSDs are also spreading in vector databases.
QLC is a NAND method that stores 4 bits per cell and is used to achieve high-capacity storage at relatively high density.
Manufacturers are expanding QLC production capacity and the supply of high-capacity SSDs, but according to TrendForce, much of the newly added supply is already secured through contracts with large customers, leaving limited volume that new customers can freely buy on the spot market and elsewhere.
In other words, a manufacturer announcing increased production does not immediately make it easier for ordinary buyers to obtain SSDs.
YMTC also offers enterprise products. The PE501, listed on its official site, is an enterprise SSD that uses QLC and supports PCIe 5.0, with capacities ranging from 15.36TB up to 122.88TB.
This confirms that YMTC offers high-capacity SSDs for data centers. However, the product page does not reveal actual shipment volumes or how production capacity is allocated. Nor does it corroborate the scale of any production allocation changes the anonymous employee reportedly described.
Meanwhile, the PC market is adjusting in the opposite direction.
According to the same September 30 TrendForce research, PC makers are working through finished goods and channel inventory built up since the start of the year, and are cutting the SSD capacity in their main models to hold down component costs. As a result, NAND purchase volumes for PCs and capacity per unit are falling, making NAND makers more willing to concede in price negotiations.
Still, TrendForce forecasts that overall NAND contract prices will rise 15–20% quarter on quarter in October–December. Enterprise SSDs are the only category where the size of the price increase is expected to widen from the previous quarter.
The price increase here refers to contract prices set between NAND makers and their customers. It does not mean that SSD prices at electronics retailers or online shops will uniformly rise 15–20%.
Demand is extremely strong in AI and data center applications, while PC buyers are moving to hold down purchase volumes. Even within the same NAND market, supply and demand conditions differ greatly by use.
New fabs starting up doesn't necessarily mean supply rises quickly
According to Micron's September 30 materials, its new NAND fab under construction in Singapore is scheduled to begin production in the second half of 2028.
The company explains that for new DRAM and NAND fabs, ramping up takes time even after production starts, and it takes several quarters before they contribute meaningful volume to the market. The time a fab begins making its first products is not the same as the time it can supply large volumes to the market.
That said, this is only the plan for Micron's new fab. It does not mean global NAND supply expansion will not begin until 2028.
The supply growth TrendForce expects in 2027 includes manufacturing technology upgrades at existing fabs and the start-up of new equipment by Chinese manufacturers. NAND makers can raise the memory capacity produced from the same number of wafers by increasing layer counts or the number of bits stored per cell. For this reason, wafer input alone does not indicate how fast actual supplied capacity will grow.
The path to better supply and demand is also not just a matter of raising production capacity.
If NAND prices stay high, replacement purchases of PCs and smartphones could slow, and manufacturers could limit the storage capacity per device. If supply grows while demand weakens, that alone would ease the tightness.
In the demand breakdown TrendForce presented in July, servers account for more than 40% of NAND bit demand, while smartphones and laptops together approach 40%. A decline in consumer-market demand is large enough to influence overall NAND supply and demand.
TrendForce also notes that if AI agents, which autonomously handle multiple tasks, spread faster than expected, demand for high-speed SSDs could rise further and absorb the newly added supply.
The "easing in the second half of 2027" that the company forecasts is therefore not a settled future. It depends heavily on assumptions about how far AI demand will grow and how far consumer demand for PCs, smartphones, and similar products will weaken.
It would not be appropriate to read the YMTC employee's "three more years" remark as meaning all SSDs will remain in short supply until then.
To gauge what comes next, it is more useful to track contract prices for enterprise SSDs and PC NAND separately, how much product new fabs can actually ship, and which markets the output added by Chinese manufacturers is directed to. These are likely to show real changes in the purchasing environment more clearly.
Only when the added NAND first fills contracts with large customers and then also flows in sufficient volume to ordinary PC makers and the SSD market will the "supply increase" be felt as an easier time securing the capacity people need.