Market Update: DDR4 rose 16.9% in July while NAND’s slide paused
Memory and storage spent most of July moving in opposite directions, an unusual split in this cycle. In the final week, NAND reversed and joined DRAM on the way up.
TrendForce’s DDR4 8Gb spot benchmark rose from $36.00 on June 30 to $42.08 on July 28, an increase of 16.9%. Its 512Gb TLC NAND benchmark fell in every published observation through July 20 before rebounding about 1.7% in the final week.
The immediate trigger behind that NAND reversal is the most useful detail in the month’s data: selective enterprise-SSD restocking, not a broad recovery in demand.
The July numbers, week by week
Two benchmarks matter here.
The DRAM series tracks DDR4 8Gb chips, quoted as 1Gx8 3200MT/s, in the component spot market. It is not a finished-module price index. Retail DIMMs may use different chip grades and sourcing channels, and their prices can also reflect contract purchases, older inventory, module assembly and channel margins.
The NAND series tracks the 512Gb TLC wafer-market benchmark, equivalent to 64GB of raw flash capacity. TLC means three bits stored per memory cell and is widely used in consumer and enterprise SSDs.
| Observation | DDR4 8Gb chip | 512Gb TLC NAND |
|---|---|---|
| Opening | Jun. 30 — $36.00 | Jun. 29 — $19.86 |
| First week | Jul. 7 — $37.14 | Jul. 6 — $19.29 |
| Mid-month | Jul. 15 — $39.80 | — |
| Third week | Jul. 21 — $41.10 | Jul. 20 — $18.93 |
| Closing | Jul. 28 — $42.08 | Jul. 27 — $19.25 |
（Some data cited from here）
DDR4 rose from $36.00 to $42.08, an increase of 16.9% over the reporting window. NAND fell from $19.862 to $18.931 before recovering to $19.250. It finished approximately 3.1% below its June 29 level but was nearly flat between the first and final July observations, declining just 0.2% from July 6 to July 27.
Method: DRAM and NAND are surveyed on different days, so the paired values are close in time but not simultaneous. No separate report was published during the week of July 13; the July 15 DRAM value appeared retrospectively in the July 22 report.
TrendForce also restated two DRAM observations in later reports. June 30 was initially published as $36.00 and later cited as $36.10; July 21 was initially published as $41.10 and later cited as $41.50. Neither revision changes the direction of the market, but both affect reconstructed percentage changes. This table consistently uses each value as it first appeared for that observation.
DRAM: rising on very little trading
The striking part of the DDR4 climb is how little transaction volume it required.
Trading remained thin throughout the month. Suppliers resisted selling at prices they considered too low, buyers were reluctant to follow quotes higher, and by the final week TrendForce still described the upper end of the market as lacking a clear consensus.
Quoted prices rose anyway. In a constrained and lightly traded spot market, a benchmark can move materially even when relatively little product changes hands.
Where the strength appeared also matters. Early gains were concentrated in low-density DDR4 and DDR3, components the industry has spent years treating as obsolete. By the end of July, inquiries had broadened across branded DDR3, DDR4 and DDR5 products.
The bid was no longer confined to a few legacy components.
NAND: selective restocking moved a thin market
Consumer demand remained weak, while suppliers reduced some quotations to clear inventory. The lower offers produced little improvement in overall purchasing activity.
The final week then turned positive. TrendForce attributed the increase mainly to selective restocking by a module house preparing for enterprise SSD orders. Similar enterprise-SSD-related purchasing had supported the market one week earlier, although it was not enough to prevent another decline.
That is the month’s most useful NAND signal.
When selective replenishment from a small part of the buyer base appears sufficient to reverse the weekly benchmark, it indicates that the underlying spot market is thin. The rebound reflected specific purchasing activity rather than a broad recovery across consumer, client and enterprise demand.
Supply conditions still mattered. TrendForce also noted limited wafer availability and firmer supplier quotations. But the immediate trigger for the late-July increase was selective replenishment, not a new industry-wide supply disruption.
Anyone treating the rebound as a confirmed bottom should first look for it to continue after the current restocking cycle ends—and for participation to broaden beyond a small number of module houses.
Why contract prices tell a different story
Spot and contract prices are often quoted as though they describe the same market. They do not.
Spot prices reflect relatively immediate transactions and quotations in the open component market. They respond quickly to available inventory, supplier behavior and short-term purchasing.
Contract prices reflect negotiated bulk transactions between suppliers and larger customers. Depending on the product, they may be tracked over monthly or quarterly periods and generally adjust more slowly than spot benchmarks.
For the third quarter, conventional DRAM contract prices are projected to rise 13–18% quarter over quarter, while NAND flash contract prices are projected to rise 10–15%. Both forecasts remain positive, although the pace of increase is expected to moderate.
NAND contract prices can therefore rise over the quarter while the weekly wafer-market benchmark falls. The series cover different transaction channels, customer groups and time periods and should not be expected to turn simultaneously.
For flash buyers, the divergence suggests that the current contract-price increase may be closer to the upper portion of the cycle than its beginning, particularly if the spot rebound does not broaden.
For DRAM, the reading is different. Spot pricing was still rising at the end of July, so the pressure behind contract-price increases had not visibly peaked.
What this looks like at retail
Finished memory products reflect upstream conditions unevenly and with a lag.
A DDR5-6000 32GB kit was listed at $389 on August 3 in this price tracker. Its table also showed $826 for a 64GB DDR5-6000 kit, $1,583 for 96GB and $2,199 for a 128GB DDR5-6400 kit.
Those figures show how expensive high-capacity retail memory has become, but the tracker is not a controlled price index. It reports the lowest available offer in each capacity and speed category regardless of manufacturer. When the page is updated, the product representing the category may change.
A comparison between two captures can therefore reflect a combination of genuine price movement, stock availability and substitution between different kits. It should not be interpreted as the price history of one consistent SKU without preserving the underlying product data.
The retail figures also should not be compared mechanically with the $42.08 DDR4 chip benchmark. They cover different memory generations, component grades, purchasing channels and inventory periods. TrendForce separately tracks mainstream chips, lower-priced eTT chips and finished module spot prices, each of which can behave differently.
The useful retail conclusion is narrower: memory remains expensive, availability is uneven, and the highest-capacity DDR5 categories are experiencing particularly severe premiums and stock constraints.
For DDR4, Tom’s Hardware places common 32GB kit prices at approximately $60 to $90 in October 2025 and $150 to $180 by January 2026. DDR4 is no longer the inexpensive fallback it once was.
Major manufacturers are reducing DDR4 production. Although Taiwanese suppliers have been expanding output, TrendForce says those additions are not sufficient to offset reductions elsewhere.
Net supply is still contracting, and July’s component data shows older DDR4 products leading price increases rather than lagging them.
The part of NAND that isn’t falling
SLC NAND—the oldest major NAND type, storing one bit per memory cell—is forecast to see contract prices rise 120–170% in the second half of 2026 compared with the first half.
That forecast covers a different product and pricing channel from the 512Gb TLC wafer-market benchmark. The two can move in opposite directions without contradicting each other.
Manufacturers have shifted mature production capacity toward denser, higher-layer 3D NAND. That has reduced the availability of legacy MLC products and pushed some industrial, automotive and networking customers toward small-capacity 4Gb and 8Gb SLC components.
Those customers cannot always respond to higher prices by changing products or reducing capacity. Their equipment may be governed by qualification requirements, fixed memory specifications and long operating lifecycles.
A part number written into a certified bill of materials is not a preference. Requalifying flash in a medical imaging system, industrial controller or signalling platform is a program, not a purchase order.
SLC is also used for boot storage and write-intensive buffers in qualified networking, industrial and data-center equipment. The headline “NAND prices are falling” does not describe the replacement market for that hardware.
What it means for equipment coming out of service
DDR4 removed from retiring servers and desktops is entering a market where supply is contracting and spot prices for components of the same generation are rising.
That reverses much of the pattern seen during the previous decade, when last-generation memory generally depreciated quickly. Working modules in storage may be worth more than internal depreciation schedules assume, making it sensible to value used server and desktop memory against current conditions.
Actual resale value still depends on the module. Capacity, speed, ECC support, RDIMM or LRDIMM form factor, rank configuration, manufacturer and current platform demand all matter. Older memory should not be valued from a single commodity benchmark alone.
Enterprise SSDs and consumer drives should also not be treated as one category. Suppliers are prioritizing enterprise demand while consumer demand remains weak. A drive’s interface, capacity, endurance rating, remaining life and intended customer now matter more than the direction of a general NAND benchmark.
Functional networking and industrial equipment that depends on qualified legacy flash may also be worth more as a complete spare than its age suggests.
The embedded SLC component is rarely economical to recover by itself. The complete system, however, may be expensive to replace where qualification requirements prevent straightforward substitution.
That is a reason to identify what is inside a chassis—and which deployed systems still depend on it—before the equipment leaves the building.
What to watch in August
The main question is whether the NAND rebound holds.
If the 512Gb TLC benchmark continues rising and purchasing broadens beyond selective module-house restocking, the market will have stronger evidence that the decline has ended. A broader recovery would also strengthen suppliers’ position in fourth-quarter contract negotiations.
If prices soften again after the current replenishment cycle ends, late July will look like a temporary interruption rather than a durable floor.
For DRAM, the near-term direction is less ambiguous. TrendForce expects supply to remain constrained through 2027. New fabrication capacity is expected to begin ramping in the second half of that year, but substantial output is not expected until 2028.
NAND is forecast to loosen sooner, potentially during the second half of 2027, as new capacity enters production while consumer demand remains weak.
NAND may get its supply response in 2027. DRAM may not get a meaningful one until 2028.
Anyone planning purchases around a single “memory prices” forecast is planning around two markets that have stopped agreeing. Our mid-2026 update covers how the cycle looked before the split became this visible.