Standard PC DRAM and NAND flash prices hit record highs in July as AI server production absorbs memory capacity, with DDR4 up 109 percent and NAND up 218 percent year to date.
Back
DRAM prices double, NAND triples as AI servers tighten supply
Standard DRAM and NAND flash prices reached record highs in July as memory makers shift production toward AI server components, pushing DDR4 PC DRAM up 109 percent and NAND up 218 percent since January.
TrendForce, the market research firm that operates the DRAMeXchange pricing service, projects DRAM supply tightness will persist into 2027 as AI server and high-bandwidth memory (HBM) demand continue to absorb production capacity.
The average fixed transaction price for DDR4 8-gigabit standard PC DRAM rose 14.3 percent month over month to $24 in July, the highest since record-keeping began in June 2016. Prices climbed from $11.50 in January through $13 in March, $16 in April, $20 in May, and $21 in June. For NAND, the 128-gigabit MLC average price reached $30.05, up 4.3 percent from June and the first time above $30.
The pricing surge benefits Samsung Electronics and SK Hynix, which dominate the DRAM and NAND markets, while squeezing PC and consumer electronics makers facing higher component costs. TrendForce expects DRAM shortages to extend through 2027, while NAND supply could ease in the second half of 2027 as new fabrication capacity comes online.
AI servers absorb capacity, squeezing legacy products
Memory makers are prioritizing HBM, server DRAM, and high-layer 3D NAND — products with higher margins tied to AI infrastructure buildouts. That reallocation has reduced output of standard PC DRAM and legacy-process NAND used in memory cards, USB drives, and embedded systems.
The supply squeeze is most acute in SLC NAND, used in automotive electronics, network equipment, and industrial controls. July prices for SLC NAND rose 35 to 51 percent month over month, far outpacing the 4 to 8 percent gains for MLC products. Inventory shortages in SLC have persisted even as prolonged price increases have raised procurement costs for downstream buyers.
TrendForce noted that notebook demand is weakening as finished goods prices rise, but suppliers still hold the upper hand in price negotiations. PC DRAM supply is expected to tighten further as server products continue to absorb production capacity.
The pricing dynamics reflect a structural shift in the memory industry. AI data center buildouts by hyperscalers including Microsoft, Amazon, and Google have created unprecedented demand for HBM and server-grade memory, which command significantly higher prices than commodity DRAM. Memory makers have responded by reallocating wafer capacity toward these premium products, leaving less output for the PC and consumer electronics segments.
Diverging outlook for DRAM and NAND
TrendForce's forecast splits the two memory categories. DRAM shortages are expected to continue through 2027, driven by sustained AI server demand and HBM adoption. NAND supply conditions, by contrast, could gradually ease from the second half of 2027 as new production capacity comes online.
For Samsung Electronics and SK Hynix, the extended pricing power translates directly into revenue and margin expansion. Both companies have been the primary beneficiaries of the AI memory cycle, with HBM commanding premium pricing. The sustained DRAM tightness suggests their pricing power remains intact through 2027.
For downstream PC and consumer electronics manufacturers, the cost pressure is mounting. Component costs have risen sharply over the past seven months, and finished goods prices are already climbing. The question is whether consumer demand can absorb further price increases or whether the memory cycle eventually cools as NAND capacity expands.
Investors tracking the memory cycle should note the asymmetry in the outlook. DRAM pricing power appears durable through 2027, supporting continued margin expansion for Samsung and SK Hynix. NAND, however, faces a potential supply normalization in the second half of 2027, which could compress pricing for products tied to that segment. The divergence suggests memory stocks may increasingly trade on product mix rather than the broad cycle.
This article is for informational purposes only and does not constitute investment advice.
