TL;DR
Memory prices for AI hardware are slowing their rise, but this is due to demand destruction from economic struggles, not supply easing. The industry remains tight, and prices are expected to stay high for years.
Memory prices for AI hardware are not recovering but are instead plateauing at high levels, driven primarily by demand exhaustion rather than supply recovery, according to recent industry data. This trend signals a cooling of the AI market, with significant implications for hardware costs and deployment strategies.
Recent data from TrendForce’s July survey shows that the growth rate of memory prices, including DRAM and NAND, has slowed significantly compared to previous quarters. Prices are rising at 13–18% quarter-over-quarter for Q3, down from approximately 60% increases in Q2. This moderation is attributed to consumer electronics makers reaching their affordability limits after months of sharp price increases, resulting in demand reduction rather than supply easing.
Despite the slowdown, supply remains limited, and prices are still at high levels. Industry experts emphasize that this is not indicative of a market recovery but rather a stabilization at elevated prices. The underlying driver is the industry’s shift toward high-bandwidth memory (HBM) for AI accelerators, which has led to a significant reallocation of wafer capacity. Major manufacturers like Samsung, SK Hynix, and Micron have prioritized HBM, which is fully booked for 2026, with SK Hynix and Micron having secured their entire capacity for the year by late 2025.
This capacity shift has contributed to notable price increases in PC DRAM contracts, which rose over 105% quarter-over-quarter in early 2026, and DDR5 chip prices have increased substantially within a single quarter. NAND prices increased by 246% in 2025, with ongoing weekly fluctuations. Industry sources indicate that monthly price increases of 10–20% may continue through the end of the year, influenced by capacity allocation decisions and ongoing shortages.
Analysts describe this situation as a ‘permanent reallocation’ rather than a cyclical fluctuation, with expectations of market relief around late 2027, when new manufacturing capacity from Micron in Idaho is projected to begin production. The current market environment is characterized by high profits driven by shortages caused by capacity prioritization, with skepticism about claims of ongoing shortages being solely due to supply constraints.
Implications of Demand-Driven Price Stabilization
This trend suggests that the current market slowdown is primarily due to demand reduction rather than increased supply, which could influence future pricing, hardware affordability, and deployment timelines for AI and high-performance computing projects. Companies dependent on memory-intensive hardware should consider the likelihood of sustained high prices over the coming years, as capacity continues to be reallocated toward high-margin AI components. The persistent high costs may encourage the adoption of memory-efficient architectures or alternative solutions, but overall, the market remains constrained by demand reduction rather than supply improvements.
high bandwidth memory (HBM) for AI accelerators
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Recent Memory Price Trends and Industry Capacity Shifts
Over the past year, memory prices have experienced significant increases driven by supply shortages and capacity reallocation toward high-bandwidth memory for AI accelerators. Industry reports from TrendForce and IDC highlight record rises in DRAM and NAND prices, driven by a strategic shift by major manufacturers to prioritize high-margin HBM over conventional memory. SK Hynix and Micron have fully booked their 2026 capacity, with HBM supply completely allocated for the year, reflecting a structural change rather than a temporary shortage.
This reallocation has resulted in notable price increases, with PC DRAM contracts rising over 105% in a single quarter and DDR5 chip prices experiencing substantial growth within a season. NAND prices have also risen sharply, with ongoing weekly fluctuations. Industry analysts warn that these price levels are unlikely to decline soon, as capacity remains constrained and demand is weakening due to economic pressures.
While some interpret the slowdown in price increases as a market recovery, experts emphasize it is a demand exhaustion phenomenon, not an easing of supply constraints. The industry anticipates relief around 2027, when new capacity is expected to come online, but current conditions suggest a prolonged period of elevated prices.
“Major manufacturers have prioritized high-margin HBM, which has led to a structural shortage of conventional memory and sustained high prices.”
— supply chain expert
Unclear Duration of High-Price Plateau
It is not yet clear how long demand reduction will persist before supply-side adjustments or new capacity come into play. Industry forecasts suggest relief may not occur before late 2027, but the exact timeline remains uncertain due to potential shifts in demand, supply chain disruptions, or technological innovations that could alter current trends.
Monitoring Capacity Expansion and Demand Trends
The next steps involve tracking capacity expansions, particularly Micron’s new Idaho fabs expected to start production in late 2027, and observing demand patterns, especially from AI and high-performance computing sectors. Industry analysts advise companies to plan for sustained high prices and to consider strategic procurement, including locking in prices through contracts or buying minimal necessary capacity within the next two quarters.
Additionally, developments in alternative architectures that require less memory could influence future demand, potentially alleviating some pressure if adopted widely. Market watchers should stay alert to these technological and capacity shifts to better anticipate price movements.
Key Questions
Why are memory prices for AI hardware slowing their growth?
The slowdown is primarily due to demand exhaustion, as consumer electronics makers reach their affordability limits after months of sharp price increases, leading to demand reduction rather than supply easing.
Will memory prices decrease soon?
Industry experts suggest that prices are unlikely to decline before late 2027, as supply remains limited and demand continues to weaken, with relief expected only after new manufacturing capacity comes online.
What is causing the current high memory prices?
The primary cause is the reallocation of wafer capacity toward high-margin high-bandwidth memory for AI, which has led to a structural shortage of conventional memory and record price surges.
How should companies plan hardware procurement now?
Experts recommend buying minimum necessary capacity, locking in prices via contracts, and acting within the next two quarters to avoid higher costs later, as the market is expected to remain constrained for years.
Could technological innovations reduce memory demand?
Yes, architectures requiring less memory are being developed, which could impact future demand, but widespread adoption is still uncertain and unlikely to significantly alter current trends in the near term.
Source: ThorstenMeyerAI.com
