A structural shortage driven by AI infrastructure demand has pushed DRAM prices up by hundreds of percent in under a year, reshaping consumer electronics, data centre economics and the semiconductor supply chain through at least 2027.
RAM is no longer a background component. Across 2026, dynamic random-access memory (DRAM) has moved from being a stable, commoditised part in every laptop and phone to one of the most volatile inputs in modern computing, and the cause is almost entirely traceable to artificial intelligence.
The scale of the price increase
The numbers involved are not incremental. According to a July 2026 Bloomberg report, DRAM spot prices have surged nearly 700% over the past year. Contract pricing, which reflects the bulk agreements between manufacturers and large buyers, has moved almost as sharply: TrendForce data pegged conventional DRAM contract increases at 50% to 55% quarter-over-quarter early in 2026, accelerating toward an 89% figure by the second quarter. Separate TrendForce guidance for the second quarter projected conventional DRAM contract prices rising 58–63% quarter-on-quarter, with NAND Flash contract prices up 70–75%.
At the component level, Tom’s Hardware’s RAM price index recorded the spot price of a 16Gb DDR5 memory chip rising from roughly $6.84 in September 2025 to about $27.20 by December 2025, a jump of nearly 298% in a single quarter. That wholesale movement reached consumers directly; a 32GB DDR5-6000 CL30 kit that sold for under $90 in early 2025 was fetching around $529 by late 2025 into 2026, roughly a fourfold increase. Even legacy DDR4 stock was affected, with a 32GB DDR4 kit that cost $60 to $90 in October 2025 listed at $150 to $180 by January 2026.
By June 2026, analyst house Gartner had converged on a similarly severe outlook, projecting a memory cost surge on the order of 130% and warning of an extended storage crunch stretching into 2027.
There are early signs of the rate of increase slowing, though not the underlying shortage. TrendForce’s most recent survey projects conventional DRAM contract prices to rise 13% to 18% quarter-over-quarter in Q3 2026, with NAND Flash contract prices increasing 10% to 15%, a marked deceleration from the roughly 60% jumps recorded in the previous quarter. Analysts attribute this cooldown to demand fatigue among consumer electronics manufacturers rather than any improvement in supply; memory remains in short supply, but consumers are no longer willing to keep paying ever-higher prices.
Why artificial intelligence is the direct cause
The shortage traces back to a specific category of memory: high-bandwidth memory (HBM), the type used in the accelerator chips that power large-scale AI training and inference. Manufacturers have been reallocating fabrication capacity away from conventional DRAM and toward HBM because it commands substantially higher margins and is contracted years in advance by AI infrastructure buyers.
According to IDC, the three companies that control the overwhelming majority of global DRAM output, Samsung, SK Hynix and Micron, together control over 95% of global DRAM production and have systematically reallocated manufacturing capacity toward HBM chips used in AI accelerators, leaving consumer-grade DRAM and NAND flash in critically short supply. Data centres are now estimated to consume 70% of all memory chips produced worldwide, a level of concentration unprecedented in the memory industry’s history.
The clearest corporate signal of this shift came from Micron. In February 2026, the company retired Crucial, its consumer-facing memory brand, in order to concentrate on data-centre and HBM products. Micron has also indicated that its HBM output is effectively sold out for 2026, a claim corroborated by CNBC reporting that HBM capacity for 2026 is entirely sold out, with manufacturers refusing new orders. Micron’s own financial results reflect the shift in priorities: the company reported record revenue of $23.86 billion, signalling memory’s move from a commodity product to a strategic asset.
TrendForce analysis published in March 2026 confirmed the mechanism directly, noting that DRAM suppliers keep reallocating capacity toward server-related applications, with cloud service providers willing to accept higher prices and sign long-term agreements to secure a stable supply, further reinforcing suppliers’ pricing power.
How the industry arrived at this point
Part of what has made the shortage so severe is the timing of the underlying investment cycle. IEEE Spectrum senior member Thomas Coughlin has noted that an extended period of capital expenditure restraint left the industry poorly positioned to respond when AI-driven demand surged in mid-2025. That restraint followed a period in which Samsung had cut production by roughly 50%, with the entire industry making little or no investment in new production capacity through most of 2024 and into early 2025.
New capacity is now being built, but not quickly enough to close the gap. IDC’s latest analysis puts global DRAM supply growth in 2026 at just 16% year-on-year, and NAND supply growth at 17% year-on-year, both well below historical norms, while new fabrication capacity from Micron and SK Hynix is not expected to reach volume production until 2027 at the earliest.
Downstream effects and outlook
The consequences are visible well beyond enthusiast PC building. IDC has warned that as memory prices continue to climb, device manufacturers will likely have to raise prices significantly, cut specifications, or both, with the impact asymmetric across the supply chain. Counterpoint Research has reported that the same dynamic is reshaping the smartphone market, with memory buyers for AI data centres effectively crowding out phone manufacturers as they compete for the same constrained pool of DRAM.
Estimates for how long this will last vary, but none are short-term. IDC and industry analysts covering Samsung, SK Hynix and Micron describe a shortage that could persist well into 2027; SK Hynix’s own chief executive warned in July 2026 that the constraint would probably persist well beyond 2030. TrendForce, separately, has projected no meaningful capacity expansion until late 2027 or 2028 for related NAND and SSD markets.
What distinguishes this cycle from previous memory shortages, which have historically been cyclical and self-correcting within a year or two, is the structural nature of the demand driving it. AI infrastructure buildout is not a temporary spike; it represents a durable shift in how fabrication capacity is allocated industry-wide, and manufacturers currently have every financial incentive to keep it that way.
Sources
Accio Business Intelligence, “DRAM Price Trend 2026: AI Demand Drives Surge,” accio.com Tom’s Hardware, “Memory price surge begins to cool as consumers hit affordability limit,” tomshardware.com Shattered.io, “RAM Prices Up 89%: AI Memory Crunch Hits Gaming [2026],” shattered.io CNBC, “AI memory is sold out, causing an unprecedented surge in prices,” January 10, 2026, cnbc.com TrendForce, “AI Server Demand to Drive Memory Contract Price Increases in 2Q26,” March 31, 2026, trendforce.com IDC, “Global Memory Shortage Crisis: Market Analysis and the Potential Impact on the Smartphone and PC Markets in 2026,” February 10, 2026, idc.com Tech Insider, “2026 Memory Chip Shortage: SK Hynix Warns It May Last Past 2030,” tech-insider.org MEXC News, “AI-driven memory shortages set to raise phone prices,” December 16, 2025, mexc.com The Register, “DRAM prices expected to double in Q1 as AI ambitions push memory fabs to their limit,” theregister.com