A strategic shift by major chipmakers away from AI-focused HBM is set to deepen a supply shortage in traditional memory, with record prices squeezing hardware markets from data centers to consumer electronics.
Back
A strategic shift by major chipmakers away from AI-focused HBM is set to deepen a supply shortage in traditional memory, with record prices squeezing hardware markets from data centers to consumer electronics.

Memory chip giants are reallocating capital from high-bandwidth memory (HBM) to more profitable traditional DRAM, a move that is causing DDR5 contract prices to surge over 20 percent in the second quarter of 2026 and tightening global supply, according to a Bank of America report.
The report, published April 24, notes that the investment logic for semiconductor producers is shifting from an "AI premium" to a "traditional supply shortage," as manufacturers decline long-term contracts to capitalize on soaring spot prices. This pivot is reshaping the production priorities at key suppliers like Samsung, SK Hynix, and Micron.
The price pressure is significant, with 16Gb DDR5 contract prices having already climbed around 200 percent in the fourth quarter of 2025 and 80 percent in the first quarter of 2026. Bank of America now forecasts NAND contract prices will approach $30 in the current quarter, while DDR5 spot prices hold near a 25-year high.
This pivot threatens to constrain the supply of HBM for AI accelerators from companies like Nvidia while driving up costs for PC and smartphone manufacturers. The situation is amplified by a potential 18-day labor strike at Samsung starting May 21, which could further disrupt an already fragile supply chain.
The capital reallocation is driven by simple economics: the profitability of conventional DRAM, such as LPDDR5, has surpassed that of HBM, according to Bank of America's analysis. The trend is compounded by a slight slowdown in orders for next-generation HBM4 and the growing use of NAND-based KV cache as a partial substitute for HBM in some AI inference workloads. As a result, capital originally budgeted for HBM expansion is now being diverted back to traditional DRAM and even NAND fabrication. This makes near-term revenue growth more dependent on conventional memory sales rather than the next wave of AI-specific chips.
The market is grappling with supply constraints that extend beyond production planning. A looming labor dispute at Samsung, where tens of thousands of workers are demanding a greater share of the company's $38 billion first-quarter operating profit, adds a layer of uncertainty. Any production slowdown at the world's largest memory maker would have immediate ripple effects, tightening inventories for everything from enterprise servers to consumer devices.
Longer-term supply relief also appears limited. Bank of America noted that most new wafer fabs planned for 2028 are now expected to add only 5,000 to 6,000 wafers per month in capacity, far below initial expectations. This conservative expansion, based on the execution risks of advanced process nodes, supports a bullish outlook for memory pricing into 2027. Omdia projects that the computing and data storage segment will lead semiconductor revenue growth in 2026, rising 90 percent year-on-year to exceed $700 billion, fueled almost entirely by elevated memory pricing rather than unit growth.
This article is for informational purposes only and does not constitute investment advice.