The construction boom in artificial intelligence (AI) data centers continues to push up demand for high-bandwidth memory (HBM). The world's three major memory manufacturers—Samsung Electronics, SK Hynix (SKHY-US), and Micron (MU-US)—are accelerating the shift of production capacity toward HBM manufacturing, tightening the supply of traditional DRAM used in consumer electronics such as PCs and smartphones, and driving a sharp increase in PC memory prices. Market observers point out that this price surge is not due to supply chain disruptions or wafer fab shortages, but rather a strategic reallocation of resources by memory makers prioritizing profitability in the AI market.
According to market data, a mainstream PC memory module that cost about $75 one year ago now sells for as high as $460 for the same specifications—an increase of over fivefold. AI customers are willing to pay premium prices, making HBM the top production priority for chipmakers, while traditional DRAM for PCs, laptops, smartphones, and automotive electronics is being relegated to lower priority.
Currently, Samsung Electronics, SK Hynix, and Micron control the vast majority of the global DRAM market. Compared to traditional DRAM, HBM not only commands higher prices but also requires more wafer capacity. Each additional HBM wafer produced means a further reduction in available supply for consumer DRAM.
Semiconductor analyst Shanaka Fernando recently noted on social platform X that current memory market tightness does not require coordinated production cuts by manufacturers—the natural market mechanism is sufficient. Because HBM yields far higher profits than standard DRAM, companies naturally allocate more capacity to AI-related products rather than lower-priced consumer memory.
Latest data from market research firm TrendForce also reflects ongoing supply-demand imbalances. In Q1 2026, contract prices for traditional DRAM rose 93% to 98% quarter-on-quarter, followed by a further 58% to 63% increase in Q2. Over the same period, NAND Flash contract prices also rose 70% to 75%, primarily due to suppliers prioritizing capacity allocation to AI-related products, thereby constraining supply to the general market.
Beyond the consumer market, large tech enterprises are also feeling memory supply pressure. Tesla (TSLA-US) CEO Elon Musk earlier this year stated the company is facing a so-called 'chip wall,' as rapid expansion of AI infrastructure has made semiconductor supply increasingly tight. Dell Technologies (DELL-US) CEO Michael Dell also noted that enterprise deployment of AI applications is driving explosive growth in memory demand.
As one of the largest HBM buyers, NVIDIA (NVDA-US) uses HBM extensively in its AI GPUs, causing global hyperscale data centers to continue absorbing market supply. The market widely expects HBM to remain one of the most critical components for AI servers over the next several years.
According to recent reports from multiple research institutions and supply chain sources, HBM production capacity is already nearly fully booked through the end of 2026, with most of 2027’s capacity also pre-reserved by major customers. NVIDIA, AMD (AMD-US), Google, and other major cloud service providers continue to expand procurement, allowing memory manufacturers to secure high-margin long-term contracts and further boost profitability.
For Samsung Electronics, SK Hynix, and Micron, shifting more capacity to HBM is reflected in their financial performance. As the share of high-priced AI memory continues to grow, the gross margins of their data center-related businesses have clearly improved over the past year, outperforming previous product mixes dominated by consumer DRAM.
However, the market does not expect this supply-demand imbalance to persist indefinitely. Micron is building new wafer fabs in Idaho and New York, while Samsung and SK Hynix are also continuously expanding HBM and DRAM capacity, with new production expected to come online over the next two years. Meanwhile, China’s CXMT (ChangXin Memory Technologies) is rapidly expanding its presence in the mainstream DRAM market, potentially increasing supply and exerting downward pressure on prices.
Analysts point out that in the short term, AI will remain the top priority for global memory capacity allocation. As long as AI data centers continue to offer significantly higher returns on investment than the consumer market, the world's three major memory makers will continue to prioritize HBM supply, leaving consumer electronics like PCs and smartphones to compete for the remaining capacity—meaning general DRAM prices will likely remain elevated in the near term.
FACT BOX
- Source: PR Times
- Category: News
- Organizations: Google
- Products / services: HBM / DRAM