Samsung Raises Q2 DRAM Prices by 30% Amid Surging AI Demand

Samsung Electronics raised its Q2 DRAM contract prices by an average of 30%, signaling that AI-driven demand continues to outpace supply.

This price surge is driven by memory manufacturers reallocating production capacity to high-demand HBM, creating a supply squeeze for conventional DRAM.

While prices are expected to continue rising, potential bottlenecks like power shortages for new data centers could moderate the pace of future demand.

삼성전자

Samsung's recent decision to raise Q2 DRAM prices by an average of 30% confirms a powerful trend reshaping the memory market.

At the heart of this story is the explosive growth in AI infrastructure. As tech giants pour billions into building data centers for AI, the demand for specialized, high-performance memory like HBM (High Bandwidth Memory) has skyrocketed. This isn't just a minor shift; it's a fundamental change in priorities.

This leads to the core causal chain. First, memory manufacturers like Samsung and SK hynix are reallocating their precious wafer production capacity to where the demand—and profit—is highest: HBM and advanced server DRAM. This makes perfect business sense, but it has a significant knock-on effect. Second, as more resources are diverted to these high-end products, the supply of conventional DRAM, used in everything from PCs to standard servers, becomes structurally tight.

Consequently, we're seeing a "catch-up" price surge for this general-purpose memory. The market already witnessed a record-breaking price jump of over 90% in the first quarter of 2026. Samsung's 30% hike is the latest chapter in this story, confirming that suppliers hold strong pricing power. While the rate of increase may be slowing compared to Q1, the absolute price levels are climbing steeply.

However, a key risk is emerging on the horizon. Reports indicate that many new data center projects in the US are facing delays due to power and equipment shortages. This infrastructure bottleneck doesn't erase the underlying AI demand, but it could slow down the pace at which that demand translates into actual memory orders. The long-term trend remains strong, but the short-term path could be a bit bumpy.

HBM (High Bandwidth Memory): A high-performance type of computer memory used in conjunction with high-performance graphics accelerators and CPUs. It's essential for AI applications that process vast amounts of data.
CAPEX (Capital Expenditures): Funds used by a company to acquire, upgrade, and maintain physical assets such as property, plants, buildings, technology, or equipment. In this context, it refers to spending on data centers and AI hardware.
DRAM (Dynamic Random-Access Memory): A type of semiconductor memory that is the most common computer memory used today. "Conventional" or "general-purpose" DRAM refers to standard modules for PCs, mobile devices, and servers, as opposed to specialized HBM.