About 20 Years of Price Decline Reversed Within Months: Study Says 2026 Memory Price Per GB Returns to 2007 Levels

Research by Daniel Lemire, a professor at the University of Quebec in Canada, reveals abnormal fluctuations in the memory market. Driven by surging demand for High Bandwidth Memory (HBM), memory unit costs have skyrocketed to 2007 levels within months, completely reversing the trend of continuous price declines over the past twenty years.

As a core component of high-performance computing, HBM is widely used in AI training and inference chips. The surge in parameters for large models places extremely high demands on memory bandwidth, causing HBM to become a key bottleneck in the AI computing power supply chain. Its demand growth directly reflects the acceleration of global AI infrastructure construction.

The drastic fluctuation in memory prices will have a direct impact on AI hardware costs. AI chip manufacturers may face higher material costs, which could subsequently drive up data center construction costs, or prompt the industry to accelerate the development of more efficient memory architectures to alleviate supply pressure.

Lemire defines this phenomenon as a "historical anomaly," emphasizing the impact of current AI computing power demand on the pricing logic of traditional hardware markets. This trend indicates that AI technology evolution is no longer limited to the software level, but is profoundly reshaping the supply and demand relationships and cost structures of underlying hardware.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.