DRAM Shortage: AI Consumes Memory and Energy

⚡

Key Takeaways

1AI, with its growing demand for DRAM, is worsening the global shortage, impacting large language models.

2The energy consumption of AI data centers could reach 12% of U.S. energy by 2028, with water consumption quadrupling.

3Giants like Nvidia and AMD are increasing their memory needs, driven by companies such as Google and Microsoft.

💡Why it matters — The DRAM shortage affects not only AI hyperscalers but also the consumer electronics market, exacerbating technological inflation.

The Impact of AI Hyperscalers on the DRAM Shortage

Samuel K. Moore's article, "How and When the Memory Chip Shortage Will End," highlights the current DRAM crisis, exacerbated by the insatiable appetite of AI hyperscalers. These digital giants, in their quest for ever more memory, are creating unprecedented pressure on available resources, directly affecting the performance of large language models.

A Significant Increase in Energy Consumption

The energy consumption of AI is a major concern. By 2028, it could account for up to 12% of the total energy consumed in the United States. In 2025, generative AI queries had already consumed 15 terawatt-hours, and this figure could reach 347 TWh by 2030. Meanwhile, the water consumption needed for cooling AI data centers is expected to double or even quadruple by 2028 compared to 2023.

HBM: A Crucial Memory Product

High Bandwidth Memory (HBM) is essential for AI processors. Manufacturers like Nvidia and AMD are increasing their memory requirements to meet the growing needs of companies such as Google, Microsoft, OpenAI, and Anthropic. These companies are financing the construction of gigantic data centers, like Meta's 5-gigawatt Hyperion site in Louisiana, posing considerable engineering challenges.

Prospects for Ending the Shortage

The question of when the DRAM shortage will end remains open. Moore indicates that signs of easing could emerge if one of the major HBM companies, such as Micron, Samsung, or SK Hynix, announces an adjustment to their production schedule. On the demand side, tech companies might adapt their supply chains by opting for hardware that requires less memory or by innovating with creative redesigns.

Consequences for the Consumer Market

The DRAM shortage is not limited to AI hyperscalers. It is also impacting the consumer device market, such as the Raspberry Pi, whose prices are rising due to pressure on electronic components. Persistent inflation and pricing changes in the United States partially mask this pressure, but the situation remains concerning for the tech sector.

Essential Media Coverage

Moore's story on HBM is both important and unique, which motivated its inclusion in a recent issue, with updates since its original publication on February 10. It is paired with an article by Matthew S. Smith, exploring how the memory chip shortage is driving up the prices of low-cost computers. The price pressure caused by the demand from AI hyperscalers across all types of consumer electronics is obscured by stubborn inflation combined with a constantly evolving pricing regime in the United States.