May 8, 2025 /SemiMedia/ — Rising AI demand is putting new pressure on the global memory market, and memory shortages may continue for several more years, according to Sanjay Mehrotra, CEO of Micron Technology.

Speaking in an interview cited by Wccftech and CNBC, Mehrotra said the AI industry is still in its early stage. As AI models become larger and inference workloads keep growing, data centers will need faster and higher-capacity memory products to support future computing demand.

Micron said demand from both traditional servers and AI servers remains strong. However, limited DRAM and NAND production capacity is making it difficult for suppliers to fully meet market demand. The company expects AI-related demand for DRAM and NAND to account for more than half of the industry’s total addressable market this year.

The rapid growth of AI inference is also increasing demand for high-bandwidth memory products. Industry suppliers are now moving toward higher-layer HBM solutions with improved performance and bandwidth for next-generation AI GPUs.

According to South Korean media reports, 12-layer HBM4 products have entered volume production ramp-up and are expected to support future AI accelerator platforms, including systems based on the Vera Rubin architecture.

Analysts cited by the Financial Times said the current supply imbalance may not ease before 2028 because new wafer fabs require years to build and qualify. Some industry forecasts also suggest global DRAM capacity could meet only around 60% of market demand by 2027 if AI growth continues at the current pace.