Micron CEO's Latest Interview: 'Storage' is an Overlooked Bottleneck in AI, Supply Constraints to Continue
Micron CEO Sanjay Mehrotra emphasizes that memory and storage are a critical, yet underestimated, bottleneck for AI development, stating that the "AI race is not just a compute race, it's a memory race." He argues that AI's evolution—driven by larger models, longer context windows, and growing token consumption—demands not only more computing power but significantly enhanced "memory" or storage capacity and bandwidth.
Mehrotra highlights a structural supply constraint, predicting tight industry supply will persist beyond 2026. Building new fabs takes 3-4 years, with subsequent ramp-up periods. Furthermore, advancing technology nodes yield diminishing bit output gains per wafer. He notes that the manufacturing complexity of memory, involving physics, chemistry, and mass production precision, is vastly underappreciated.
Despite announcing a $200 billion U.S. manufacturing investment, Mehrotra underscores that such decisions are based on disciplined analysis of data, technology trends, and customer needs, not speculation. He expresses no doubt about the long-term opportunity in memory, driven by AI's early-stage growth, while emphasizing the need for resilience, agility, and long-term focus to navigate industry cycles.
marsbit06/10 02:10
