Elon Musk Identifies Memory as Primary Constraint for Agentic AI

Elon Musk identified memory as the primary silicon-based constraint for the development of agentic AI, noting that memory rather than compute is the current rate limiter. This shift occurs as artificial intelligence evolves from prompt-based generative systems toward agentic AI capable of independent planning and execution.

These advanced systems exponentially increase the demand for high-capacity, high-bandwidth DRAM and NAND flash storage to handle state tracking, tool outputs, and vector data. While memory stocks for providers such as Micron Technology, SK Hynix, and SanDisk Inc. experienced a pullback in July 2026 due to market skepticism and profit-taking, the technical requirements for agentic AI suggest a long-term growth cycle.

Goldman Sachs Private Wealth Management estimates that agentic AI token usage will reach approximately 120 quadrillion tokens per month by 2030. This projection indicates a sustained up-cycle for memory and storage providers as the industry moves toward more autonomous AI agents.