Micron CEO: Memory Supply Tightness to Extend Beyond 2026, the AI Race Is Fundamentally a Memory Race
Taylor Wilson
Micron CEO Sanjay Mehrotra says the memory supply crunch is a structural constraint, not a short-term mismatch, and will persist beyond 2026. He frames memory as AI infrastructure's core bottleneck — which means the market may need to re-price memory stocks from cyclical to secular growth.
Why isn't the supply crunch temporary?
A new fab takes three to four years from groundbreaking to first wafer, then more time to ramp. This means → even fabs started today won't deliver meaningful output until around 2029.
The deeper bottleneck is technical: as process nodes advance, each wafer yields less additional storage capacity. In plain terms = fabs keep getting more expensive, but each new fab adds less memory than the last.
Mehrotra says Micron saw this coming around 2021 and moved early into HBM — high-bandwidth memory, a chip designed specifically for AI workloads. At the time, HBM was less than 1% of the memory industry.
Why does AI keep demanding more memory?
Three forces are pulling at once: models are getting larger, inference demand is surging, and agent AI — autonomous systems that execute tasks on their own — is rising fast.
Mehrotra highlights token economics: longer context windows, growing KV-cache requirements, and ever-larger models. In plain terms = AI doesn't just need computing power to think — it needs "recall" to hold conversations and context.
He believes this point has been chronically underestimated: "All of this is still at a very, very early stage." This reflects a structural trend — AI memory demand isn't a one-time spike but scales continuously as models evolve.
What's the logic behind a $200 billion bet?
Micron previously announced a $200 billion investment to build a domestic memory manufacturing system in the U.S. Mehrotra says the decision rests on "discipline" — investment grounded in data and fundamentals.
Asked whether he ever had doubts, his answer was direct: "We had no self-doubt. We absolutely believe in the memory opportunity."
This means → Micron isn't betting on a short-term AI wave. It's betting on memory's long-term structural role as AI infrastructure. Investment pace will be continuously assessed against demand forecasts and technology progress.
The stock tripled, then pulled back — what now?
Micron surged 87.8% in May alone; its market cap briefly topped $1 trillion for the first time. It has since pulled back about 11% in June, but the year-to-date gain is still roughly 203%.
Two immediate triggers drove the pullback: Broadcom's earnings cooled AI-sector sentiment, and May nonfarm payrolls came in at 172,000 — far above the 80,000 economists expected — fueling fears the Fed may resume rate hikes.
This means → the pullback reflects short-term sentiment and macro pressure, not a fundamental deterioration. The unanswered question: whether structural supply constraints can hold long enough as AI demand keeps expanding — that will determine if Micron's valuation thesis plays out.
Content is for reference only, not financial advice.