Micron CEO Says AI Memory Demand Could Keep Chip Supply Tight Beyond 2027

Micron Technology is betting that the memory shortage powering the AI boom will last longer than the current semiconductor cycle. Sanjay Mehrotra, Chairman, President and CEO of Micron, said on CNBC's "Mad Money" on August 20 that supply constraints could continue beyond 2027 as AI systems demand more advanced memory.

"Memory has become today a key enabler in the AI era," Mehrotra told host Jim Cramer. He said every AI system needs more memory and higher memory performance to reach its full potential, creating what he described as long-term demand across data centers, smartphones, PCs, automobiles and other applications.

The comments come as Micron's financial results show how sharply the economics of memory have changed. The company reported $41.46 billion in fiscal Q3 2026 revenue, compared with $23.86 billion in the previous quarter. Gross margin reached 84.6%, while net income rose to $28.24 billion. Micron forecast fiscal Q4 revenue of about $50 billion.

The shift is also visible in the contracts Micron is signing. The company said it had completed 16 Strategic Customer Agreements (SCAs) by its June earnings release, covering data center, consumer and automotive customers. Its latest filing says the agreements generally include binding multi-year volume commitments, with pricing either fixed or subject to minimum and maximum levels.

AI Is Reshaping the Memory Market

The demand is putting pressure on memory supply beyond high-end AI systems. A previous analysis of the shortage found that AI data center expansion was competing with consumer electronics and other industries for memory capacity.

Mehrotra said Micron's customers themselves had struggled to forecast the scale of the current demand. He also pointed to the industry's reduced investment during the 2023 downturn, when memory prices fell sharply and weakened manufacturers' ability to add capacity.

Micron is now trying to reverse that supply constraint through large manufacturing and research investments. Mehrotra said the company plans to invest $200 billion across the US, while expanding manufacturing and R&D globally. He said the investments will take years to translate into meaningful new supply because new fabs require greenfield construction and lengthy qualification processes.

Micron's current manufacturing plans illustrate that lag. The company expects its first Idaho fab to produce initial wafers in 2027, with production ramping primarily in 2028 after customer qualification. Its New York project is also moving through construction, adding further capacity later in the decade.

The Shortage Is Reaching Beyond Data Centers

The supply squeeze is already affecting hardware markets outside AI infrastructure. Conventional DRAM prices have risen sharply this year, while manufacturers have been allocating more capacity toward high-bandwidth memory (HBM) used in AI systems.

That pressure has already reached enterprise PC procurement, where memory costs have become a significant component of device prices.

Micron is trying to protect its exposure to those markets even as data centers become its largest memory customer base. Mehrotra said about 40% of Micron's revenue comes from consumer, automotive and industrial markets, making diversification important as the company expands its AI business.

HBM is central to that strategy. Micron said HBM4 was already in high-volume shipments for its lead customer's platform in its June earnings release, with HBM4E development underway and volume production expected in 2027.

The result is a memory market increasingly shaped by AI infrastructure demand, long-term customer commitments and years-long capacity expansion cycles.