1 hour ago
22
Micron Technology just told the market something it didn’t expect to hear: the global memory chip shortage isn’t going away anytime soon. The company’s fiscal Q3 2026 earnings, reported on June 24, revealed that supply constraints previously expected to ease by early 2027 will now persist well beyond that date.
The culprit is AI. Demand for high-bandwidth memory and DRAM products built for data centers has surged past every prior forecast, and new manufacturing capacity simply can’t come online fast enough to close the gap.
The numbers behind the squeeze
Micron’s latest guidance represents a meaningful revision from earlier projections. The company and its competitors, Samsung and SK Hynix, had previously suggested the tight supply environment would begin loosening around early 2027.
That timeline is now obsolete. Micron’s management has updated its outlook to reflect an acceleration in AI inference and training workloads that continues to outstrip supply additions across the entire industry.
Micron’s new fabrication facility in Idaho is expected to begin initial output around mid-2027. But reaching significant production levels won’t happen until 2028, thanks to the lengthy qualification periods required before a fab can churn out chips at scale. These facilities take more than two years to go from groundbreaking to meaningful production.
Micron’s stock price jumped following the earnings announcement, buoyed by the company’s pricing power and long-term contracts reportedly totaling around $100 billion.
Why AI keeps eating memory
The three major memory suppliers, Micron, Samsung, and SK Hynix, are all seeing the same pattern: orders growing faster than their ability to fulfill them.
The innovation wildcard
The concern, flagged alongside Micron’s otherwise stellar results, is that innovation by major buyers could eventually optimize how memory is used in AI workloads. This could take several forms: more efficient model architectures that require fewer parameters, novel compression techniques that reduce memory footprints, or custom silicon designs that integrate memory more tightly with compute, reducing the need for external DRAM or HBM.
If demand growth decelerates while new capacity, like Micron’s Idaho fab, finally comes online in 2028, the market could shift from shortage to surplus faster than anyone currently expects.
The pricing power and contract visibility that drove Micron’s stock higher are real, at least through 2027. The company’s roughly $100 billion in long-term contracts provides a revenue floor that most chipmakers would envy.
Investors should watch two signals closely. First, capacity timelines: any further delays at Micron’s Idaho facility or comparable projects from Samsung and SK Hynix would extend the shortage and support pricing. Second, customer innovation: announcements from major cloud providers about memory-efficient AI architectures or custom memory solutions would be early warning signs that the demand runway may be shorter than current projections suggest.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·