The semiconductor memory industry is entering a new era where previous cyclical supply and demand dynamics are giving way to structural changes. Micron CEO Sanjay Mehrotra called memory "strategic infrastructure for the AI era," declaring a fundamental shift in the market's economics.
Speaking at the site of a new semiconductor plant in Boise, Idaho, Mehrotra outlined the contours of a future in which memory ceases to be just a commodity. The company is already implementing a massive $250 billion investment plan in U.S. manufacturing and research, building two new facilities.
Why memory is becoming infrastructure
Traditionally, the memory market developed according to a classic cycle: rising demand stimulated the construction of new capacity, oversupply crashed prices, and everything repeated itself. However, according to Mehrotra, artificial intelligence has shattered this model. Demand is now more sustainable in nature, and buyers can no longer choose a supplier solely by price, as was the case with ordinary goods.
The key factor is the deep integration of memory chips with specific processors and systems. Server performance now directly depends on this pairing, which changes the very nature of procurement.
"That's why I call memory strategic infrastructure for the AI era," Mehrotra said. "Today, it's impossible to imagine AI without memory. AI-based systems require more and more memory."
The numbers confirm the shift
Micron's financial results fully confirm these words. Revenue for the third fiscal quarter soared to $41.46 billion, compared to $9.30 billion in the same period last year. Gross margin reached 84.6%, compared to 37.7% a year earlier, and in the fourth quarter the company expects to reach approximately 86%.
The main bottleneck right now is production. Mehrotra admitted that factories cannot keep up with customer demand: data centers alone require about 50% more chips than Micron can produce. In the future, demand will only grow—among prospective buyers, the company head named manufacturers of autonomous vehicles, robots, and AI-powered home appliances.
However, not everyone shares this optimism. In June, a lawsuit was filed against Micron, Samsung, and SK Hynix: plaintiffs accuse the manufacturers of collusion and inflating DRAM prices. Chinese companies such as CXMT are also ramping up output, which could bring back the fierce price competition—the very thing the industry, according to Mehrotra, has moved away from thanks to AI.
My take: Mehrotra's statements are not just corporate optimism but a marker of deep transformation. If memory was once a cyclical asset, it is now becoming a strategic resource comparable in significance to advanced lithography processes. However, growing pressure from China and antitrust risks could adjust this idyllic picture. Investors should closely monitor the balance between structural demand and new manufacturing capacity.