AI rewrote the rules of the memory market: Micron's CEO on the shift in economic paradigm
The semiconductor memory market is entering a new era, and this is not just another cyclical turn. Sanjay Mehrotra, CEO of Micron, during a recent speech at the site of a plant under construction in Boise, Idaho, stated that there has been a fundamental shift in the industry's economics. According to him, artificial intelligence has transformed memory from a routine commodity into a strategic resource, which fundamentally changes the rules of the game for all market participants.
Memory as Infrastructure
Mehrotra emphasizes that the traditional model, where memory developed cyclically—from demand booms to overproduction and price collapses—is becoming a thing of the past. Today, customers are increasingly refusing to choose suppliers solely based on price. Instead, chips are being tailored to specific processors and server architectures, making memory a critically important element of overall system performance.
"I call memory the strategic infrastructure of the AI era," he stated. "AI-based systems require ever more memory, and without it, their existence is impossible." This statement is backed by impressive financial results. Micron's 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%, up from 37.7% a year earlier, and the company forecasts it will rise to 86% in the fourth quarter.
Shortage and New Challenges
The main issue right now is production. Mehrotra admits that factories cannot keep up with demand: data centers alone require approximately 50% more chips than Micron can produce. In the future, he expects even stronger growth in demand from manufacturers of autonomous vehicles, robotics, and "smart" home appliances.
However, not everyone shares this optimism. In June, a lawsuit was filed against Micron, Samsung, and SK Hynix, accusing the manufacturers of collusion and artificially inflating DRAM memory prices. Chinese companies, such as CXMT, are also ramping up production, which could bring back the intense price competition that, according to Mehrotra, the industry supposedly moved away from thanks to AI.
My take: Mehrotra's statements sound convincing, but the memory market has already gone through periods of "structural shortage" that ended in overproduction. The key question is whether AI can truly smooth out cyclicality, or whether we are simply witnessing the peak of another cycle, followed by a correction. Investors should closely monitor the actions of Chinese manufacturers and the outcomes of antitrust proceedings.