The semiconductor memory industry is undergoing a fundamental transformation. Micron CEO Sanjay Mehrotra stated that the era of cyclical booms and busts, which defined the economics of this market for decades, is coming to an end. The key driver of change is the rapid proliferation of artificial intelligence systems.
Speaking at the construction site of a new plant in Boise, Idaho, Mehrotra emphasized that memory has become a strategic resource without which AI infrastructure cannot function. Micron, as a reminder, is implementing an ambitious capital investment plan of $250 billion in manufacturing capacity and research within the United States.
Why the old model no longer works
Traditionally, the memory market developed along a classic cycle: rising demand prompted the construction of new capacity, which ultimately led to overproduction and price collapses. However, according to Mehrotra, AI has shattered this logic. Demand has become more sustainable and predictable, and purchasing has become less price-sensitive.
Modern memory chips are now designed in close conjunction with specific processors and systems. This means that buyers can no longer simply choose the cheapest supplier, as in a typical commodity market. The performance of an entire server now critically depends on memory characteristics, which changes customer priorities.
"That is why I call memory the strategic infrastructure of the AI era. Today, it is impossible to imagine AI without memory. AI-based systems require ever more memory," he stated.
Figures confirming the shift
Micron's financial results fully confirm this thesis. The company's revenue for the third fiscal quarter soared to $41.46 billion, whereas in the same period a year earlier, this figure was only $9.30 billion. Gross margin reached an impressive 84.6% compared to 37.7% a year earlier, and in the fourth quarter the company expects it to rise to approximately 86%.
The main problem now is not a lack of demand, but a shortage of manufacturing capacity. Mehrotra admits that factories cannot keep up with customer requests. Data centers alone require approximately 50% more chips than Micron is currently able to produce. In the future, according to his forecasts, demand will grow even further—driven by manufacturers of autonomous vehicles, robots, and household appliances with built-in AI.
However, not everyone shares this optimism. In June, a class-action 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 harsh price competition to the market—the very competition that, according to Mehrotra, the industry has moved away from thanks to AI.
My analysis: Mehrotra's statements look convincing, but it should not be forgotten that high margins are the best incentive for new players to emerge. If Chinese manufacturers can overcome the technological gap, market cyclicality may return, but at a new, more complex level. For now, AI has indeed created a "golden age" for memory manufacturers, and the capitalization of companies in the sector reflects this.