A massive imbalance in memory supply and demand could benefit Micron through at least 2027, according to a William Blair analyst.

Memory chips have emerged as a key bottleneck to the artificial-intelligence buildout, and Micron Technology can keep cashing in on soaring prices for at least the next two years, according to an analyst.

The supply of memory products has been squeezed as AI chip makers have started rolling out advanced chips on an annual cadence. At the same time, memory and storage companies have been reluctant to add more capacity due to fears of oversupply in a historically cyclical industry. This severe supply-and-demand imbalance has given memory and storage makers more pricing leverage.

William Blair analyst Sebastien Naji expects tight supply to continue into 2027, setting Micron up to benefit from "significant" increases in average selling prices that boost the company's profitability.

In his view, Micron is the No. 2 memory supplier, and he expects its high-bandwidth-memory revenue to grow 164% in 2026 and 40% in 2027. Some of the company's other memory offerings should also benefit from the AI boom, he said.

Over the next two years, Naji said, Micron's adjusted earnings per share could grow more than 275%.

Naji initiated coverage of Micron with an outperform rating on Thursday, pointing to the "memory supercycle" that has sent the chip maker's shares up 256% over the last year.

Micron, Samsung Electronics and SK Hynix are the top three suppliers in the world of HBM, a type of dynamic random-access memory that is in high demand from Nvidia and other AI-chip makers to meet the needs of increasingly advanced AI models.

As large language models get longer context windows, or the amount of data they can process at one time, HBM has become "a critical enabler of the AI platform shift," Naji said. The growing need for HBM in next-generation chips means "memory represents a growing share of the AI-infrastructure bill of materials."

While Naji forecast a near quadrupling of adjusted EPS over a two-year span, he acknowledged several factors that could hamper growth in that metric, and in free cash flow. For one, he noted, Samsung, which had fallen behind on the HBM curve, now has its HBM3e qualified for use in Nvidia's chips. The South Korean chip maker also has a good relationship with Alphabet and Broadcom, Naji said, which work together on Google's tensor processing units that power the tech giant's Gemini AI models.

Meanwhile, imbalances of supply and demand are expected to even out in the medium to long term, Naji said.

Micron's stock is "not necessarily expensive" compared with other AI-semiconductor plays, Naji said. At a forward price-to-earnings multiple of 9.7x his estimate for 2026, the stock trades below its historical average of 11x, he noted.