Micron surges 233% in 2026 as AI memory demand outstrips supply

Micron Technology has sold out its high-bandwidth memory chips for the full year, a sign that AI infrastructure demand is overwhelming supply capacity across the semiconductor industry.

Micron Technology Inc. (Nasdaq: MU) has emerged as one of the biggest beneficiaries of the artificial intelligence infrastructure buildout, with shares surging 233% year to date and 918% over the past 12 months. The memory chip maker's high-bandwidth memory (HBM) — a critical component in Nvidia Corp.'s AI accelerators — is already fully contracted for 2026, allowing the company to raise prices and expand margins at a pace rarely seen in the cyclical semiconductor industry.

"The memory trade is alive and well," Cantor Fitzgerald analyst C.J. Muse said, raising his price target on Micron to $1,500 and predicting memory chips will remain undersupplied through 2028. Wells Fargo also increased its price target following the company's latest earnings report.

In its fiscal second quarter of 2026, Micron reported revenue of $23.9 billion, up 195% from a year earlier and 76% from the prior quarter. Gross margin in its cloud memory business reached 74%, compared with 55% in the same period last year, while data center gross margin hit 74%, up from 47%. For the current quarter, the company forecast revenue of $33.5 billion — a 40% sequential increase — with gross margin expanding to 81% and earnings per share of $18.90, up from $12.07 last quarter.

The stock touched $1,079 on June 3 before a broad semiconductor selloff — triggered by Broadcom Inc.'s earnings report and a stronger-than-expected US jobs report that reduced the case for rate cuts — pushed it down about 12% to $874. It has since recovered to around $947. Even after the pullback, Micron trades at a forward price-to-earnings ratio of 10, a level that last appeared in August 2025 when the stock was at $119 per share. Its five-year price-to-earnings-to-growth (PEG) ratio stands at 0.37, a level typically associated with undervalued stocks.

Why HBM became the bottleneck

High-bandwidth memory stacks DRAM chips vertically to deliver data to AI processors at speeds traditional memory cannot match. As Nvidia, Advanced Micro Devices Inc., and a growing roster of custom chip designers compete for HBM supply, Micron — alongside South Korea's SK Hynix Inc. and Samsung Electronics Co. — controls nearly all of the available capacity.

Nvidia confirmed that Micron will supply HBM for its Vera Rubin AI chip architecture, a design win that locks in demand through the next product cycle. The partnership with Nvidia, combined with Jensen Huang's recent Asia tour that produced multiyear agreements with SK Hynix and other Korean technology firms, has reinforced the view that memory supply will remain tight for years. Huang said the industry is still at "the outset of the AI revolution."

Valuation disconnect or new normal?

The stock's forward P/E of 10 sits in value territory by traditional measures, yet the company is growing revenue at triple-digit rates. The disconnect reflects lingering skepticism about whether Micron can sustain this growth trajectory — a question that hinges on the durability of AI capital spending by hyperscale cloud providers.

Micron's market capitalization has swelled to $1.1 trillion, placing it among the largest US semiconductor companies. Cantor Fitzgerald's $1,500 price target implies roughly 58% upside from current levels, based on the thesis that memory undersupply will persist through 2028. The company's HBM supply is locked under long-term contracts, providing revenue visibility that the cyclical memory industry has rarely offered investors.

For investors, the central question is whether the current valuation already prices in the HBM boom or leaves room for further gains. With a PEG ratio below 1 and a forward P/E that has historically preceded rallies — the stock rose 135% in the five months after its forward P/E last touched these levels — the data suggests the market has not fully priced in the duration of the AI memory cycle.

This article is for informational purposes only and does not constitute investment advice.