Micron CEO Says AI Rewrote Memory’s Boom and Bust Cycle as MUB Gains 5.6%

Micron Technology CEO Sanjay Mehrotra says artificial intelligence is fundamentally changing the economics of the memory industry, turning memory from a frequently oversupplied component into strategic computing infrastructure.

His comments coincided with a strong session for Micron, although there is insufficient evidence to attribute the rally entirely to the interview.

MU gained 3.97% Thursday to close at $974.33, giving Micron a market capitalization of approximately $1.1 trillion. The stock was indicated near $987.81 in Friday premarket trading, according to Google Finance.

Micron’s tokenized stock also advanced. MUB traded at $984.02 on Binance, rising 5.57% over a rolling 24 hour period. The token generated approximately $27.4 million in trading volume.

MUB’s reported market capitalization of $55.3 million represents the value of its circulating token supply. It should not be confused with Micron’s corporate market value.

Micron CEO Says Memory’s Value Equation Has Changed

Speaking with CNBC’s Jim Cramer, Mehrotra argued that AI has changed how customers use and value memory.

“Today there is no AI without memory,” Mehrotra said, explaining that AI systems require greater capacity, faster performance and lower power consumption.

“The value of memory, that equation has totally changed,” he added.

The memory industry has historically moved through severe boom and bust cycles. Rising prices encouraged Micron, Samsung and SK Hynix to expand production. Once new supply exceeded demand, memory prices collapsed, forcing manufacturers to reduce output and capital spending.

Mehrotra’s argument is not that oversupply has become impossible. It is that memory now plays a more important role in overall system performance, particularly as AI models require increasingly large quantities of data.

Modern accelerators cannot operate efficiently without fast access to high bandwidth memory. Demand is also expanding beyond data centers into autonomous vehicles, robotics and AI powered consumer devices.

Mehrotra consequently described memory as strategic infrastructure for the AI era rather than a basic component selected primarily according to price.

Five Year Contracts Could Make Earnings More Durable

The strongest evidence supporting Mehrotra’s argument may be the way customers are securing future supply.

According to the CEO, Micron’s data center customers want approximately 50% more memory than the company can currently commit to delivering. He also said customers across Micron’s end markets are prepared to purchase everything the company can produce.

Micron disclosed 16 Strategic Customer Agreements during its latest earnings call and has signed additional agreements since then.

Most of the committed volume runs through 2030 under five year terms, while smaller automotive agreements can run for three years. The contracts contain binding purchase commitments and are expected to eventually cover approximately 50% of Micron’s revenue, according to details presented during a recent technology leadership forum.

The agreements also contain pricing floors and ceilings. Those provisions could protect profitability if spot market prices weaken while giving customers greater certainty about future costs.

This structure does not eliminate the memory cycle. However, it could reduce the severity of future downturns by lowering the risk that Micron builds capacity without customers committed to purchasing the resulting output.

AI Is Making Memory a Performance Constraint

Micron executives recently said processors can spend approximately half their operating time waiting for data. This makes memory capacity and bandwidth increasingly important to overall AI system performance.

High bandwidth memory addresses part of this problem by moving information to AI processors considerably faster than conventional memory. However, HBM is more complicated to manufacture and consumes more wafer capacity, limiting how quickly producers can increase supply.

Micron began high volume production of HBM4 during the first quarter of 2026. The product was designed for Nvidia’s Vera Rubin platform and provides more than 2.8 terabytes per second of bandwidth, according to Micron’s announcement.

Customized HBM4E products are also under development. Micron expects memory market conditions to remain tight beyond 2027 as structural supply constraints struggle to keep pace with AI demand.

The shift is already visible in Micron’s financial results. Fiscal third quarter revenue reached a record $41.46 billion, compared with $9.3 billion one year earlier. Non GAAP gross margin expanded to 84.9%, while adjusted earnings reached $25.11 per share, according to the company’s official earnings release.

Micron Expands Its Long Term US Investment

Micron is supporting its outlook with one of the semiconductor industry’s largest expansion programs.

The company plans to invest more than $250 billion in US manufacturing and research through 2035. Its Idaho development will eventually contain two major fabrication facilities, with the first expected to begin wafer production during the middle of 2027 and the second targeted for late 2028.

Micron also unveiled Micron Research Labs, a Boise based research institution backed by a planned $10 billion investment over the next decade.

The company expects to begin construction in 2027. The institution will focus on memory technology, advanced computing architectures, packaging and future semiconductor manufacturing.

This investment is separate from Micron’s previously announced commitment of more than $250 billion.

What the Comments Mean for MU and MUB

Mehrotra’s comments strengthen the long term bullish case for Micron because they are supported by record revenue, constrained supply and binding customer commitments.

However, saying AI has changed the memory cycle is not the same as proving that the cycle has disappeared.

Micron shares fell more than 7% earlier this week before recovering, demonstrating how quickly investors can reduce exposure when interest rates, semiconductor valuations or AI spending become concerns.

New factories will eventually increase supply. Samsung, SK Hynix and Chinese memory manufacturers are also expanding their capabilities. Meanwhile, Micron stock has risen approximately 700% during the past year, leaving limited room for weaker demand or execution problems.

For MUB holders, the primary catalyst remains the underlying Micron stock rather than an independent cryptocurrency development.

If AI memory demand continues exceeding supply and Micron’s customer agreements protect pricing, both MU and MUB could remain supported. The decisive test will be whether those contracts make the next memory downturn meaningfully less severe or merely postpone it.