Micron CEO makes the case for a longer AI-driven memory cycle

Micron memory cycle outlook was the focus of CEO Sanjay Mehrotra’s interview with Jim Cramer on CNBC’s “Mad Money.” Mehrotra argued that artificial intelligence is changing demand for memory chips and that new supply will take time to arrive. For investors watching Micron and the broader chip group, the key question remains whether those forces can prevent the familiar swing from shortage to oversupply.

According to CNBC, Mehrotra said memory has become a key enabler of AI. His argument rests on three connected points: AI needs more memory in more places, a specialized AI memory product is taking up capacity, and new production facilities do not begin supplying customers quickly.

Why does Micron think the memory cycle could last longer?

Mehrotra argued that the current AI buildout could produce a different outcome from prior memory booms. That is a management view, not a confirmed forecast. The traditional bear case is straightforward: when memory makers build enough capacity to exceed demand, pricing power and profitability can fall sharply. CNBC said even a small supply surplus has historically pressured prices.

1. AI demand reaches beyond data centers

Mehrotra told Cramer that AI systems need more memory, as well as higher-performance and lower-power memory. He said the demand is not limited to the data centers used to train and run AI models. Phones, personal computers and self-driving cars also need increasing amounts of memory as they take on more AI-related tasks.

CNBC pointed to the change in smartphone memory as one illustration: the original iPhone had 128 megabytes of dynamic random-access memory, or DRAM, while current models have 8 to 12 gigabytes. One gigabyte equals 1,024 megabytes, according to the report.

Mehrotra also mentioned robotics as a potential future source of demand. CNBC presented that as a longer-term possibility rather than a measured current market driver.

2. HBM is competing for DRAM capacity

The leading AI chips from Nvidia, AMD and Google use high-bandwidth memory, or HBM, CNBC reported. HBM is a specialized type of DRAM, described in the report at a high level as individual DRAM wafers stacked together.

Micron, SK Hynix and Samsung are directing more capacity toward HBM, according to CNBC. The report said this contributes to tight supply because AI chips, smartphones, laptops and other products compete for underlying DRAM capacity. That capacity shift is central to Mehrotra’s case that adding demand from AI can affect a wider range of memory products.

3. Added capacity has a multiyear delay

Micron is building two fabrication plants in Boise, Idaho, and another complex in Clay, New York, CNBC reported. But the report also emphasized a multiyear gap between beginning construction and delivering chips to customers.

That delay matters to the cycle argument. More construction can eventually increase supply, but it does not immediately relieve a shortage. Mehrotra’s position is that demand can remain ahead of available supply during that buildout.

The test for this view is still supply and demand. A slowdown in AI-related spending could weaken demand, while enough new output could restore the oversupply conditions that have defined earlier downturns. Mehrotra made the case that AI has raised memory’s role in computing; CNBC’s account leaves unresolved how long that imbalance will persist.

This story draws on original reporting from CNBC.