Micron Technology ((MU Quick QuoteMU - Free Report) ) and Sandisk Corporation ((SNDK Quick QuoteSNDK - Free Report) ) have become two of the defining stocks of the current AI infrastructure trade. After years of being viewed as highly cyclical commodity memory companies, both stocks have exploded higher as investors reprice the value of DRAM, NAND, high-bandwidth memory and enterprise storage in the AI era.

The move has been extraordinary to say the least. Micron has surged as demand for high-bandwidth memory has overwhelmed available supply, while Sandisk has become one of the market’s most dramatic winners as NAND flash and enterprise SSD demand have tightened sharply. Sandisk is up nearly 500% year to date and more than 3,300% in the last year, while Micron has gained 730% in that time.

The reason is simple: memory has become one of the key bottlenecks in the AI buildout. Both stocks boast a Zacks Rank #1 (Strong Buy) rating, but can investors still reasonably consider buying shares after such a massive rally?

Image Source: Zacks Investment Research

Memory Is No Longer an Afterthought

For much of the last decade, investors focused primarily on GPUs, cloud providers and hyperscale data centers as the center of the AI trade. But as AI workloads scale, memory is increasingly becoming one of the most critical bottlenecks in the entire AI infrastructure stack.

Modern AI systems are extraordinarily memory intensive. GPUs like those from Nvidia ((NVDA Quick QuoteNVDA - Free Report) ) can only perform efficiently if they are constantly fed data at extremely high speeds. That is where high-bandwidth memory (HBM) comes in. HBM sits physically next to the GPU inside advanced AI servers and provides the ultra-fast data transfer speeds required for training large language models and running inference workloads.

Without enough HBM, even the most powerful GPUs can become bottlenecked.

Traditional DRAM also remains essential throughout the stack. DRAM is used in AI servers, networking equipment, CPUs and cloud infrastructure to temporarily store and process active workloads. As models become larger and context windows expand, the amount of DRAM required per server rack continues to rise sharply.

NAND flash and enterprise SSDs are equally important. AI systems require enormous amounts of storage capacity to hold training datasets, model weights, vector databases and inference data. Enterprise SSDs allow hyperscalers to quickly retrieve and move these massive datasets throughout the data center. As inference scales globally, the demand for fast, power-efficient storage continues to accelerate.

The result is that memory is no longer just a commodity attached to PCs and smartphones. It has become critical infrastructure for the AI technology stack.

This is why companies like Micron and Sandisk have seen such dramatic stock moves. The market is beginning to realize that AI demand is not just about compute. It also requires vast amounts of high-speed memory and storage, and supply has struggled to keep pace.

The Fundamental Momentum Is Real

The rally in both stocks has been supported by extraordinary earnings growth and improving industry fundamentals, not simply speculative enthusiasm.

Micron’s recent results highlighted just how tight the market has become. Quarterly revenue nearly tripled year-over-year, with the company reporting record revenue across DRAM, NAND and HBM products. Management also confirmed that much of its 2026 HBM capacity is already sold out, giving the company unusual pricing power and visibility for a memory producer.

Image Source: Zacks Investment Research

That visibility is a major reason Micron stock has exploded higher. Historically, memory companies operated with little visibility and highly volatile pricing. Today, AI customers are effectively reserving supply years in advance, changing the economics of the business, at least for now.

Sandisk’s numbers were arguably even more dramatic. The company recently reported fiscal Q3 revenue of $5.95 billion, up roughly 250% year-over-year, while earnings surged to $23.41 per share, crushing expectations. Data center revenue jumped more than 200% sequentially as demand for AI-related SSD and NAND products accelerated sharply.

Image Source: Zacks Investment Research

The company’s guidance also stunned investors. Sandisk forecast next-quarter revenue between $7.75 billion and $8.25 billion, far ahead of expectations, while announcing major long-term supply agreements and a $6 billion share repurchase authorization.

Importantly, both companies are benefiting from the same underlying dynamic: demand is growing faster than supply. High-bandwidth memory, enterprise SSDs and NAND flash have become critical infrastructure for AI systems, and producers are struggling to expand capacity quickly enough to meet demand.

That does not mean the cycle has disappeared. Memory has always been cyclical, and eventually supply responds. But this cycle already looks structurally different from prior booms because AI demand appears more durable, customers are signing longer-term contracts, and memory has become strategically critical to the broader compute ecosystem.

But Memory Cycles Still Exist

The key caveat is that memory has always been cyclical. Historically, these stocks have gone through violent boom-and-bust periods. Tight supply leads to higher prices, higher prices lead to aggressive capacity expansion, and eventually supply catches up just as demand growth slows. When that happens, pricing can fall quickly and earnings can collapse.

The current cycle is unique because AI demand appears more durable, supply additions are harder, and advanced memory is becoming more strategic. But it is still a cycle. The question is not whether the cycle exists. The question is whether this cycle lasts longer, peaks higher and resets from a structurally better level than prior cycles.

Should Investors Buy Shares in MU and SNDK?

Micron and Sandisk can continue to work if memory prices keep rising, AI infrastructure demand remains strong, and supply remains constrained. For now, that appears to be the case.

Both stocks still boast notable top Zacks Ranks, but after such enormous runs, investors should avoid pretending these are risk-free compounders. Memory stocks can look cheapest near peak earnings and most expensive near trough earnings. That means valuation alone can be misleading.

The best way to think about Micron and Sandisk is as two of the strongest beneficiaries of the AI memory shortage, but still within a cyclical industry. This cycle may be different in scale, duration and strategic importance. But cycles have not been repealed.