AI Memory Squeeze Puts Micron and Sandisk at the Center of a Booming Data Center Cycle

Artificial intelligence workloads are reshaping the memory and storage industry, and two companies are emerging as prominent beneficiaries of the resulting supply crunch: Micron Technology and Sandisk.

The dynamics behind the squeeze are straightforward. As AI training runs scale up and inference workloads move into production, each new rack of accelerators demands more memory bandwidth positioned close to the GPU. At the same time, converting existing wafer fabrication lines to produce high bandwidth memory (HBM) is a slow and expensive process, keeping supply constrained even as demand accelerates. On the storage side, NAND flash supply is similarly tight — not because of manufacturing limits alone, but because hyperscale operators are generating enormous volumes of data that must be persisted after it leaves the GPU.

Micron Technology sits on both sides of this demand wave. The company manufactures the HBM that stacks directly on top of GPU packages, as well as the server DRAM that fills out the rest of the AI compute stack. Investors appear to be taking note: Micron shares traded at $1,016.59 recently, up 6.58% from the prior close of $953.84, giving the company a market capitalization of roughly $1.15 trillion.

Sandisk, meanwhile, is exposed to the same cycle through the storage layer. The company develops and sells NAND flash-based solutions, including enterprise solid-state drives deployed in the same data centers driving the memory buildout. Sandisk stock recently changed hands at $1,740.00, a 12.3% gain over its previous close of $1,549.40, for a market cap near $186.4 billion.

The broader infrastructure context matters here as well. NVIDIA, whose GPUs anchor most AI training and inference deployments, recently traded at $223.67, down 0.73% on the day, with a market capitalization of approximately $5.56 trillion. Continued spending on accelerator deployments ripples directly into demand for the memory and storage that surround those chips in modern AI systems.

While both Micron and Sandisk are riding favorable supply-demand conditions, their exposure differs: Micron spans both HBM and conventional server DRAM, while Sandisk’s fortunes are tied more narrowly to NAND and enterprise flash demand. Pricing power in both segments will depend on how quickly new wafer capacity comes online relative to AI infrastructure spending.

What to watch

• Upcoming quarterly earnings from Micron and Sandisk for updated commentary on HBM and NAND pricing trends
• Hyperscaler capital expenditure plans, which drive enterprise flash and server DRAM orders
• NVIDIA’s product roadmap and data center shipments, a key demand indicator for adjacent memory suppliers
• Announcements regarding new fab capacity or HBM line conversions, which could ease supply constraints

Source: original release