← All signals

markets

Sandisk Stock Gains as Analyst Cites AI-Driven Shift in NAND Demand Dynamics

Sandisk rose 6.8% to $1,887.04 after Rosenblatt began coverage with a buy rating and $2,400 target, arguing AI is turning NAND flash from a commodity into a critical component and that multiyear supply deals could smooth the memory cycle.

Julia Peters3 min read Reviewed by an editor
Image published by MarketWatch.com - Top Stories with the original report
MarketWatch.com - Top Stories

Shares of Sandisk (SNDK) rose 6.8% on Tuesday, Sept. 22, to close at $1,887.04, a day after Rosenblatt analyst Kevin Cassidy began covering the stock with a buy rating and a $2,400 price target. The stock had already soared this year on demand for memory and storage from artificial-intelligence systems.

Cassidy's argument goes beyond near-term demand. He contends that AI is changing what NAND flash memory, Sandisk's core product, is worth to the customers who buy it.

From commodity to critical component

NAND flash stores data without needing a constant supply of power. For most of its history it has been sold as a commodity. In personal computers, smartphones and other consumer electronics, Cassidy said, demand has been driven by two things: rising density, meaning how much data fits in a given physical area of the chip, and a falling cost per bit, the unit of data stored in a NAND cell. Buyers mostly wanted more storage for less money.

AI changes those priorities. Larger AI models and inference, the data-intensive process in which trained models generate answers and predictions, "increasingly prioritize density, performance, endurance and supply certainty over the lowest absolute price," Cassidy wrote in his Monday note. In his view, new AI systems "are creating an opportunity to reposition" NAND as "a more system-critical component of AI infrastructure."

If that holds, Sandisk would compete on performance and reliability of supply, not only on price per bit.

Sandisk's technology and revenue outlook

Cassidy pointed to Sandisk's technology roadmap as support for that shift. It includes the BiCS8 and BiCS10 architectures, developed with Japan's Kioxia, which stack memory cells vertically to fit more storage into a smaller space. He said the technology allows for higher density at favorable costs, which should "sustain a performance-and-cost advantage" in storage for large-scale AI, and lets NAND "increase its value to the overall" AI system.

He also put numbers on the opportunity. For Sandisk's fiscal 2027, which began in late June, he expects data-center revenue to reach $21.7 billion, rising to $28.6 billion in fiscal 2028.

Long-term supply agreements

The third part of Cassidy's case is a change in how Sandisk sells. The company has signed what it calls new business model agreements with eight major NAND customers. These multiyear deals let customers secure supply at set prices, and they have emerged as AI-driven NAND shortages have persisted.

Cassidy said the agreements will cover a large share of Sandisk's production capacity and "could increase demand visibility and reduce historical NAND-cycle volatility." The memory industry has long moved in boom-and-bust cycles, as producers add capacity during shortages and then cut prices when that capacity arrives. Contracts that lock in volume and price for several years would make Sandisk's revenue less dependent on where the cycle stands at any moment.

Data-center revenue and supply contracts

  • Data-center revenue in Sandisk's quarterly reports, measured against Cassidy's $21.7 billion estimate for fiscal 2027.
  • Whether more customers sign multiyear supply agreements beyond the eight so far, and how much of Sandisk's capacity they cover.
  • Signs of whether NAND shortages persist, since they have driven both pricing and the willingness of customers to sign long-term deals.
  • Progress on the BiCS10 architecture developed with Kioxia.

Source: marketwatch.com