- Published on
AI Inference Ignites Global NAND Demand: SanDisk Says Customers Underestimate Flash Capacity Needs, Long-Term Orders ...
- Authors
- Name
- DeepHorus
SanDisk said on its fiscal 2027 first-quarter earnings conference call that AI inference is reshaping the supply and demand landscape of the NAND flash market, and that customers generally underestimate the NAND capacity needed to run AI infrastructure. The company said customers are aggressively adding long-term orders, and its new business model (NBM) order pipeline has extended to 2030 and even further. This statement pushed the storage demand changes brought by AI inference into the market spotlight.
NBM is based on committed purchase volumes, contractual constraints, and pricing mechanisms. SanDisk disclosed that about 50% of bit shipments in fiscal 2027 have already been locked in by NBM agreements, and that this proportion is expected to rise to about two-thirds in fiscal 2028. By the end of this decade, NBM reserved demand may exceed the company's available supply.
On the technology side, SanDisk announced that its first high-bandwidth flash (HBF) memory die has completed tape-out and entered the manufacturing stage. HBF targets AI inference scenarios, with the goal of providing read bandwidth close to HBM while also offering NAND-level large capacity. The product can support larger models, longer context, and continuously growing KV caches, and the first samples are planned for delivery in the next calendar year.
SanDisk also pointed out that the market is exploring the use of SLC NAND to replace some high-cost DRAM. In AI inference, flash memory can play the role of a capacity tier, helping systems retain broader context and reduce expensive repeated computation. This direction is seen as an important signal of NAND expanding its applications in AI infrastructure.
The earnings report showed that SanDisk's quarterly revenue was $11.03 billion, and non-GAAP earnings per share were $49.51, both above market expectations. The company also announced a 1-for-8 stock split, which will take effect on November 19. The better-than-expected results and guidance did not fully eliminate the market's profit-taking sentiment.
During the regular trading session on October 29, SanDisk closed at $1,330.05, down $74.72 for the day, a decline of 5.32%. After the earnings release, the stock price rose at one point but then fell back. As of the evening of October 29, U.S. Eastern Time, the after-hours price was approximately $1,282.50 to $1,295.05, down about 2.6% to 3.6% from that day's closing price, and down about 7.8% cumulatively from the previous closing price.
Looking ahead, NBM long-term orders and the advancement of the HBF product will enhance SanDisk's visibility in AI storage demand. However, whether supply can keep up, the pace of customers' long-term purchasing, and the capital market's repricing of performance still require continued observation. The pull from AI inference on storage capacity and bandwidth may become an important variable for the NAND market in the next stage.
