Over the past year, data storage companies have moved sharply higher. Western Digital, Seagate, and SanDisk have all benefited from a market narrative that is easy to understand: if artificial intelligence is going to require an extraordinary buildout of compute infrastructure, it is also going to require an extraordinary buildout of storage infrastructure.
That thesis strikes me as directionally correct. AI models do not merely require GPUs. They require vast repositories of training data, intermediate outputs, synthetic data, user interactions, logs, backups, compliance archives, and low-cost cold storage. Inference at scale will only add to the data exhaust. The more useful AI becomes, the more data it will generate, store, retrieve, and preserve.
This thesis has been part of the Situational Awareness theme championed by Leopold Aschenbrenner, whose investment fund has profited handsomely from investments in SanDisk and other players. (Read our related article, along with selected undervalued investment ideas that are already seeing an acceleration in demand.)
As often happens, however, the market has first rewarded the most visible beneficiaries. The device makers are obvious. Their products sit directly in the path of demand. If hyperscalers need more nearline storage, the first order of analysis points to the companies that make hard disk drives and flash storage products.
The more interesting question, at least for value investors, is what lies beneath them.
If Western Digital, Seagate, SanDisk, Kioxia, Micron, Samsung, SK Hynix, and the tape vendors are the visible storage layer, then there is a less visible materials layer below them. Modern storage devices depend on specialty glass, cobalt, platinum, ruthenium, helium, tantalum, silicon, copper, rare earth magnets, and other materials that are produced by mining, refining, and specialty-materials companies around the world.
Some of those companies remain available at value prices.
This is not a call to buy a basket of commodity stocks simply because AI exists. The purpose of the exercise is more specific: to identify small-cap public companies whose products connect to the data-storage supply chain, whose valuations remain modest, and whose upside could be enhanced if storage demand becomes a larger and more strategic driver of incremental demand.
In other words, the storage story may already be in the stocks of the device makers. It may not yet be fully reflected in selected upstream suppliers.
