SKHY
STX
Not every AI upgrade is a new chip
Nvidia gets most of the AI glamour shots, but Seagate is making the case that the real bottleneck might be what happens after the model is built. In a new white paper with SK hynix, the company says tiered storage can help inference and agentic AI workloads reuse context instead of constantly recomputing it.
That matters because the whole AI stack is basically a very expensive game of musical chairs. If data can move through memory, SSDs, and hard drives more efficiently, GPUs spend less time doing repetitive work and more time on the stuff that actually creates revenue. Seagate’s pitch is simple: storage doesn’t just sit there looking pretty in the server rack — it can help AI infrastructure do more with the same hardware.
The hidden AI tax
CEO Dave Mosley said on the company’s fiscal fourth-quarter earnings call that KV cache is central to the idea. That cache keeps context handy so models don’t need to regenerate it every time, which means less wasted compute and, in Seagate’s telling, more demand for hard drive storage.
That’s a neat twist for investors:
- AI growth still helps Nvidia, obviously.
- But it also creates a second-order beneficiary in storage.
- And if cloud customers are extending supply commitments into 2029 and beyond, this isn’t just a one-quarter story.
Why this matters for your portfolio
Seagate says cloud data centers now make up roughly 90% of its exabyte shipments, which is a fancy way of saying the big AI buyers are already here and they’re hungry. If the industry keeps leaning into longer-context AI and agentic workflows, storage could become a bigger piece of the infrastructure spend than a lot of people expected.
Big picture: the AI boom may still be a Nvidia headline, but Seagate is reminding everyone that the supporting cast can get paid too.