Toshiba is doubling AI data-center HDD capacity in the Philippines by FY2027 — their first major HDD investment in roughly five years.
The commitment? Around ¥60 billion. New nearline production lines are being installed, with a roadmap targeting 65TB drives by ~2030 and 100TB further out.
The thesis is simple: HBM and SSDs can't economically store every byte AI generates. As model training scales and inference workloads multiply, cold storage demand becomes structural — not cyclical.
This isn't about cutting-edge performance. It's about cost per terabyte at hyperscale. And in that game, spinning rust still wins for archival and tiered storage.
Watch the AI infrastructure trade beyond chips. Storage, power, and data-center build-out are the next layer.
The commitment? Around ¥60 billion. New nearline production lines are being installed, with a roadmap targeting 65TB drives by ~2030 and 100TB further out.
The thesis is simple: HBM and SSDs can't economically store every byte AI generates. As model training scales and inference workloads multiply, cold storage demand becomes structural — not cyclical.
This isn't about cutting-edge performance. It's about cost per terabyte at hyperscale. And in that game, spinning rust still wins for archival and tiered storage.
Watch the AI infrastructure trade beyond chips. Storage, power, and data-center build-out are the next layer.