AI is a data system: Western Digital's Irving Tan on the storage supercycle

Key points

Key takeaways from a theCUBE interview at the CASPA CEO Summit 2026 with Irving Tan, CEO of Western Digital (October 2026):

AI as a data system. Tan's core argument: AI is too often framed as compute and memory, but intelligence only persists if the data behind it is stored and tiered well — AI must be treated as a full data system to deliver its ROI.

Data tidal wave. Citing IDC, he says data generation will triple to over 700 zettabytes within three years, with associated storage needs doubling to more than 26 zettabytes.

Hot, warm, cold. Data constantly moves within the data centre: metadata lives on hard drives, training and inference workloads cache into flash for IOPS, then generated data flows back onto hard drives — tiering persists but the mix shifts by workload.

Physical AI is next. Humanoid robotics, autonomous vehicles and video-heavy workloads bring a new wave of data, much of it under regulatory retention requirements — reviving demand for archival tiers including tape.

Throughput innovation. Western Digital is raising hard-drive bandwidth and throughput from 2x up to 8x with doubled IOPS, alongside higher-capacity drives for better rack density, TCO and power.

Demand outlook. Storage demand is growing at least 25%+ CAGR over the next five years; the company is investing in capacity through technology rather than simply building more drives.

System-level design. With hyperscalers and AI labs as customers, WD optimises at device, software and data-centre-system level, co-designing with the ecosystem for power and latency.

Directive. "Innovate, move with speed and agility" — Tan says the edge, including on-device AI, is on the cusp of arriving.

Read more: theCUBE on YouTube