Cloud operators have already claimed most of its nearline capacity through 2028
The AI infrastructure boom isn't just eating all the memory – it's also driving demand for storage, giving high-capacity hard drives a new lease of life, according to Seagate.The storage veteran says cloud operators are buying ever more mass-capacity hardware as AI applications generate and retain growing volumes of data."Datacenter demand now represents approximately 90 percent of our exabyte shipments," chairman and CEO Dave Mosley told analysts during a call to discuss Seagate's Q4 and full-year results.
"Based on the long-term supply agreements in place today, the vast majority of our nearline exabytes are now allocated into calendar 2028," he said.
Back in February, Seagate warned that its production capacity of nearline drives was fully allocated for this year, and it had started accepting orders for 2027."Importantly, we are not seeing customers pull back on planning horizons," Mosley added, suggesting hyperscalers remain keen to reserve hard drive capacity years in advance.AI agents that carry out tasks with varying degrees of autonomy are the latest source of demand for nearline drives – at least according to Seagate's CEO."With the transition from AI model training to inference to agentic applications, more data is generated and retained for historical context, compliance and future reuse. As these datacenter environments become larger and more complex, customers must balance performance, energy consumption and cost."Cloud operators already address these challenges through tiered storage combining high-performance memory and SSDs with mass capacity hard drives to optimize performance and economics at scale, he added."Our recent white paper with SK hynix illustrates the importance of tiered storage for inference and agentic AI workloads, which show a direct benefit to hard drive storage. These workloads rely on persistent context across user interactions, and key-value or KV cache is used to retain and reuse that context efficiently."By distributing KV cache data across memory, SSD, and hard drive tiers, organizations can retain more context and avoid recomputing previously generated data.
