Large firms prioritise storage scale & reliability in AI

WD has published survey findings showing that large organisations are prioritising storage scale, cost control and reliability in AI infrastructure planning. The results are based on responses from customers in hyperscale cloud, cloud service provider and enterprise markets.

The survey found that 87% of respondents prioritise capacity expansion and total cost of ownership optimisation when planning AI infrastructure. It also found that 74% cited total cost of ownership, capacity and scalability as the main advantages of hard disk drive-based infrastructure, while 70% said their storage environments remain majority HDD.

The figures suggest that, for many large organisations, AI infrastructure is becoming as much a data retention and storage management issue as a computing one. As businesses move from pilot projects to production systems, the need to retain training datasets, inference logs, embeddings and outputs is shaping long-term storage design.

Among respondents, 66% said they had deprioritised or were considering deprioritising newer technologies in favour of infrastructure with consistent reliability and predictable performance at scale. Reliability and availability ranked alongside support for AI training and inference workloads as the top infrastructure priorities, both at 69%.

Latency ranked much lower, with 7% of respondents placing it ahead of scalability, reliability and operational efficiency. This suggests many organisations are giving greater weight to sustained data movement and retention than to the fastest possible response times across all workloads.

Storage mix

The survey covered 200 of WD's top global customers across hyperscale cloud, cloud service provider and enterprise segments. Of those, 80 respondents represented organisations responsible for enterprise infrastructure strategy, data centre operations and storage architecture, while individual questions drew different response totals.

One finding highlighted the continuing role of hard drives in large data estates. Among respondents with visibility into their storage mix, 35% said HDDs accounted for more than 75% of total storage capacity.

That reflects the economics of storing growing volumes of AI-related data over long periods, especially in environments expected to reach very large scale. WD described this as part of a broader shift towards tiered storage architectures, in which businesses balance faster media for performance-sensitive tasks with lower-cost storage for bulk retention.

"HDDs remain part of our long-term strategy because they deliver reliable, scalable storage at a lower cost, making them ideal for large data volumes and long-term retention," said Abish Mohamed, Amstergi Middle East.

The survey also included anonymous responses on the role of hard drives in infrastructure planning. One respondent said HDDs should be seen as part of a combined approach rather than in opposition to solid-state drives.

"HDD is not a legacy product, but it is a strategic capacity solution. HDD is also ideal for data growth and deliver the lowest cost per TB in the market. The future is not HDD vs SSD, but HDD and SSD," the respondent said.

AI economics

WD used the survey to argue that AI infrastructure choices are being shaped as much by operational economics as by computing demand. Compute resources can be reused across training and inference cycles, but the data generated by those systems continues to accumulate over time.

That changes the financial profile of AI deployments. Rather than focusing only on peak processing needs, organisations are increasingly assessing how to retain and move larger volumes of data over the longer term without sharply increasing overall costs.

Ahmed Shihab, chief product officer at WD, said the findings reflect a shift in how customers are thinking about AI systems. "AI is fundamentally a data systems challenge, not just a compute challenge. Our customers are on the front lines of solving it, and their needs directly shape our innovation roadmap and the technologies we build for the AI era and beyond."

He added that storage planning would become more central as AI use matures. "While compute is reused, data persists-and grows. The organisations that win in the next phase of AI will be the ones that build infrastructure designed for continuous data systems at scale, not just peak compute performance."

Another anonymous respondent made a similar point about the economics of long-term storage. "HDDs stay in a long-term strategy because they solve a problem that newer technologies still don't beat on economics and scale. In simple terms: they're the cheapest, most reliable way to store huge amounts of data for a long time. Even as SSDs dominate performance-critical workloads, HDDs remain unmatched for bulk storage."