A new global survey by Western Digital highlights a major shift in enterprise AI infrastructure planning, with organisations increasingly focusing on long-term data scalability, operational economics, and infrastructure reliability as AI deployments move from experimentation to production scale.
The findings suggest that AI infrastructure is evolving beyond a pure compute challenge into a continuous data systems challenge, where the persistent growth of datasets, inference logs, embeddings, and AI-generated outputs is driving demand for scalable storage architectures.
According to the survey, enterprises are increasingly favouring proven, operationally stable infrastructure over emerging but untested technologies. Around 66% of respondents indicated they are deprioritising newer infrastructure technologies in favour of systems that deliver predictable performance and reliability at scale.
The study also revealed that reliability and support for AI workloads are now equally critical priorities, with 69% of respondents emphasising both AI training/inference support and infrastructure availability. Notably, latency optimisation ranked significantly lower than scalability and operational efficiency, reflecting a growing preference for throughput-orientated infrastructure designed for sustained data movement and retention.
A major insight from the report is the continued strategic importance of HDD-based storage architectures within AI ecosystems. Despite the rise of SSDs for high-performance workloads, enterprises continue to rely heavily on HDD infrastructure for large-scale data retention and cost optimisation. Around 70% of respondents reported operating HDD-majority environments, while 35% indicated HDDs accounted for more than 75% of their total storage capacity.
This reflects the emergence of tiered AI storage architectures, where SSDs handle performance-intensive compute tasks while HDDs provide economically viable, large-capacity storage for long-term AI data accumulation. The report underscores that storage infrastructure decisions are increasingly driven by total cost of ownership (TCO), scalability, and lifecycle economics rather than peak performance alone.
The survey also found that 87% of enterprises prioritise capacity expansion and TCO optimisation in infrastructure planning, highlighting the growing operational challenge of managing continuously expanding AI datasets across cloud, enterprise, and hyperscale environments.
According to Ahmed Shihab, organisations entering the next phase of AI adoption will need infrastructure designed around continuous data growth and persistent storage requirements, rather than short-term compute peaks.
The findings reflect a broader industry transformation where AI infrastructure is increasingly being architected as a long-lived, data-centric operational environment, combining compute, storage, governance, and scalability into integrated systems capable of supporting exabyte-scale AI ecosystems.