Western Digital said in a recent analysis that the buildout of AI data centers is shifting away from a narrow focus on GPU capacity and toward competition over storage capability. The company argued that storage planning is now a core part of AI infrastructure as data generated across the AI lifecycle keeps accumulating rather than disappearing after compute jobs end.
Citing an IDC forecast, the analysis said annual global data creation will reach 718 zettabytes by 2030. Western Digital said AI systems continuously produce and retain training data, model checkpoints, embedding vectors, inference logs, prompts, outputs, and evaluation datasets. Those records later become important assets for model iteration, quality assessment, and compliance audits.
The company also warned that many infrastructure plans still pay too much attention to GPU utilization while overlooking long-term data accumulation. As datasets move into the petabyte and even exabyte range, a single storage architecture becomes harder to sustain. Western Digital said enterprises should adopt tiered storage strategies, using high-performance flash for training and real-time inference, while relying on high-capacity HDDs and object storage for long-term retention, historical archives, and low-frequency access. It added that future AI infrastructure competition will be judged not only by GPU counts, but also by storage cost per PB, power consumption, recovery efficiency, and lifecycle data management.
BlockBeats reported on Aug. 15 that Western Digital, in a recent analysis, said the rapid expansion of artificial intelligence applications is pushing AI data center construction beyond a simple race for GPU compute and into a contest over data storage capacity. The company said storage planning has become a core part of AI infrastructure.
Citing an IDC forecast, the analysis said global annual data creation will reach 718ZB by 2030. Western Digital said data produced by AI systems does not disappear once a computing task ends. Training datasets, model checkpoints, embedding vectors, inference logs, prompts, outputs, and evaluation data continue to accumulate over time.
Storage is becoming central to AI infrastructure planning
Western Digital said many current AI infrastructure plans pay too much attention to GPU utilization while overlooking the buildup of data across the AI lifecycle. Data generated during training and inference becomes an important asset for later model iteration, quality evaluation, and compliance audits. Storage costs, the company said, will directly shape the long-term operating efficiency of AI systems.
As data volumes move into the PB and even EB range, a single storage architecture is unlikely to meet demand. The analysis said enterprises need a tiered storage strategy, using high-performance flash for training and real-time inference, while high-capacity HDDs and object storage are better suited for long-term retention, historical records, and low-frequency access scenarios.
Key metrics extend beyond GPU counts
According to the analysis, the main benchmarks in future AI infrastructure competition will not be limited to the number of GPUs. Storage cost per PB, power consumption, recovery efficiency, and data lifecycle management will also matter.
Western Digital added that companies that still treat storage as a secondary step after compute may face uncontrolled data costs and slower model iteration efficiency.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
