AI data growth outpaces compute planning, making storage the new infrastructure bottleneck

Huo Xing Cai Jing reports that on August 15, Western Digital’s latest analysis indicates that as the scale of AI applications expands rapidly, AI data center construction is shifting from pure GPU compute competition to competition in data storage capacity, making storage planning a core component of AI infrastructure. The article cites IDC’s forecast that global annual new data generation will reach 718 ZB by 2030. Data generated by AI systems does not disappear after computational tasks end—training data, model checkpoints, embedding vectors, inference logs, prompts, output results, and evaluation data continue to accumulate. Western Digital notes that many current AI infrastructure plans overly focus on GPU utilization while neglecting data accumulation throughout the AI lifecycle. Data generated during training and inference will become critical assets for model iteration, quality assessment, and compliance auditing, with storage costs directly impacting the long-term operational efficiency of AI systems. As data scales enter the petabyte and even exabyte ranges, a single storage architecture can no longer meet demands. Enterprises must adopt tiered storage strategies, using high-performance flash for training and real-time inference, and high-capacity HDDs and object storage for long-term data retention, historical records, and low-frequency access scenarios. The analysis suggests that future key metrics for AI infrastructure competition will no longer be limited to GPU count, but will include cost per petabyte of storage, energy consumption, recovery efficiency, and data lifecycle management capabilities. Companies that continue to treat storage as a secondary afterthought to computation risk uncontrolled data costs and reduced model iteration efficiency.