Backblaze supplies $335M HDD storage to CoreWeave AI cloud

Backblaze and CoreWeave have finalized a five-year storage agreement valued at $335 million, introducing large-scale HDD-based object storage tiers into CoreWeave’s AI cloud platform. The arrangement allows CoreWeave customers to access additional storage capacity without modifying their applications, addressing a key operational constraint in AI infrastructure: managing the vast volumes of data generated by training, inference, and retrieval-augmented generation workloads.

The partnership reflects a broader shift in AI infrastructure priorities. While GPU capacity has dominated industry discussions, storage architecture has emerged as a critical factor in cost control and performance optimization. CoreWeave’s integration of Backblaze’s HDD-based storage aims to reserve high-performance storage for latency-sensitive workloads while offloading less demanding data to more economical tiers.

Storage as a strategic layer

AI workloads generate diverse data types, each with distinct performance requirements. Model checkpoints, training datasets, prompt logs, and retrieval-augmented generation corpora exhibit varying access patterns and latency tolerances. Treating all data as premium storage is prohibitively expensive, while relegating everything to cold archive risks operational inefficiency. The Backblaze-CoreWeave agreement seeks to address this challenge by introducing tiered storage options within CoreWeave’s AI Object Storage platform, including environments leveraging CoreWeave’s LOTA distributed cache technology.

For CoreWeave, the deal expands its managed storage capabilities without disrupting existing customer workflows. The company serves nine of the top ten AI model providers, positioning it as a key infrastructure player in the AI ecosystem. Backblaze, which reports over 100,000 customers, brings expertise in large-scale, cost-efficient cloud storage, though the agreement’s success will depend on execution factors such as utilization rates, performance predictability, and data placement strategies.

AI infrastructure beyond GPUs

The deal underscores the maturation of AI infrastructure beyond initial GPU-centric investments. Early AI spending prioritized access to GPUs, power, and data center capacity, but operators now face challenges in storage placement, caching, networking, orchestration, and cost control. These factors are critical to transforming AI infrastructure from experimental projects into scalable, service-oriented businesses.

The multi-exabyte commitment raises operational questions. Deployment timelines, pricing adjustments in response to shifting demand, and regional data control requirements will influence the agreement’s long-term impact. Regulators and enterprise risk teams are increasingly scrutinizing data residency and storage practices, particularly for AI training data, generated outputs, and customer information.

For Backblaze, the partnership provides a high-profile entry into the AI infrastructure market, where storage has often been overshadowed by compute narratives. CoreWeave gains an additional layer of managed storage without imposing migration burdens on its customers. The next phase will focus on execution—filling the capacity, maintaining cost predictability, and ensuring storage operations remain seamless enough to avoid developer friction.