Scality Research Highlights Object Storage as Foundation for 91% of Private AI Deployments

A recent independent research study conducted by Freeform Dynamics, commissioned by Scality, a leader in cyber-resilient storage software, has revealed that object storage is a cornerstone technology for enterprises implementing private AI in production environments. The study, titled "Storage Infrastructure for Enterprise AI: Lessons From Seasoned Adopters on Building Scalable Sovereign Environments," found that a significant 91% of enterprises with private AI in production are meaningfully utilizing object storage. Furthermore, 44% use it extensively, and 47% use it quite a bit, positioning it ahead of file-based storage and significantly above block-based storage in these critical AI infrastructures. The research, announced on August 12, 2026, surveyed 504 enterprises and emphasizes that data and storage infrastructure are emerging as primary constraints as AI initiatives scale from experimentation to operational deployment. This report is highly significant for cloud and DevOps practitioners, particularly those involved in architecting and managing AI/ML workloads. It validates that object storage has matured beyond its traditional roles (e.g., cold storage, data lakes) to become a central, high-performance component in demanding AI pipelines. For organizations embracing private or sovereign AI—driven by needs for data control, compliance, and cost management—this means that the choice and configuration of their object storage solution directly impact the success and efficiency of their AI initiatives. IT leaders, infrastructure architects, and data engineers must now view object storage not merely as a repository but as an active, performance-critical layer that can either accelerate or bottleneck their AI models. The findings affect any enterprise moving AI from pilot to production, especially those with stringent data governance requirements. This trend aligns perfectly with the broader evolution of data infrastructure in the age of AI. As AI models become more sophisticated and data volumes explode, the demand for scalable, cost-effective, and highly available storage has intensified. Object storage, with its inherent scalability, metadata capabilities, and cost-efficiency for massive unstructured datasets, is uniquely positioned to meet these demands. The shift towards private AI, where enterprises seek to retain control over their data and models, further amplifies the need for robust on-premises or hybrid object storage solutions. This is not a new phenomenon; for years, the industry has been moving towards data-centric architectures where storage is a first-class citizen in the compute continuum. The integration of object storage with AI frameworks, data lakes, and data lakehouses has been a consistent theme, with providers continually enhancing performance, security, and management capabilities to support AI workloads. For instance, Google Cloud's recent "Cloud Storage Rapid" features aim to bridge the gap between object storage reliability and specialized AI storage performance, indicating a market-wide recognition of object storage's pivotal role in AI. Similarly, discussions at events like the Supermicro Open Storage Summit 2026 highlight how modern data lakes built on object storage infrastructure are now natively supporting AI workflows, including inference and context memory serving. The increasing focus on immutable object storage for ransomware protection also underscores its importance in securing the critical data assets that fuel AI. Practitioners should recognize that object storage is now a strategic investment for AI. This implies several practical considerations. Firstly, evaluate object storage solutions not just on capacity and cost, but critically on their performance characteristics for AI workloads, including throughput, latency, and I/O operations. Secondly, prioritize solutions that offer strong data governance, immutability, and lifecycle management features, as these are crucial for compliance and data integrity in AI pipelines. Thirdly, consider the integration capabilities of object storage with existing AI/ML platforms, data processing frameworks, and security tools. The research highlights that storage performance is as critical as compute or GPU availability in production AI environments, meaning bottlenecks at the storage layer can negate investments in high-end compute. Therefore, architects should focus on optimizing data access patterns and ensuring that their object storage infrastructure can keep pace with the demands of AI training, inference, and data preparation. Finally, for those building private AI, the ability to deploy and manage object storage on-premises or in hybrid configurations with cloud-like scalability and features will be paramount.