Intel and Supermicro develop hybrid storage-as-a-service architectures for AI workloads

The shift toward hybrid storage-as-a-service architectures addresses the technical bottleneck of moving massive AI datasets between disparate environments. By standardizing on Redfish and BMC protocols, infrastructure providers are reducing the friction of managing high-performance compute at the edge while maintaining cloud-like consumption models. For the streaming ecosystem, this evolution suggests a move away from pure-cloud reliance toward a more distributed, high-bandwidth infrastructure capable of handling real-time analytics and metadata processing. As AI workloads grow, the industry should monitor the adoption rates of Scality ADI and similar purpose-built frameworks to see if they become the standard for low-latency data management.

Scality has been aggressively expanding its footprint in AI-optimized storage infrastructure throughout 2026. The company's Artesca platform, designed for object storage at scale, has been positioned as a foundational layer for AI data pipelines that span edge and core environments. Scality's ADI (Artificial Data Intelligence) framework was specifically built to handle the metadata-intensive workloads that AI training and inference demand, providing a unified namespace across distributed storage tiers. This aligns with the broader industry movement toward disaggregated storage architectures where compute and capacity scale independently.

The competitive landscape for AI infrastructure storage is intensifying, with multiple vendors racing to capture enterprise budgets. Ericsson launched its AI in RAN commercial software subscription on June 11th, claiming up to 20% higher downlink throughput and up to 10% better spectral efficiency across more than 15 live deployments. While that announcement targets radio access networks rather than storage directly, it illustrates the same pattern: vendors are packaging AI-ready infrastructure as subscription services with measurable performance guarantees. Iron Mountain's participation in the Supermicro initiative signals that traditional data center operators are repositioning their physical assets as AI-ready infrastructure nodes, bridging the gap between legacy colocation and modern GPU-accelerated compute environments.

On the technical side, the use of Redfish and BMC protocols for hardware management represents a deliberate choice to maintain vendor neutrality in the storage stack. Ericsson's strategy emphasizes hosting AI inference inside the network itself, positioning infrastructure as part of the execution environment rather than merely carrying AI traffic. That same principle applies to storage: the goal is to make storage nodes active participants in AI workflows rather than passive repositories. Supermicro's Open Storage Summit, where this collaboration was detailed, has become a key venue for demonstrating how commodity hardware combined with software-defined storage layers can match or exceed proprietary appliance performance for AI workloads. The emphasis on open management interfaces like Redfish ensures that operators can integrate these systems into existing orchestration frameworks without vendor lock-in, a critical consideration for streaming platforms evaluating .