Nebius AI Cloud 3.6 Enhances Developer Experience and Governance for Production AI

Nebius today unveiled Nebius AI Cloud 3.6, also known as "Aether," a major update designed to bolster the developer experience, strengthen security measures, and improve governance capabilities for teams operating AI workloads in production. The release focuses on providing a more streamlined and efficient platform for the entire AI lifecycle. A standout new feature in AI Cloud 3.6 is Nebius Echo, an innovative AI agent seamlessly integrated into the Nebius AI Cloud web console. This agent empowers users to manage their cloud infrastructure through natural language commands, significantly simplifying complex operational tasks and enhancing user interaction. The update also brings more refined controls for Soperator, Nebius's managed Slurm-on-Kubernetes service, and for its Managed Kubernetes offering. A crucial enhancement to Managed Kubernetes is the introduction of the "Bring Your Own Image" (BYOI) feature. This allows users to deploy their own hardened or compliance-validated base images, providing greater flexibility and meeting stringent security and regulatory requirements. Furthermore, Nebius AI Cloud 3.6 now includes comprehensive support for Kubernetes v1.34, ensuring that users can leverage the latest advancements and features in container orchestration. This keeps the platform current with industry standards and best practices. Addressing critical performance and cost challenges inherent in large-scale AI workloads, the release incorporates local SSD disks directly on GPU servers. This strategic addition is engineered to eliminate I/O bottlenecks that can occur during intensive model training and inference processes, thereby accelerating AI development cycles. In terms of storage optimization, a new Intelligent Object Storage Class has been introduced. This class automatically migrates archived data to a lower-cost tier, effectively reducing storage expenses without imposing additional request or egress fees. Other storage improvements include a 30% increase in read bandwidth for single-thread Object Storage connections and up to a 100x improvement in IOPS for metadata-heavy workloads on Shared Filesystem, with validated filesystem clusters now supporting up to 100 PB. For enhanced financial governance, a new Budgets feature allows FinOps teams to establish specific spending targets and configure threshold-based alerts across various projects. This provides greater control and visibility over cloud expenditures, helping organizations manage costs more effectively. These extensive updates collectively underscore Nebius's dedication to delivering a robust, efficient, and user-friendly full-stack platform that supports every stage of the AI lifecycle, from initial data processing and model training to final production deployment.