Nebius AI Cloud Aether 3.5 release for real world AI

Nebius announced Aether 3.5 for its AI Cloud platform. The update introduces serverless capabilities along with new GPU options and tools aimed at applied AI workloads in robotics, industrial systems, biotech pipelines and scientific computing. The serverless services include DevPods for development environments and early prototyping in private preview with general availability expected in three months. Jobs and Endpoints for batch workloads such as model training and inference are available in public preview through the web console and CLI. Nebius added the NVIDIA RTX PRO 6000 Blackwell Server Edition GPU with 96 GB of high-speed GDDR7 memory. The GPU targets cost-effective compute for physical simulations and drug discovery with initial access through account managers. A Data Transfer Service entered private preview. It allows users to move and replicate datasets between S3-compatible storage and across regions with general availability expected later next month. Managed Soperator received a configuration wizard and expanded options to define Node Sets with different GPU or CPU setups along with partitions and storage selections. The Applications section of the console was updated to improve discovery of curated images, frameworks and preconfigured environments including NVIDIA NIM Microservices models. Platform changes added support for Kubernetes 1.33. Encryption coverage expanded to additional boot disk configurations and Kubernetes Audit Logs became generally available with Node Events accessible under defined retention policies.