공시 • Jun 25
Nebius Unveils Nebius AI Cloud 3.6 With New Developer Experience, Security, Governance, And Storage Enhancements Nebius unveiled Nebius AI Cloud 3.6, adding new developer experience, security and governance improvements, as well as enhanced storage capabilities for teams running AI in production on Nebius’s full-stack AI cloud platform. Version 3.6 of Nebius AI Cloud “Aether” introduces Nebius Echo, an AI agent for controlling cloud infrastructure using natural language. Built directly into the Nebius AI Cloud web console, Echo can answer questions and execute core operations, with guardrails to prevent unintentional actions. Nebius Echo runs on open-source models hosted on Nebius Token Factory, Nebius’s production inference platform. The first-look version released covers core cloud services, with infrastructure debugging and multi-step infrastructure-as-code deployments among upcoming features on the roadmap. Aether 3.6 delivers a broader set of developer experience improvements targeting the time teams lose before the real work begins. This includes one-click access and deeper integration with SkyPilot, with no control plane to maintain; a new step-by-step instance creation workflow; global search across all services and tenants; a unified Notification Center; and more. Nebius is opening the Nebius Builder Program in early preview, offering developers and ML engineers credits across Nebius AI Cloud, Nebius Token Factory, and Nebius Academy, plus engineer office hours and a builder community. Members of the Nebius Builder Program get early access to the new Nebius Certifications program developed by Nebius Academy, starting with AI Cloud Ops (Associate), available now. Additional tracks for AI application builders, solutions architects, infrastructure engineers, and business leaders across the Nebius ecosystem will be introduced in the coming weeks. The new release of Nebius AI Cloud extends the security and governance controls that teams need to run sensitive workloads in regulated environments. A new Key Management Service (KMS) with customer-managed encryption keys (CMEK) and cryptographic erasure puts control of data encryption into customers’ hands, while Workload Identity Federation enables credential-free authentication across environment boundaries. Other enhancements include more granular controls for Soperator, Nebius's managed Slurm-on-Kubernetes service, and Managed Kubernetes, including a new “Bring Your Own Image” (BYOI) feature for hardened or compliance-validated base images. The release introduces a new Budgets feature, letting FinOps teams set spending targets and threshold-based alerts across projects. Nebius AI Cloud 3.6 delivers local SSDs on GPU servers for high-performance caching, removing I/O bottlenecks during training and inference. A new Intelligent Object Storage Class automatically moves archived data to a lower-cost tier, with no request or egress fees. Using the Enhanced class, Object Storage shows 30% more read bandwidth for single-threaded client connections. On the shared filesystem, 4K file operations deliver 3x more IOPS, metadata-heavy workloads deliver up to 100x more IOPS, and validated cluster sizes now reach 100 PB. Nebius AI Cloud Aether 3.6 — at a glance: Developer experience: Nebius Echo, an AI agent for natural-language infrastructure control; Managed Service for SkyPilot; Unified Notification Center; Global Search; New instance creation workflow; Updated pricing calculator. Security and governance: KMS with customer-managed encryption keys (CMEK); Workload Identity Federation; Soperator IAM-based access; Managed Kubernetes API access control; Budgets; Bring Your Own Image (BYOI); Datadog Log Management integration; Disk snapshots. Storage and other platform improvements: Intelligent Object Storage Class with automatic cost-tier migration; 30% read Object Storage bandwidth improvement for single-thread connections; Local SSD disks on GPU servers; 3x more IOPS for 4K files operations and up to 100x more IOPS for metadata-heavy workloads for Shared Filesystem; Validated filesystem clusters size up to 100 PB; Kubernetes v1.34 support. 공시 • Jun 09
Nebius Launches Physical AI Living Lab For UK And European Robotics Startups Built With NVIDIA Technologies Nebius announced the Physical AI Living Lab, a six-month program that equips British and European robotics startups with NVIDIA's physical AI development tools and Nebius's AI cloud infrastructure. Physical AI depends on large-scale simulation, synthetic data, and accelerated compute that most early-stage robotics companies cannot assemble on their own. The Physical AI Living Lab removes that barrier, putting the same class of tooling and compute used to build physical AI at scale into the hands of founders, so they can move from simulation to real-world deployment faster. The Lab builds on Nebius and NVIDIA's collaboration to create a cloud platform for robotics and physical AI. The companies intend to extend the Physical AI Living Lab to other regions over time and bring in further cohorts as the program grows. Participating startups in the Physical AI Living Lab will work hands-on using NVIDIA technologies to deploy their physical AI workloads — NVIDIA OSMO for workload orchestration, NVIDIA Cosmos world foundation models, and NVIDIA Isaac Sim and NVIDIA Isaac Lab for robot simulation and training — together with the NVIDIA Physical AI Data Factory Blueprint, all running on Nebius infrastructure. Synthetic data generation is provided through Voxel51's FiftyOne integration, built on Cosmos world foundation models. The first phase of the Physical AI Living Lab will run on Nebius's UK-based infrastructure, built on NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. Beyond the simulation and training tools, participants build, fine-tune and run their models on Nebius AI Cloud — the company's full-stack platform for taking AI from data and training through to production deployment. Applications run through the NVIDIA Inception pipeline. The first cohort will begin in September 2026. Engineers from both Nebius and NVIDIA will provide technical guidance throughout the program. 
