AMSTERDAM–(BUSINESS WIRE)–Nebius today unveiled Nebius AI Cloud 3.5, adding significant new capabilities to its full-stack cloud platform that reduce operational friction and enable AI builders to prototype, test, and ship products faster.
The introduction of serverless features gives developers the ability to launch workloads almost instantly, eliminating the need for AI teams to spend significant time configuring infrastructure before they can run experiments, train or serve models in production. Infrastructure configuration and runtime management are handled by the Nebius platform, enabling developers to focus on building applications instead of managing environments.
Alongside serverless capabilities, Nebius is expanding its GPU offering with NVIDIA RTX PRO 6000 Blackwell Server Edition for a range of workloads including AI inference, industrial robotics, physical AI simulations, visual computing, and drug discovery.
Version 3.5 of Nebius AI Cloud “Aether” also introduces Nebius’s Data Transfer Service, which reduces data management overhead for teams working across environments by simplifying data migration and replication between external S3-compatible storage systems and Nebius cloud regions.
Configuration setup for Managed Soperator, Nebius’s fully managed Slurm-on-Kubernetes solution, has also been overhauled to offer more options and granularity when creating a Slurm cluster for self-service users. Managed Kubernetes observability has also been updated to give teams additional cluster-level control.
The AI application marketplace has also been redesigned to help users access faster tools, models and applications required in their workflows.
Other updates in Aether 3.5 include improved user administration and role-based permissions, making it easier for organizations to manage access across teams. New public APIs for billing data streamline the export process for finance and operations teams.
All the new features that the Aether 3.5 release delivers are available now on the global Nebius AI Cloud infrastructure, with the serverless service available in public preview. NVIDIA RTX PRO 6000 Blackwell Server Edition is available today.
Nebius AI Cloud Aether 3.5 — at a glance
Serverless AI
- Elastic, pay-as-you-go compute accelerated by NVIDIA
- Simplified access to AI workloads without managing infrastructure
- Designed for prototyping, experimentation, and model inference evaluation
NVIDIA RTX PRO 6000 Blackwell Server Edition
- GPU option designed for a range of workloads including AI inference, industrial robotics, physical AI simulations, visual computing, engineering research, and drug discovery
- Enables cost-efficient AI inference and simulation-heavy workloads
Data Transfer Service
- User-friendly tool for data transfer and replication across Nebius regions and S3-compatible object storage services
Managed Soperator
- An updated cluster configuration wizard for Nebius’ fully managed Slurm-on-Kubernetes solution
Platform enhancements
- Updated navigation for the AI/ML application marketplace
- Improved disk encryption, boot image management, and Kubernetes-level observability
- Expanded controls for user administration and role-based permissions
- Public API for exporting billing data in standardized formats
Additional resources
- Blog post from our Product Management team
- Blog post on NVIDIA RTX PRO
- Blog post on our new serverless AI
- Webinar and live Q&A registration
About Nebius
Nebius, the AI cloud company, is building the full-stack platform for developers and companies to take charge of their AI future — from data and model training to production deployment. Founded on deep in-house technological expertise and operating at scale with a rapidly expanding global footprint, Nebius serves startups and enterprises building AI products, agents and services worldwide.