The Zhitong Finance App learned that Dutch artificial intelligence infrastructure provider Nebius (NBIS.US) has launched Nebius AI Cloud 3.1, which introduces the next-generation Nvidia (NVDA.US) Blackwell Ultra computing technology and enhanced operational capabilities into its latest full-stack artificial intelligence cloud platform.
The company said Nvidia's HGX B300 and GB300 NVL72 systems (Nvidia Quantum-X800 InfiniBand with 800 Gbps) are capable of delivering next-generation performance for large-scale artificial intelligence training and inference.
Nebius indicated that the company is deploying Nvidia Blackwell Ultra infrastructure globally and currently has Nvidia GB300 NVL72 systems and HGX B300 systems for customers to use. Nebius also said the company is now the first cloud service provider in Europe to run both platforms simultaneously in a production environment.
Nebius' AI cloud platform is also the first company in the world to use 800 Gbps Nvidia Quantum-X800 InfiniBand interconnect technology to run GB300NVL72 systems — a move that triples the throughput of distributed workloads.
The company notes that to meet the enterprise's need for operational transparency, version 3.1 introduced Capacity Blocks (Capacity Blocks) and real-time capacity dashboards (Capacity Dashboard) to enable customers to understand their reserved GPU capacity and availability across all data center regions.
Nebius added that version 3.1 was developed on the basis of the Nebius AI cloud “Aether”, adding transparent capacity management capabilities and expanded infrastructure to provide customers with the required operational visibility and resource planning capabilities to help them expand AI applications in production environments. Additionally, Nebius said Aether 3.1 also provides enhanced AI/ML development tools, extended ecosystem integration, and fine enterprise-level governance controls, which accelerate value realization while reducing risk.
Nebius' competitor CoreWeave (CRWV.US) also provides Nvidia's Blackwell chips for artificial intelligence infrastructure. In July of this year, CoreWeave said it was the first cloud platform to fully launch the Nvidia RTX PRO 6000 Blackwell server instance. In addition to the RTX PRO 6000, CoreWeave also offers the broadest range of Nvidia Blackwell infrastructure products on the market, including the Nvidia GB200 NVL72 system and the Nvidia HGX B200 platform.