• Nebius has been featured in the keynote by NVIDIA’s CEO Jensen Huang multiple times: as a conference sponsor, a European AI cloud provider, a cloud with infrastructure in the UK and an NVIDIA DGX Cloud Lepton provider.

• NVIDIA announced the expansion of NVIDIA DGX Cloud Lepton — an AI platform featuring a global compute marketplace for developers building agentic and physical AI apps, with GPUs now available from us.

• ‘Really, really great team. Really like the people, ’ is how NVIDIA’s CEO Jensen Huang sees Nebius. It was great having him at our booth at NVIDIA GTC Paris. Watch the short video and explore other content from the event, including customer interviews, insights from our leaders and even our own podcast with Sifted.

• We announced various integrations with NVIDIA’s accelerated computing platform, including expansion of our cloud with the NVIDIA AI Enterprise software stack. It’s a robust suite of tools to accelerate AI development and deployment, including NVIDIA NIM microservices, NVIDIA NeMo and other foundational components.

SWE-rebench dataset: more than 21,000 verifiable interactive tasksSWE-rebench dataset: more than 21,000 verifiable interactive tasks

Our AI R&D team announced the open-source release of the SWE-rebench dataset of more than 21,000 real-world, interactive software engineering tasks. For a detailed methodology and technical report, please see the accompanying paper on arXiv.

Our ISEG2 system is #13 on the Top500 of the world’s fastest supercomputersOur ISEG2 system is #13 on the Top500 of the world’s fastest supercomputers

That makes ISEG2 the most powerful commercially available system in Europe — and the second most powerful commercially available supercomputer anywhere in the world. It achieves this at the outstanding energy efficiency of 38.189 GFLOPs/watts, earning it #67 on the Green500.

AI Discovery Award: the winners are in!AI Discovery Award: the winners are in!

In June, we held the ceremony for the AI Discovery Award, our annual award recognizing startups leveraging AI in healthcare and life sciences. We brought the participants together in Amsterdam for this special occasion. 257 applications, 103 reached the semifinals, 28 of those received honorable mentions, and 12 became winners across four categories.

New products, features and tech docs tutorialsNew products, features and tech docs tutorials

• Managed Soperator is our fully managed Slurm-onKubernetes solution. It’s now available for everyone in selfservice, which means you can get a ready-to-work Slurm training cluster literally in minutes.

• It was a thrilling moment for us to announce a partnership and a joint GPU-as-a-Service (GPUaaS) solution with WEKA, integrating its advanced high-performance data storage software with our full-stack Nebius AI Cloud.

• Compute adds new deep-dive content for GPU workloads. Learn about the network topology of GPU clusters and how to manage it. Plus, new docs on boot disk images, troubleshooting initialization errors and accessing VM serial logs can help you get your VMs back on track faster.

• The Nebius AI Studio docs now cover how to generate consistent image styles and characters using LoRA. A great place to start if you want more consistency and control over your image generation workflow.

• Slurm and Soperator docs keep growing. You can now track metrics and monitor or analyze logs for both Pro Solution and self-deployed clusters. New how-tos include deploying clusters, creating and deleting Managed Soperator clusters and stopping and starting them.

• IAM updates: learn how to rename tenants, manage static access keys and work with custom groups.

• In Observability, you can now ingest logs from Managed Kubernetes into Logging using the Nebius O11y Agent.