Nebius AI Cloud

GPU-first AI cloud platform with S3-compatible object storage as one tier of a unified AI infrastructure stack (GPU droplets, managed Kubernetes, managed Postgres, block volumes, shared filesystem, container registry, serverless AI inference). The storage tier is specifically engineered to feed datasets to GPU clusters at maximum sustained throughput, using high-speed shared storage for multi-host training checkpoint writes/reads. Partnership with NVIDIA covers early access to Rubin, Vera CPUs, BlueField storage, and RTX PRO 6000 Blackwell Server Edition GPUs.

Generic clouds (AWS / GCP / Azure) treat GPU compute and object storage as separate procurement decisions, with separate billing dimensions and separate optimization targets. Nebius's bet: bundle the entire AI training/inference stack — GPU + S3-compatible storage + checkpoint-optimized shared filesystem — into one platform where the I/O paths are co-engineered for AI workloads. The 2026 framing (Nebius AI Cloud 3.5 "Aether 3.5") adds serverless AI inference + Data Transfer Service for cross-region replication.

Large-scale LLM training where checkpoint I/O dominates wall-clock cost, multi-host distributed training on NVIDIA Rubin / Blackwell GPUs, AI inference serving with frictionless serverless deployment, S3-compatible data staging for ML training corpora, and customers needing early access to next-generation NVIDIA hardware before it lands on the hyperscalers.