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.
Definition
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.
Recent developments
Nebius AI Cloud 3.5 "Aether 3.5" launched March 2026. Added NVIDIA RTX PRO 6000 Blackwell Server Edition GPU + the Nebius Data Transfer Service for moving and replicating datasets across S3-compatible storage and across Nebius regions. Per Nebius — Aether 3.5 launch.
Multi-generation NVIDIA partnership: Rubin, Vera CPUs, BlueField storage. Early-access deployment of multiple generations of NVIDIA infrastructure including the Rubin platform, Vera CPUs, and BlueField storage systems. Per NVIDIA + Nebius partnership announcement.
S3-compatible storage engineered for GPU throughput. Storage tier specifically designed to feed datasets to GPU clusters at maximum speed; high-speed shared storage for writing and reading checkpoints during multi-host training. Per Nebius services/storage.
Serverless AI introduced March 2026. Nebius AI Cloud 3.5 introduces serverless AI to give developers frictionless compute for real-world AI workloads, eliminating provisioning overhead for inference deployments. Per Nebius newsroom — serverless AI.
Production-grade AI cloud profile. Datacenters.com lists Nebius as a provider of AI Cloud, GPU Infrastructure, Storage & Managed AI Services with full enterprise-grade SLA structure. Per Datacenters.com — Nebius.
Nebius AI Cloud shipped "Aether 3.6," succeeding the 3.5 release already tracked here. New pieces include Nebius Echo (an AI agent built into the console), Nebius Agents Blueprint, and a Data Lab feature inside Token Factory — announced alongside the ~$27B long-term Meta infrastructure supply agreement. Per nebius.com (2026-06-09).
Object storage is priced at roughly $0.025/GB/month. GPU on-demand rates run A100 $2.10-2.50/hr, H100 $2.95/hr, H200 $3.50/hr, B200 $5.50/hr, with $0.10/GB outbound data — giving the storage tier a concrete price point alongside the compute pricing. Per Nebius Review 2026: Pricing, Performance, Pros & Cons.
Nebius reportedly retains request data for up to 30 days in some configurations, unlike zero-retention providers. Worth flagging for compliance-sensitive workloads evaluating the storage/inference stack against EU alternatives like Scaleway. Per EU LLM APIs Compared 2026: Mistral, Scaleway, Nebius & JuiceFactory.
Q1 2026: revenue up 684% year-over-year, plus an acquisition spree — Tavily, Eigen AI, and Clarifai. The Q1 shareholder letter reports revenue of $399M (vs $50.9M a year earlier), NVIDIA Exemplar Cloud status on GB300 NVL72, expansion toward 800MW–1GW of connected power by year-end, and a new Pennsylvania site with up to 1.2 GW. The three acquisitions fold agentic search (Tavily, ~$275M), inference optimization (Eigen AI, ~$643M), and computer-vision tooling (Clarifai) into the platform. Per Nebius Group Letter to Shareholders Q1 2026 (Nebius).
The hyperscaler backlog now anchors the balance sheet: Microsoft (up to $19.4B, Sept 2025), Meta (up to ~$27B, March 2026), and a $2B NVIDIA investment. To fund the buildout, Nebius priced an upsized $4.0B private offering of convertible senior notes (1.250% due 2031 and 2.625% due 2033) in March 2026, with proceeds earmarked for data-center construction and hardware. Per Nebius convertible notes pricing announcement (Nebius) and Nebius Group Q1 2026 coverage (Pomegra). Sources: nebius.com · Nebius services/storage · Nebius docs — Object Storage · Aether 3.5 launch · NVIDIA + Nebius partnership