Nebius

Nebius is a cloud platform built specifically for AI workloads, covering the full lifecycle from model training to production inference. Its core offering is large-scale, non-virtualized NVIDIA GPU infrastructure connected with InfiniBand networking, which the company positions as suited to training foundation models and running AI agents at scale. On top of that infrastructure, Nebius offers managed and serverless inference through a service called Token Factory, letting teams run open-source large language models and image models via API without managing the underlying hardware themselves. The platform is aimed at AI developers and enterprises building production AI systems, with named use cases spanning healthcare, media, financial services, robotics, and e-commerce. Nebius reports operational reliability figures for its infrastructure, including a mean time between failures of 56.6 hours across a 3,000-GPU deployment and a 12-minute mean time to resolution, and backs enterprise customers with a team of AI experts, 24/7 support, and proof-of-concept assistance. It holds NVIDIA Reference Platform Cloud Partner status. Nebius B.V. is part of Nebius Group, the renamed successor to Yandex N.V.: in 2024 Yandex N.V. sold its Russian business operations and rebranded its remaining international assets as Nebius Group, later resuming trading on Nasdaq. Pricing is usage-based but not broken out in detail on the public site.