Run:ai
GPU orchestration for AI workloads on Kubernetes.
Acquired — still operating
acquired by NVIDIA late 2024 ($700-800M)
About
Run:ai is a Kubernetes-based platform for scheduling and managing AI workloads across GPU infrastructure. It pools and shares GPUs with fractional allocation, workload prioritization, organizational quotas, and utilization monitoring, covering interactive development, model training, and inference.
Founded in Israel in 2018 by Omri Geller and Ronen Dar, the company was acquired by NVIDIA in late 2024 for a reported $700 to $800 million after regulatory review. NVIDIA has integrated it into its AI infrastructure stack including DGX Cloud.
Lifecycle note: in October 2026 NVIDIA released the Run:ai platform as open source, including the KAI Scheduler under Apache 2.0, contributed to the Cloud Native Computing Foundation as a Sandbox project, alongside modular components for GPU fractioning and resource management.
What it can do
- GPU pooling and fractional allocation
- Workload prioritization and quotas
- KAI Scheduler gang scheduling
- Topology-aware GPU sharing (MIG, MPS)
- Utilization monitoring and RBAC
Setup
Deploy the open-source Run:ai platform on a Kubernetes cluster via NVIDIA's developer resources.
Prerequisites
Kubernetes cluster with NVIDIA GPUs; cluster admin access.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 117; Trending score 84/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.