Rafay

Rafay Systems, based in Sunnyvale, California, sells the Rafay Platform. It started as a Kubernetes Operations Platform and has grown into an orchestration layer for both Kubernetes and GPU infrastructure. It is not a host itself. Platform teams use it to manage cluster lifecycle across EKS, AKS, GKE, OKE, private clouds and on-premises environments, with standard blueprints, OPA-based policy, role-based and zero-trust access, a GitOps pipeline service, and backup and restore. Newer modules, AI Factory and Token Factory, let telcos, sovereign clouds and GPU neoclouds offer bare-metal GPUs, VMs, Kubernetes, SLURM clusters, notebooks and token-metered inference to their own tenants, with quotas and chargeback. Reported customers include TELUS, MoneyGram, Guardant Health, Samsung and Palo Alto Networks. Pricing is contract-based: a platform fee plus a per-node or per-GPU/CPU unit fee, with volume discounts. Standard support during US business hours is included, and 24x7 support costs an extra 20%. On AWS Marketplace, a 12-month contract for 50 nodes is listed at $240,000 and a self-hosted controller at $100,000 a year. Rafay connects to Slack, Microsoft Teams, PagerDuty, Jira, ServiceNow, Datadog, Elasticsearch and Loki, and offers a Terraform provider, the RCTL CLI and an API. Its main limitation is cost and complexity. Reviewers describe it as an enterprise governance tool with a learning curve and licensing above many Kubernetes tools, and one competitor's review calls it overkill for teams with only a few clusters.