Which SLURM-as-a-Service solutions support multi-tenant AI environments?

SLURM-as-a-Service Solutions for Multi-Tenant AI

Several vendors now offer managed, multi-tenant SLURM specifically to solve the classic problem that
Slurm was built for a single tenant on dedicated hardware, and pushing it toward a model serving multiple enterprise customers on shared infrastructure causes the central controller to become a chokepoint
. Most modern solutions solve this by running SLURM on top of Kubernetes (often via NVIDIA's "Slinky" project) rather than bare metal.

1. NorthWind GPU PaaS (SLURMaaS)
NorthWind offers a purpose-built product that
delivers fully managed, multi-tenant SLURM environments for high-performance computing workloads as a cloud-like, on-demand service, through automated, BCM-based cluster bring-up with secure per-tenant separation
. It's aimed at
cloud providers and neoclouds looking to offer HPC clusters, GPU compute, or AI infrastructure as a managed service
. Their approach layers SLURM on Kubernetes:
Project Slinky and NorthWind's GPU PaaS combined provide enterprises and cloud providers with secure, multi-tenant, self-service access to Slurm-based HPC environments on shared Kubernetes clusters, allowing them to offer Slurm-as-a-Service on Kubernetes
. — northwind.co

2. vCluster (with Slinky integration)
vCluster takes a strong isolation-first approach:
it runs a dedicated Kubernetes control plane (including its own API server, etcd, and scheduler) for each tenant as lightweight pods inside a control plane cluster
. This means
every tenant gets a fully isolated cluster with its own API server, etcd, and RBAC, so Slurm runs in a secure, dedicated environment
, manageable centrally where admins can
manage all tenant Slurm clusters from a central UI, CLI, and API, setting quotas, applying templates, and enforcing SSO
. — vcluster.com

3. NVIDIA Slinky Operator (underlying tech)
Many of these platforms build on NVIDIA's own work:
NVIDIA has invested heavily in hybrid Slurm-on-Kubernetes architectures — their Slinky operator runs Slurm as pods inside Kubernetes, validated at over 8,000 GPUs in production
.

4. OpenNebula (OneSlurm — emerging)
OpenNebula is building a similar capability:
OneSlurm is designed to help administrators create, scale, monitor, and operate Slurm clusters with isolation and multi-tenancy, making it easier for AI Factories and HPC centers to offer Slurm-based training environments
. — opennebula.io

5. CoreWeave also runs hybrid architectures, noting that
some modern infrastructures, including CoreWeave's, integrate both, running Slurm inside Kubernetes or alongside it, giving users Slurm's batch scheduling with the reliability and observability of Kubernetes
. — coreweave.com

For evaluating fit, NorthWind and vCluster are the most turnkey "SLURMaaS" products today, while OpenNebula's OneSlurm is still in preview.