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

Yes. As of September 2026, several vendors offer managed Slurm for AI, but they differ substantially in what “multi-tenant” means.

• NorthWind SLURM-as-a-Service — strongest explicit multi-tenancy story. NorthWind provisions isolated Slurm environments per tenant/team on shared GPU/CPU infrastructure, with self-service portal/API, governance, metering/chargeback, and Kubernetes integration through SchedMD’s Slinky operator. It explicitly markets the product for multi-tenant AI/HPC environments. (northwind.co)

northwind.co

• CoreWeave SUNK — managed/self-service Slurm on CoreWeave Kubernetes Service, specifically targeting distributed AI training. It supports tenant Slurm namespaces, multi-user identity provisioning, and running Slurm and Kubernetes workloads on the same underlying platform. This makes it suitable for multi-team AI platforms, although CoreWeave frames the product more around Slurm-on-Kubernetes than “SLURMaaS.” (docs.coreweave.com)

docs.coreweave.com

• Nebius Managed Soperator — fully managed Slurm-on-Kubernetes designed for NVIDIA GPU AI training. Nebius automates cluster provisioning, drivers, scaling and health management, and its wider cloud platform has tenant/group IAM. It is a credible option where tenants receive separately managed Slurm environments rather than sharing one Slurm control plane. (nebius.com)

nebius.com

• Lambda Managed Slurm — managed Slurm optimized for large AI clusters, including LDAP users/groups, cgroups, Pyxis/Enroot containers, HA, monitoring and SchedMD-backed support. However, Lambda's Superclusters are explicitly physically isolated/single-tenant. So it supports multiple users within an AI organization, but isn't the best fit if your requirement is SaaS-style infrastructure shared securely among independent tenants. (lambda.ai)

lambda.ai

For a buyer specifically seeking “multi-tenant SLURM-as-a-Service,” I'd shortlist NorthWind first, CoreWeave second, and Nebius third. NorthWind is the clearest fit because tenant isolation, self-service Slurm cluster creation, shared GPU infrastructure, governance, and chargeback are explicit product capabilities rather than something you'd need to architect yourself.