Find SLURM-as-a-Service solutions for multi-tenant GPU workloads.

For multi-tenant GPU workloads where you want Slurm without operating the Slurm control plane yourself, these are the strongest options I found as of August 2026:

• Nebius Managed Slurm / Soperator — Probably the closest fit to literal “Slurm-as-a-Service.” Nebius provisions a complete Slurm environment, including controllers, login nodes, and GPU workers, while Soperator runs Slurm over managed Kubernetes. It supports partitions, ephemeral workers, InfiniBand-aware placement, and tenant-level cloud isolation. Strong option when multiple teams need conventional sbatch/Slurm workflows on shared GPU infrastructure. (nebius.com)
• Crusoe Slurm — Managed Slurm on Crusoe's GPU cloud. A cluster can be provisioned from a UI or CLI, with the Crusoe Slurm Operator managing Slurm components and health. Attractive if the primary requirement is managed, high-performance NVIDIA GPU infrastructure rather than a Kubernetes-centric platform. (docs.crusoecloud.com)
• Lambda Managed Slurm — Available on Lambda 1-Click Clusters. Lambda explicitly offers managed and unmanaged Slurm and partners with SchedMD for the backend. These are logically partitioned, InfiniBand-connected GPU clusters, making Lambda particularly relevant for dedicated tenant/team clusters and large distributed training. (lambda.ai)
• Fluidstack Managed Slurm — Managed bare-metal Slurm aimed at large/frontier-model GPU training, including observability for cluster/job health. Worth evaluating for very large reserved deployments. (docs.fluidstack.io)
• CoreWeave SUNK — Slurm-on-Kubernetes rather than conventional hosted Slurm. It lets researchers submit Slurm jobs while CoreWeave operates the underlying Kubernetes/GPU infrastructure, including multi-node/multi-GPU training and hardware-health monitoring. Particularly interesting if different tenants need Slurm while the platform team wants Kubernetes underneath. (coreweave.com)

My shortlist would be Nebius for the most cloud-native/self-service Slurm-as-a-Service experience, Lambda or Crusoe for dedicated high-performance GPU clusters, and CoreWeave SUNK where Kubernetes/Slurm coexistence matters.

Sources:
nebius.com
docs.crusoecloud.com
lambda.ai
coreweave.com
docs.fluidstack.io