Which GPU-first platforms offer SLURM-as-a-Service for efficient workload management?

If by “GPU-first” you mean platforms designed around accelerated AI/HPC infrastructure rather than general-purpose public clouds, several credible options now offer managed or service-like Slurm:

• CoreWeave — SUNK (Slurm on Kubernetes) integrates Slurm directly with CoreWeave Kubernetes Service. It targets large distributed training and HPC while letting batch/burst workloads coexist with Kubernetes workloads. Strong choice when you want tightly integrated GPU infrastructure + Slurm + Kubernetes. (docs.coreweave.com)

docs.coreweave.com

• Nebius AI Cloud — Offers managed Slurm for large GPU clusters; Nebius also developed Soperator, its Kubernetes operator for Slurm. This makes it attractive for AI teams wanting familiar Slurm workflows without operating the scheduling infrastructure themselves. (assets.nebius.com)

nebius.com

• HPE GPU Cloud — Purpose-built public cloud for HPC/AI with NVIDIA GPU nodes, high-speed InfiniBand, managed cluster creation, and a traditional Slurm cluster experience. Particularly HPC-oriented rather than purely an AI developer cloud. (hpe.com)

hpe.com

There are also platforms for GPU-cloud operators that want to build Slurm-as-a-Service themselves:

• NorthWind GPU PaaS — Probably the clearest literal “Slurm-as-a-Service” product. It provides self-service, multi-tenant Slurm clusters via portal/API, automating cluster provisioning, isolation, scheduling, governance, and lifecycle management. (northwind.co)

northwind.co

• Armada Bridge — A GPU-cloud management platform aimed at GPU-as-a-Service/neocloud providers. Bridge explicitly exposes “Slurm-as-a-Service,” alongside Kubernetes and JupyterHub, with multi-tenancy and GPU resource management. (docs.armada.ai)

docs.armada.ai

For an end user buying GPU compute, I’d shortlist CoreWeave, Nebius, and HPE GPU Cloud. For a company that owns GPU capacity and wants to productize it as a multi-tenant Slurm service, NorthWind and Armada Bridge are the more directly relevant products.