Regions & Providers
Spheron sources GPU capacity from six providers across North America and Europe.
Provider Capabilities
| Provider | Instance Types | Volume Support | K8s Addon | GPU Tiers | Notes |
|---|---|---|---|---|---|
| Voltage Park | Dedicated, Cluster | NFS volumes (up to 64 TB) | Yes (Cluster only) | H100 SXM5: Cluster (InfiniBand 3.2 Tbps or Ethernet 100 Gbps) | Dallas TX, external NVMe |
| DataCrunch | Spot, Dedicated | NFS volumes (up to 10 TB, multi-instance) | No | A100, H100 | EU (Finland FIN-01/02/03) |
| TensorDock | Dedicated | No | No | RTX 4090, A100 | US/EU, VM-based |
| Sesterce | Dedicated | No | No | Various | EU regions |
| Spheron AI | Spot, Dedicated | No | No | H200 SXM, H100, A100, L40S, L40, A6000, RTX Pro 6000, A4000 | Canada, Norway, US (Tier 3 DC) |
| Massed Compute | Dedicated | No | No | RTX A6000, others | US regions |
Provider Details
Voltage Park
Voltage Park provides Dedicated and Cluster servers in Dallas, TX. Voltage Park is the only provider that supports:
- The Kubernetes addon for managed K8s clusters (requires a Cluster instance type with H100 SXM5 offers)
- Multi-GPU bare-metal in two networking variants: InfiniBand (3.2 Tbps) for maximum DDP/DeepSpeed throughput and Ethernet (100 Gbps) for standard multi-GPU workloads, both deployed as Cluster instances
- External NVMe drives for high-speed local dataset storage
- NFS persistent volumes for cross-deployment data sharing (up to 64 TB; 1 volume per instance)
Compliance certifications: SOC 2 Type II, SOC 2 Type I, HIPAA, ISO 27001.
DataCrunch
DataCrunch operates in EU regions across three Finland data centers: FIN-01, FIN-02, and FIN-03. It provides both Spot and Dedicated instances with NFS volume support.
DataCrunch volumes (NVMe_Shared NFS type) support multi-instance attachment: a single volume can be mounted on multiple GPU instances simultaneously (unlike Voltage Park, which restricts a volume to one instance at a time). Each instance supports at most one shared volume.
Local storage for Dedicated instances: When deploying a DataCrunch Dedicated instance you choose between two root-disk sizes: 500 GB (default) or 1 TB. Select 1 TB when your workload requires more local dataset or checkpoint space than the default provides.
TensorDock
TensorDock provides Dedicated VM-based GPU instances in US and EU regions with RTX 4090 and A100 GPUs.
Volume mounting is not supported on TensorDock. Use instance-local NVMe storage for datasets.
Sesterce
Sesterce provides Dedicated GPU instances in EU regions with various GPU tiers (GDPR, SOC 2 Type II, ISO 27001 certified). Volume mounting is not currently supported.
Spheron AI
Spheron AI provides on-demand and spot GPU instances across North America and Europe, sourced from Tier 3 data centers. Available GPU models include H200 SXM, H100 SXM5/NVL/PCIe, A100 SXM4/NVL/PCIe, L40S, L40, A6000, RTX Pro 6000, and A4000 across Canada (Montreal), Norway (Bergen), and the United States.
Startup scripts (cloud-init) are fully supported. Volume mounting is not currently supported; use instance-local storage for datasets.
OS images: 11 OS images are available, including Ubuntu 22.04/24.04, AlmaLinux 9, Debian 12, and CUDA-enabled variants (R535, R550, R570). The default is Ubuntu Server 22.04 LTS with CUDA 12.8 and Docker pre-installed.
SSH access: Port 22 opens automatically on every new Spheron AI instance; no firewall configuration is needed for SSH.
Massed Compute
Massed Compute provides Dedicated GPU instances in US regions (including Des Moines, IA) with RTX A6000 and other GPU configurations. Volume mounting is not currently supported.
Compliance certifications: HIPAA, SOC 2 Type II.
Region Selection Guidance
- Lowest SSH latency: Choose the provider and region geographically closest to you
- GDPR / EU data requirements: Use DataCrunch (Finland, ISO 27001) or Sesterce (EU)
- H200 GPUs: Spheron AI (Canada, Norway, US)
- Largest H100 bare-metal clusters: Voltage Park Dallas
- Shared persistent volumes: Voltage Park (1 volume per instance) or DataCrunch (multi-instance NFS attach within the same region)
- Reserved GPU requests: Select "Any Location" when submitting reserved GPU requests to maximize provider competition and get the best price
Additional Resources
- API Reference: Query available providers and GPU offers programmatically
- Instance Types: Spot vs Dedicated vs Cluster
- Kubernetes addon: Voltage Park K8s setup
- Volume Mounting: NFS storage for Voltage Park and DataCrunch
- External Storage: NVMe drives on Voltage Park
- Reserved GPUs: Long-term GPU reservations