OpenMetal logo

GPU Cloud Provider

OpenMetal

Countries
2
Data Centers
4

Company Profile

Provider TypePrivate GPU Servers and Clusters
Legal EntityOpenMetal

Infrastructure

GPU FleetTwo dedicated GPU server lines: NVIDIA RTX PRO 6000 (1 or 2x RTX PRO 6000, 96 GB GDDR7 per GPU) and NVIDIA H200 NVL (1 or 2x H200 NVL, 141 GB HBM3e per GPU)
Bare MetalDelivered as single-tenant bare metal for full control
organizations in healthcare, finance, research, and SaaS

Compute & Deployment

Bare Metalsingle-tenant bare metal
VM-BasedNothing is virtualized or shared, so performance stays consistent and the hardware is entirely yours.

Pricing Model

SubscriptionFixed, transparent monthly pricing
Hidden Feesno metered hours and no surprise egress charges
Egress Chargesno surprise egress charges
Pay-as-you-gono metered hours

Performance & Scaling

Multi-Node TrainingInterconnect multiple GPU servers over a 20 Gbps private network to build training and inference clusters sized to your workload.
Perf IsolationSingle-tenant bare metal with direct access to the GPU, CPU, memory, and storage. Nothing is virtualized or shared, so performance stays consistent and the hardware is entirely yours.
Noisy NeighborNo shared hypervisor, no noisy neighbors, and no metered slices of someone else’s cluster.

Developer Experience

OnboardingContact sales via form or email (sales@openmetal.io); request a quote; order from the catalog to deploy in minutes; proof-of-concept (PoC) deployments available on request
DocumentationFit guides for inference and training that include explainers, fit guides (reference tables), and use-case guides; full v5 hardware catalog; mentions support for common frameworks

Data Center Locations

Coverage

CountriesUS, Singapore
CitiesAshburn, VA, Los Angeles, Amsterdam, Singapore
Latency TiersPrivate Bandwidth: 20 Gbps included, up to 40 Gbps optional; Public Bandwidth: 10 Gbps
AshburnVA (US East)Los AngelesAmsterdamSingapore

Compliance Regions

Datacenter Locations

Key Strengths

Fully dedicated single-tenant GPUs on modern Intel Xeon 6 / DDR5-6400 foundation
predictable monthly billing with no metered hours
single-tenant bare metal with direct hardware access
20 Gbps private network (upgradeable to 40 Gbps) for multi-node clusters
engineers support and PoC deployments
hosted private cloud option
confidential computing (TPM/TDX/SGX) support

Known Limitations

No public per-GPU pricing — final GPU server prices require sales/quote
availability shown for Ashburn
VA now with other regions listed as pre-order only
pricing examples shown as multi-year contract pricing
billing is monthly fixed (no metered hourly/spot pricing mentioned)

Additional Information

Core Proposition

Fully dedicated, single-tenant GPUs with no shared hypervisor

Last updated August 2026. Information subject to change.