You don’t need a greenfield project to add GPUs. OpenMetal lets you attach RTX Pro 6000 or H200 GPU servers to an existing Hosted Private Cloud or bare metal deployment, on the same private network as your current compute and storage. This is the path for teams that already run on OpenMetal and want to add AI training or inference next to their existing applications, data, and Ceph storage — without a separate environment or a second vendor.
Key Takeaways
- Same private network — GPU nodes join your existing private VLANs, so they sit next to your application servers, databases, and Ceph storage with low-latency, unmetered east-west traffic.
- No separate environment — augment what you already run; keep one provider, one private network, one billing model.
- Pick the right GPU — attach RP6000 nodes for cost-efficient training and inference, or H200 nodes for the largest-memory, bandwidth-bound work. Mix as needed.
- Single-tenant bare metal — attached GPU nodes are dedicated, with full root access and IPMI, just like the rest of your OpenMetal footprint.
- Fixed monthly pricing — added GPU capacity carries the same predictable, no-“idle-silicon-tax” economics as the rest of your deployment, with included egress.
Ready to Add GPUs to Your Deployment?
Tell us about your existing OpenMetal footprint and the AI workload you want to add, and we’ll help you attach the right GPU nodes.
Common Scenarios
Add inference to an existing application cluster
Attach RP6000 nodes next to your existing OpenMetal Hosted Private Cloud so your application can call local inference endpoints over the private network — no public round-trip, no per-GB egress between app and model.
Add training capacity to a bare metal footprint
Add H200 or RP6000 nodes to an existing bare metal deployment for model training and fine-tuning, pulling datasets from your existing storage over the private 40 Gbps mesh.
Burst or augment with GPUs
Add GPU nodes for a defined project (a fine-tuning run, a new inference feature) and keep them on the same private network as your existing services for the duration.
RAG and data-adjacent AI
Place GPU nodes next to the data they serve — vector stores, document repositories, Ceph object storage — so retrieval and generation happen within the private network.
How Attachment Works
Attached GPU servers are provisioned as dedicated bare metal nodes and joined to your existing private network/VLANs, so they share low-latency, unmetered east-west connectivity with your current OpenMetal compute and storage. Each GPU node is a single-tenant server with dual Intel Xeon 6530P (64C/128T), 1TB DDR5-6400, a 6.4TB Micron 7500 MAX NVMe data drive, and 40 Gbps private / 10 Gbps public networking. See the RP6000 and H200 spec pages, and OpenMetal GPU clusters for multi-node GPU builds.
Security and Compliance
Attached GPU nodes are single-tenant bare metal — physical isolation, no shared hypervisor on the accelerator. Intel SGX is available on the host CPUs. Intel TDX and GPU passthrough cannot be combined in a single trust boundary. GPU nodes attached to Ashburn (NTT DATA VA1) deployments are in a HIPAA-compliant facility; OpenMetal offers BAAs at the organizational level.
Ready to Add GPUs to Your Deployment?
Tell us about your existing OpenMetal footprint and the AI workload you want to add, and we’ll help you attach the right GPU nodes.
Deployment Options
- Attach to Hosted Private Cloud — add GPU nodes to an existing OpenStack + Ceph private cloud.
- Attach to bare metal — add GPU nodes to an existing bare metal footprint.
- Dedicated GPU cluster — for larger multi-node GPU builds, see OpenMetal GPU clusters.
Where to deploy
GPU nodes are available now in Ashburn, Virginia (US-East), with advance reservations for Los Angeles, Amsterdam, and Singapore. Proof of Concept clusters are available for testing; ramp pricing is available for migrations.
Get a GPU Quote
Ready to deploy? Tell us about your existing OpenMetal deployment and the GPU workload you want to add, and we’ll provide a custom quote.
- Attach RP6000: cost-efficient training and inference next to your existing services
- Attach H200: largest-memory, bandwidth-bound AI alongside your current footprint
- GPU cluster: multi-node GPU builds on a private 40 Gbps mesh
All deployments include fixed monthly pricing, included egress, a 99.96%+ network SLA, and DDoS protection.
Related OpenMetal Answers
Product specifications, pricing, and availability may change due to market conditions and other factors. For the most current information, please contact the OpenMetal team directly.