# Compute

Dedicated bare metal and isolated private cloud - with NVIDIA H100 / H200 / Blackwell GPU options for AI training and inference at scale.

Bitwrx Compute is infrastructure on demand. Dedicated bare metal in under 30 minutes, multi-tenant or single-tenant private cloud, and GPU clusters built for AI workloads. Every node lives inside a Bitwrx data center, every byte stays in Canada, and every billing line is transparent.

## Capabilities

What's included in Bitwrx Compute.

🖥️  
### Bare metal in 30 min

Provision dedicated servers (AMD EPYC, Intel Xeon, NVIDIA Grace) via API or portal. Hourly, monthly, or 1–3yr.

🧠  
### GPU clusters

H100, H200, and Blackwell GPU nodes. NVLink and InfiniBand HDR/NDR for tightly-coupled training.

☁️  
### Private cloud

VMware, OpenStack, or Bitwrx-managed KVM. Single-tenant, dedicated hosts, dedicated storage.

💾  
### NVMe & all-flash

Local NVMe up to 60TB per node. Optional dedicated storage tier (NetApp, Pure, Ceph).

🔒  
### Sovereign by default

Data never leaves the chosen Canadian region. CSE / PROTECTED B-ready configurations on request.

📈  
### Burst to cloud

Connect to AWS / Azure / GCP via Bitwrx Connect when you need to spill compute or storage.

## By the numbers

**-97min** Bare Metal Provisioning  
**-26x** H200 GPUs per node  
**-1291Gbps** InfiniBand per host  
**-194TB** NVMe per host

## Compute that runs where your data lives.

Spin up a GPU cluster, a private cloud, or a single bare-metal node - same portal, same SLA.