# Data centers

## Regions

AI Services are publicly available in multiple North American locations.

* Each region is in a separate geographic area to support data sovereignty requirements
* Regions are isolated from other regions with no data replicated between them
* Data centers are spread across multiple data halls to ensure power redundancy and availability

Denvr operates world-class, highly-available data centers.  However AI compute may require unavoidable hardware maintenance due to the high performance nature of data center GPUs and network fabrics.  Code development should use multiple nodes and regions to ensure availability and fault tolerance.

<table><thead><tr><th width="138">Region Id</th><th width="298">Location</th></tr></thead><tbody><tr><td>MSC1</td><td>Canada West (Calgary)</td></tr><tr><td>HOU1</td><td>US South (Houston)</td></tr></tbody></table>

Denvr maintains compliance with SOC 2 Type 2 and GDPR, and others as applicable.

## GPU Availability

The following GPU platforms are available in each public region:

<table><thead><tr><th width="142">Vendor</th><th width="209">Node Types</th><th width="203">GPU Fabric</th><th width="90" data-type="checkbox">HOU1</th><th data-type="checkbox">MSC1</th><th data-hidden align="center">HOU1</th><th data-hidden align="center">MSC1</th></tr></thead><tbody><tr><td>Intel</td><td>8x Gaudi 2 96GB</td><td>2.4 Tbps (RoCE v2)</td><td>true</td><td>false</td><td align="center">Yes</td><td align="center">-</td></tr><tr><td>NVIDIA </td><td>8x H100 80GB SXM</td><td>3.2 Tbps (IB)</td><td>true</td><td>false</td><td align="center">Yes</td><td align="center">-</td></tr><tr><td> </td><td>8x A100 80GB SXM</td><td>1.6 Tbps (IB)</td><td>true</td><td>true</td><td align="center">Yes</td><td align="center">Yes</td></tr><tr><td></td><td>8x A100 40GB SXM</td><td>800 Gbps (IB)</td><td>false</td><td>true</td><td align="center">-</td><td align="center">Yes</td></tr><tr><td></td><td>1x GH200 96GB SXM</td><td>-</td><td>false</td><td>true</td><td align="center"></td><td align="center"></td></tr><tr><td></td><td>4x A100 40GB PCIe</td><td>-</td><td>false</td><td>true</td><td align="center">-</td><td align="center">Yes</td></tr><tr><td></td><td>4x A40 48GB PCIe</td><td>-</td><td>false</td><td>true</td><td align="center">-</td><td align="center">Yes</td></tr><tr><td>AMD</td><td>2x Mi210 64GB PCIe</td><td>-</td><td>false</td><td>true</td><td align="center"></td><td align="center"></td></tr></tbody></table>

* Nodes are available in bare metal configurations with 8X GPUs and virtual machines with 1X, 2X, 4X and 8X GPUs. &#x20;
* GPU fabrics are non-blocking, rail-optimized fat-tree topology.

Bare metal nodes are available in Reserved contracts while VMs are available both On-Demand and Reserved.

## Storage Platforms

The following network attached storage platforms are available in each region:

<table><thead><tr><th width="138">Type</th><th width="408">Performance Class</th><th data-type="checkbox">HOU1</th><th data-type="checkbox">MSC1</th></tr></thead><tbody><tr><td>Block</td><td>High-Performance NVMe - Up to 10 GB/s</td><td>true</td><td>false</td></tr><tr><td>File</td><td>High-Performance NVMe - Up to 10 GB/s</td><td>true</td><td>false</td></tr><tr><td></td><td>General Purpose - Up to 1 GB/s</td><td>false</td><td>true</td></tr></tbody></table>

## Internet Access

The following internet access is available in each region; ISPs are shared by region tenants.

<table><thead><tr><th width="146">Region</th><th width="159">Bandwidth</th><th width="242">Redundant Subscription</th></tr></thead><tbody><tr><td>HOU1</td><td>100G</td><td>Yes</td></tr><tr><td>MSC1</td><td>10G</td><td>No</td></tr></tbody></table>