02Infrastructure

ONE AMBITION.
THREE REGIONS.

Customer equipment and GPU deployments, supported through our partner data-center operations. Explore each region’s active platforms and construction programs.

Regional overview · locations are generalized
01 / PARTNER DATA-CENTER OPERATIONS

North Western Canada

Our partner data-center operations in Vancouver and Calgary support GB300 clusters and customer equipment colocation. The Vera Rubin cluster program in Western Canada is under construction.

Operational

Vancouver & Calgary

NVIDIA GB300

Under construction

Western Canada

NVIDIA Vera Rubin

SYSTEM ARCHITECTURE

FOUR LAYERS.
ONE WORKING SYSTEM.

01

Allocate the working set

Model weights are only one part of GPU memory. Account for KV cache, activations, runtime buffers and concurrency, then test the selected precision and software build.

  • GPU count, edition and usable memory
  • CPU, host RAM and NUMA layout
  • Driver, CUDA, framework and model versions
NEXT / Under construction

Building the next capacity.

Two programs extend our partner infrastructure footprint: Vera Rubin cluster construction in Western Canada and the Osaka data-center development. Both are under construction. Installation, integration, acceptance and customer allocation are separate milestones.

Under construction

New data center

Osaka

Under construction

NVIDIA Vera Rubin

Western Canada

READ / ENGINEERING

BEFORE THE FIRST JOB.

02
DEPLOYMENT / ACCEPTANCE / 2 min read

Accept a cluster with evidence.

The inventory, software baseline, communication tests and workload results that belong in the handover record.

04
REGIONAL DEPLOYMENT / 2 min read

Plan the data path across regions.

Site, platform, data location, support access and exit: the five parts of a regional deployment record.

06
CONSTRUCTION / ROADMAP / 2 min read

Commission Vera Rubin, gate by gate.

The facility, integration, workload and operating records required before a construction project becomes a compute service.