The Pacific platform
A repeatable unit of AI infrastructure.
A physical platform: dedicated GPU compute, integrated support systems, and an operating model carried from the factory to the site.
The building block
Meet Pod 32.
The reference module combines 32 Supermicro HGX B300 nodes, 256 NVIDIA B300 GPUs, and rail-optimized 400G Ethernet. The cluster is single tenant, with bare-metal access.
- Storage
- 983 TB raw NVMe per module
- Fabric throughput
- Configurable 3.2 or 6.4 Tbps per node
- Thermal system
- Direct-to-chip liquid cooling, redundant CDUs
The deployment model
The module meets the site.
A factory-built system is installed on a prepared pad with a power feed. Factory work and site preparation proceed in parallel, with timing confirmed through the individual deployment plan.
Factory, site, and integration acceptance gates connect the reference specification to recorded results. Pacific operates remotely with contracted regional field service.
See the deployment processThe right configuration
Choose the path around the requirement.
Reserve a cluster, assess owned GPU hardware for colocation, or evaluate bridge capacity while a permanent site is completed. Each path has its own fit review and agreed responsibilities.
The next conversation
Start with your site. Your workload. Your constraints.
Review the capacity, site requirements, and acceptance plan with Pacific.
Discuss your deployment