Your next 256GPUs, reserved
Single-tenant bare metal built around 32 Supermicro HGX B300 nodes, with a deployment plan confirmed for your site and schedule.
Fit your workload to the cluster
Inference
Reserved capacity for predictable, high-volume inference service on dedicated hardware.
Fine-tuning
Single-tenant fabric and bare-metal control for training and fine-tuning runs.
Neocloud
Whole-pod capacity for multi-year offtake by neoclouds and GPU marketplaces.
Pod 32 reference specification
Named components and configuration ranges. The written specification, not the render, is the authority on every value.
- Compute
- 32 Supermicro HGX B300 nodes
- GPUs
- 256 NVIDIA B300
- Racks
- 4
- Operating load
- 645 kW
- System memory
- 128 TB
- Storage
- 983 TB raw NVMe per module
- Fabric
- Rail-optimized 400G Ethernet
- Fabric throughput
- Configurable 3.2 or 6.4 Tbps per node
- Cooling
- Direct-to-chip liquid, redundant CDUs
- Site water
- No site water requirement
- Tenancy
- Single-tenant bare metal
Measured before it ships
Acceptance is evidence, not a brochure. Every module carries witnessed results from a defined test program.
24 stages
A factory, site, and integration test program covering the module from build to handoff.
Gates F-1 to F-7
Named evidence gates carried into acceptance, with witnessable outputs at each gate.
NCCL pass bars
Measured NCCL fabric-performance acceptance criteria on the deployed cluster fabric.
Factory and site work run in parallel
Each module is factory-built and burned in off-site, arrives complete on one truck, and is craned onto a prepared pad. Commissioning is remote. A site-specific plan sets the target schedule - targeted to be operational within roughly three months of pre-payment, confirmed site by site.
- Site scope
- Prepared pad + power feed
- Factory
- Build and burn-in off-site
- Arrival
- Complete on one truck
- Placement
- Craned onto the prepared pad
- Commissioning
- Remote sequence
- Scale
- Pod by pod
Questions before you reserve
How fast?
The schedule is a target confirmed site by site, never a guarantee. Site preparation and factory build run in parallel.
Network?
Choose the 3.2 or 6.4 Tbps per node fabric configuration with our engineering team during the deployment plan.
Commercials?
Discussed privately on a qualified call. Pacific does not publish pricing.
Qualify, then choose a time
Three quick questions - role, workload, timeline - then the calendar opens. No long forms before you see proof.
Your role
Answer the three questions and the calendar opens here. Prefer to skip ahead? Book directly on Cal.com.
Before you book
Who owns and operates the hardware?
Pacific builds, deploys, and operates each module. You get single-tenant bare-metal access to the cluster. Operation is unstaffed on site, run through remote operations with contracted regional field service.
Which network configurations are available?
The fabric is rail-optimized 400G Ethernet with a configurable 3.2 or 6.4 Tbps per node. The configuration is selected with our engineering team during the deployment plan for your workload.
How fast can capacity be live?
Each module is targeted to be operational within roughly three months of pre-payment. That is a design target confirmed site by site, not a guarantee. Site work is limited to a prepared pad and power feed and runs in parallel with the factory build.
What evidence comes with a deployment?
Every module ships with witnessed acceptance evidence from a 24-stage factory, site, and integration test program, including measured NCCL fabric-performance pass bars at gates F-1 through F-7.
What does it cost?
Commercial terms are discussed privately on a qualified capacity call. Pacific does not publish pricing.