Pacific
Pacific deploys portable AI data centers on pre-developed construction sites with unused grid connections, turning idle power into revenue within days. Each factory-built pod integrates 256 NVIDIA B300 GPUs with direct-to-chip liquid cooling and single-tenant bare-metal access, targeting high-volume inference, training, and whole-pod offtake workloads.
Funding
Raised to date
$500KRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Founder details are not available yet.
Product
Problem
Delayed construction sites in the US lose $1 billion per week while sitting on unused grid connections and valuable power capacity. Meanwhile, data centers face multi-year waits for power access, creating a major bottleneck for AI infrastructure deployment.
Solution
Pacific deploys portable AI data centers on pre-developed construction sites with unused grid connections, turning idle power into revenue within days. Each factory-built module integrates compute, cooling, and power distribution into a single machine, replacing the complexity of one-off construction projects. The company designs and builds modular factories that can be stood up quickly near power, labor, and demand globally. The pods support high-volume inference, training and fine-tuning workloads, and whole-pod offtake for neoclouds or GPU marketplaces, with single-tenant bare-metal access and rail-optimized 400G Ethernet fabric.
Target Audience
Primary customers are AI workloads requiring dedicated infrastructure, including enterprises running high-volume inference, organizations performing training and fine-tuning, and neoclouds or GPU marketplaces seeking multi-year whole-pod capacity.
Features
- Factory-built pods integrating 32 Supermicro HGX B300 nodes with 256 NVIDIA B300 GPUs in 4 racks, operating at 645 kW load
- Direct-to-chip liquid cooling with redundant CDUs and no site water requirement, with diamond wafers in development to eliminate evaporative water use in arid environments
- Single-tenant bare-metal tenancy with 128 TB system memory and 983 TB raw NVMe storage per module
- Configurable rail-optimized 400G Ethernet fabric at 3.2 or 6.4 Tbps per node
- 24-stage factory, site, and integration test program with witnessed acceptance evidence including measured NCCL fabric-performance pass bars at gates F-1 through F-7
- Factory-as-a-product model enabling rapid deployment of modular factories globally and a franchise model for operators