AI INFRASTRUCTURE • GIGAPOD
Rack-ready AI compute pod.
GigaPod is our factory-built, pre-engineered compute module for every major Al accelerator family.
what is GigaPod?
The white space module where AI compute runs
What's traditionally assembled on-site, piece by piece, now ships in a single module. Pre-tested and ready to rack.
A 45-foot module with 4N/3 power redundancy, hot aisle containment, and air and liquid cooled solutions to support any ratio of low and high density racks.
The unit arrives plug-and-play, scaling to 90 or 135 feet with a deployment.
No on-site fabrication. Ready to rack.
The gigabase system
The core of Giga’s complete AI data center system.
GigaPod is the compute heart. On a full GigaBase campus, it connects to our other pre-built modules — power, cooling, and backup — through proprietary glue modules.
Chillers
Maglev compressor-driven chillers that reject system heat and feed a closed cooling water loop to each Giga Pod, maintaining stable temperatures under full load.
Heat rejection
UPS
A 3.0 MW lithium-ion battery backup that bridges any gap between utility interruption and generator response.
Power Module
GigaPod
A 3.6 MVA transformer paired with a 5,000A low-voltage switchboard that steps down and distributes voltage to the site.
MV → LV + switchboard
Generator Block
A pre-commissioned compute module with 45 feet of rack space, up to 1.9 megawatts of capacity, and integrated liquid cooling built to support GPU and TPU clusters.
GPUs + CDUs + Fans
Glue Modules
A diesel generator providing 24-hour backup power that comes online when utility power fails, holding the site operational until the grid is restored.
Backup power
Enjoy ongoing support
Proprietary electrical and cooling interconnects that tie every module together and manage power distribution and cooling across the full system.
Pre-built interconnects
Specs at a Glance
Engineered for
GPU density
- Cooling
CDUs + Fan Walls
Supports high-density liquid-cooled and low-density air-cooled configurations.
- Power density
Configurable kW/rack
Racks installed on-site.
- Redundancy
4N/3 distributed redundancy
Any single train can fail and the other three carry the full load.
- Row length
45, 90, or 135 feet
Scales with deployment needs
- COMPATIBILITY
Chip agnostic
NVIDIA · Google · AMD · Cerebras
AI Compute Pod
Built in our factory.
Delivered to your site.
Pre-built, pre-tested, and compatible with every major AI accelerator.
One partner
GigaPod is the centerpiece of a complete Giga-built infrastructure stack. Giga engineers and manufactures everything from power modules to chillers to switchgear.
Plug & play commissioning
Every module ships pre-built, pre-tested, and commissioned from our factory, ready to drop into the whitespace onsite.
Scalable architecture
Our pre-engineered design scales from 8.4 MW IT to full campuses. Start or scale with our reliable building blocks.
Hardware Compatibility
Chip agnostic
Rack what you want.
Power and cooling for every major Al accelerator family — NVIDIA, Google, AMD, and Cerebras.
NVIDIA
NVIDIA is the current standard for AI compute, and our infrastructure is built to power it.
TPUs have specific demands. We've done the engineering work necessary to meet them.
AMD
AMD is moving into the AI space, and our infrastructure is ready to support it.
Cerebras
Cerebras' wafer-scale compute requires special configurations. Giga is prepared to accommodate them.
Talk to Sales
Ready to see a GigaPod on your site plan?
Share a few details and our AI data center team will follow up. If you're in San Francisco, ask about a tour of our Innovation Hub to see the infrastructure up close.
