Fog Hashing is now Fog Computing.Read the announcement
AI data center ยท FogStack

AI data center,
edge to hyperscale.

FogStack is fully integrated, prefabricated infrastructure for high-density AI compute, scaling from a single pod to a hyperscale campus.

Talk to sales
Aerial rendering of a 10 MW FogStack site: the data hall with its module rows behind it and a line of chillers and buffer tanks along its front, the substation and retention pond beyond, trucks at the gate and the access road running past the fence

Built in the factory,
assembled on site.

The 10 MW reference design runs 160 racks, 64 of them compute, with N+1 cooling, 4N/3 power redundancy and diesel generators for backup.

Rendering of a FogStack site from the air: the hall with its open-top compute pods along a piping spine, a row of power modules and a row of diesel generators behind it, and on the other side the hydro pod containers against the wall with the air-cooled chillers and their buffer tanks beyond them

Containerized data hall integrating compute racks, busway, CDUs and piping, network cabling and aisle containment. Liquid and air cooling are laid out together as one system.

Prefabricated all-in-one power module, integrating LV switchboards, UPS and batteries, LV output distribution, cooling and fire suppression. Installed outdoors with no construction needed.

Diesel gensets in a container, with controls, fuel system and exhaust management built in. Prime or backup power for the site.

A container that houses the fluid support system, including pressure stabilization, water treatment, ventilation and pumps. It cuts on-site work and minimizes operational disruption.

Magnetic-bearing chillers with free cooling and a thermal backup tank. Low WUE, built for harsh and variable climates.

Built for
Frontier AI Engines

NVIDIA GB300 NVL72 rack, compute and switch trays cabled at the frontNVIDIA GB300 NVL72
NVIDIA GB200 NVL72 rack, compute and switch trays cabled at the frontNVIDIA GB200 NVL72
NVIDIA HGX B300 baseboard, eight GPU modules on the trayNVIDIA HGX B300
HGX platform in a rack, servers stacked in a Supermicro cabinetHGX Platform

One system,
built to expand.

Every module is standardized and shares one design, so a site expands by adding modules rather than by redesigning it.

Designed for superintelligence

Liquid cooling takes GB300 NVL72 and B300 racks at full density, and cooling and power both run N+1 for Tier III reliability.

Prefabricated in the factory

Power, cooling and the compute pods are built and tested before they ship, while the site is prepared. What is left on site is assembly.

Backed by the people who build it

On a supply chain of industry leaders, we work with you directly from the start to power on. You reach a person who knows your project.

Rendering of the FogStack IT hall: liquid-cooled rack rows on both sides, coolant hoses arcing overhead down the cold aisle to the lit doors at the far end

How we compress
the schedule.

Power and cooling set the date a project goes live, not the GPU order behind it. Building them in a factory while the site is prepared is where the months come back.

Site design
Permitting
Site construction
Manufacturing
Shipping
MEP installation
Testing and commissioning
The Fog WayFast-tracked conventional build6 months saved

Build by the module.

Talk to salesSee FogPod Air