Skip to content
Aerial view of black modular data center containers with ground mounted fluid coolers and a transformer

Modular data centers

Dense compute,
built in the factory.

Modules engineered for GPU workloads and delivered in 3 to 4 months, from a single edge module to a multi megawatt campus.

Delivered in
3 to 4 months
One platform
Edge to campus
Cooling, by GPU class
Air or liquid
Request specs and pricing

Request specs and pricing

  1. Project
  2. Contact
Modular data centers: project details

0 of 2000 characters. Longer notes can go in an attached document.

The reality

Your GPUs need power density, not a construction project

Halls designed for air cooled servers were not built for today's GPU racks, and a new building means a long construction project. A module designed for dense, liquid cooled compute skips the retrofit.

A black BiXBiT modular data center container on a concrete pad at blue hour

Why modular

What a factory built module changes

Capacity in 3 to 4 months

The module is built in the factory while site work proceeds, instead of one waiting on the other.

Density a legacy hall cannot cool

Cold plate liquid cooling handles the GPU racks that air cooled halls were never designed for.

One system, not a parts list

Power, cooling and monitoring arrive integrated in one unit instead of being assembled from parts on site.

Scale one block at a time

Start with a single edge module and grow on the same platform to a multi megawatt campus.

Runs on less overhead

Liquid cooling moves heat more efficiently than air, so more of the power you pay for reaches the GPUs.

See every circuit and rack

Circuit level power telemetry and sensors on every rack report into your monitoring systems.

Stop waiting on a building. Ship the data center instead.

One platform

From a single edge module to a campus

The same design scales by adding modules, so the first deployment is not a dead end.

Edge

Single module

A compact module for edge sites, inference and small clusters.

AI module

Liquid cooled

A module built around dense, liquid cooled GPU racks.

Block

Grouped modules

Modules grouped into a block with shared power distribution and redundancy.

Campus

Multi megawatt

Multiple blocks on one site, built out as demand grows.

Technical specifications

Built for GPU density

Standard container form factors, with power and cooling configured for the GPU class you deploy.

Power

Input voltage
480 V, 3 phaseOther distribution voltages configured per deployment
Distribution
Busway with a tap off per rackSized to the IT load
Redundancy
N, N+1 or 2NConfigured per deployment
UPS
Online double conversionBattery chemistry selected per project
Power monitoring
Circuit level telemetryReports to standard building and DCIM systems
Transfer switch
Automatic transfer switchFor utility and generator sources
Generator ready
Generator connection provisionsGeneration available from BiXBiT USA

Cooling

Cooling modality
Air, hybrid direct liquid or full liquidSelected per GPU class
Liquid cooling
Coolant distribution unitsSized to the load, with redundant pumps on AI configurations
Heat rejection
Dry or adiabatic fluid coolersSelected for the site climate and water availability
Air side
Contained hot and cold aislesHandles the air cooled share of the load
Thermal monitoring
Sensors on every rackTemperature, flow, pressure and leak detection
Cooling redundancy
N, N+1 or 2NConfigured per deployment

Cooling by GPU class

Edge, inference and enterprise
Precision air with containment
H100 and H200 class clusters
Hybrid: direct to chip liquid plus air
GB200 and GB300 NVL72 class
Cold plate direct liquid cooling

Final cooling design depends on rack configuration, GPU generation and site conditions. NVIDIA and other third party names are trademarks of their owners.

Form factor and delivery

  • 20 ft and 40 ft ISO container form factors
  • Ships on a standard flatbed and sets on a prepared pad
  • Racks, power, cooling and monitoring integrated in the factory

The process

From concept to capacity

  1. 01

    Site and load review

    We review your power availability, GPU class, redundancy level and site conditions.

  2. 02

    Configuration and drawings

    Our engineering team prepares the configuration, pricing and drawings for your load and site.

  3. 03

    Factory build

    Racks, cooling, power distribution and monitoring are integrated in the factory as one unit.

  4. 04

    Delivery and startup

    The module ships on a standard flatbed, connects to power and cooling, and is commissioned with our team.

Delivery time: 3 to 4 months

Rows of liquid cooled GPU racks inside a modular data center

Standards and compliance

Built to standards that matter

NEC
National Electrical Code for all power distribution
UL 2755
Safety standard for modular data center equipment
NFPA 75
Fire protection for IT equipment
IBC
Structural design for seismic zones and wind loads
ASHRAE TC9.9
Thermal guidelines for data processing environments

Modules are designed to these standards. Certification scope is set per module at final engineering review, and third party certification is available on request.

Components

Established parts in every subsystem

Each subsystem is sourced from established OEM vendors.

Electrical
UPS, switchboards and power distribution from established OEMs
Cooling
CDUs, fluid coolers and air handling from established OEMs
Monitoring
Power and thermal telemetry with DCIM integration
Racks and cabling
Industry standard racks and structured cabling
Delivered in
3 to 4 months
One platform
Edge to campus
Cooling, by GPU class
Air or liquid
Redundancy
N, N+1 or 2N

Questions

Common questions about modular data centers

How fast can a module be delivered?

Delivery takes 3 to 4 months. Site work proceeds while the module is built in the factory.

How big can a deployment get?

One platform runs from a single edge module to a multi megawatt campus, adding modules as your load grows.

What rack density do you support?

From standard enterprise racks to liquid cooled NVL72 class GPU racks. Tell us your GPU class and we size power and cooling to it.

What cooling options are there?

Air, hybrid direct liquid cooling or full liquid, selected for the GPU class you are deploying.

How efficient is it?

Liquid cooled AI configurations run with far less cooling overhead than air cooled halls. Design PUE is set per deployment and depends on the cooling mode and site climate.

What redundancy can I get?

N, N+1 or 2N for both power and cooling, configured per deployment.

What power does a module need?

480 V, 3 phase input, with other distribution voltages configured per deployment. We also supply the transformers, switchboards and generation that feed it.

Which standards are the modules designed to?

NEC, UL 2755, NFPA 75, IBC seismic and wind, and ASHRAE TC9.9. Certification scope is set per module at final engineering review.

How fast do I get a quote?

Submit your IT load, GPU class and site, and our engineering team comes back with a configuration and pricing.

What is the warranty?

12 to 24 months, with extended warranties available.

Configure your module

Tell us your IT load and GPU class, and we will come back with a configuration and pricing.