Factory-built AI data centers at 2.4 and 4.8MW blocks — power, compute, and cooling integrated, sited, and ready to run.
What we build
A conventional site of this size takes years of custom construction. A CGiX site arrives factory-built and tested — power, compute, and cooling integrated into standardized blocks that come online in months, at the power you already have.
CGiX · Block
2.4 MW
The standard deployment. Two self-contained 1.2 MW compute blocks, each with dedicated power, cooling, and standby generation — sized for customer-furnished GB300-class racks.
| Configuration | 2 x 1.2 MW blocks |
|---|---|
| Redundancy | 2N CDU · N+1 UPS |
| Compute | GB300-class |
| Footprint | ~2,200 m² |
CGiX · Block
4.8 MW
The flagship module. Two 2.4 MW pods combine into a single 4.8 MW ComputeBlock with pooled redundancy across the whole package — built for scale deployments that grow in multiples.
| Configuration | 2 x 2.4 MW pods |
|---|---|
| Redundancy | 4-to-make-3 pooled |
| Compute | GB300-class |
| Footprint | Scales from 2.4 MW |
Designed for scale
Module · 01
Contained white-space for GB300-class racks — busway, liquid-cooling manifolds, containment, and environmental monitoring, prepared for customer-furnished racks.
Module · 02
MV switchgear, transformers, IT and mechanical UPS, and LV distribution — factory-assembled and FAT-tested as one prefabricated module.
Module · 03
Liquid-first heat rejection with residual air. CDU / fan-wall skids and adiabatic dry coolers, factory-matched and sequenced to the load.
Module · 04
Factory-packaged standby gensets with controls and fuel systems — backup power for the site, deployable outdoor or indoor.
Every module is factory-integrated and FAT-tested before shipment.
See the subsystem
Built for redundance
Both sizes are engineered for high availability with redundant power and cooling — but they get there two different ways. The right architecture depends on the site.
2.4 MW standard site
2N CDU • N+1 UPS • N+1 pumps/fans
Each 1.2 MW block carries its own dedicated power, cooling, and standby generation — redundancy is contained within the block rather than shared across the site. A fault in one block never reaches the other.
4.8 MW module
4-to-make-3 pooled
The larger module shares redundancy across the whole package — a “4-to-make-3” topology where any one of four units can drop without losing capacity. More efficient at scale, with resilience spread across the module.
Runs compute
Purpose-built for customer-furnished GB300-class racks, with liquid-first cooling designed for continuous, near-full-load operation — ready for training and inference the moment the site is energized.
Ready for the grid
Grid-tied by design, with standby generation for resilience. The modular architecture goes to where power and land are already available — collapsing the path from megawatts to usable AI capacity.
Returns energy
Integrated battery storage lets every unit participate in the energy market — peak shaving, arbitrage, and demand response built in.
Tell us your power, your workload, and your site. We’ll scope the right configuration.