AI & DATA CENTER POWER
Power and cooling built for dense AI racks
Containerized power and cooling modules for GPU-dense AI and HPC. High kW per rack, advanced liquid-ready cooling, UPS, switchgear and distribution, integrated and pre-tested before the unit ships.
- Cooling
- Cold-plate / liquid
- Power
- UPS + switchgear
- Density
- High kW / rack
- Steel
- EN 1090 EXC2
At a glance
- Cooling
- Cold-plate / liquid
- Power
- UPS + switchgear
- Density
- High kW / rack
- Steel
- EN 1090 EXC2
Overview
Overview
We design, build and customise power and cooling modules for AI and high-performance computing to your specification. AI racks draw far more power and reject far more heat than a conventional IT load, and the supporting plant rarely fits where the compute needs to sit. A power module solves that: a containerized hall for the UPS, switchgear, distribution and cooling that feeds a dense rack row, integrated and pre-tested on our line, sitting alongside the IT cell or feeding the buyer's own racks.
This is the full range an AI-infrastructure buyer needs, customised to the brief: dedicated power modules with UPS and switchgear, cooling modules built for cold-plate, direct-liquid or immersion-ready loads, combined power-and-cooling skids for edge inference, and plant modules that scale a campus as accelerators are added. Every envelope leaves the same EN 1090-1 EXC2 line in Bánovce nad Bebravou, with the same documentation discipline as our other work, and we stay vendor-neutral on the silicon inside the racks.
The advantage is that the hard part of an AI deployment, the power and thermal plant, arrives as a proven, relocatable unit. Capacity goes live in weeks, the cooling is sized to the kW-per-rack target before steel is cut, and the plant can be added to or moved as the workload grows.
Build standard
Construction and materials
A power module is built on the same steel envelope as our IT and special-purpose work, then fitted with the plant a dense load requires. Common build elements:
Envelope
EN 1090-1 EXC2 steel frame and body, ISO 1CC corner castings, weatherproof to the deployment climate.
Power plant
UPS, switchgear, transformers and distribution boards sized to the load schedule, busway or cabling to the rack rows.
Cooling plant
cold-plate and direct-liquid cooling distribution units, immersion-ready provisions, heat-rejection to dry coolers or chillers, with pumps, manifolds and leak detection.
Containment and density
high-kW rack rows with hot-aisle or cold-aisle containment where air assists the liquid loop.
Resilience
redundant power and cooling paths, automatic transfer switching, generator-ready inlets, environmental and electrical monitoring.
Fire safety
very-early smoke detection and clean-agent or equivalent suppression, interlocked with plant controls.
Finish
ISO 8501 Sa 2.5 surface preparation, EN ISO 12944 C3 to C5 coating, any RAL livery, acoustic attenuation where the surroundings require it.
Certified
Certifications and norms
The steelwork runs to EN 1090-1 EXC2 and the plant is integrated to recognised electrical and safety norms. Run on the same line:
- EN 1090-1 EXC2 (execution class for the steel structure)
- EN ISO 3834-2 (welding, production level)
- EN ISO 9606 welder qualification (MAG 135 and related processes)
- ISO 8501 Sa 2.5 surface preparation, EN ISO 12944 C3 to C5 coating
- PN-HD 60364 low-voltage electrical design for the power path
- EN 10204 3.1 mill certificates on traceable steel
- ISO 9001 quality and ISO 14001 environmental management
- Factory acceptance test of the power and cooling plant before dispatch
Grid connection, cooling-medium and any sector-specific certification requirements are confirmed at the specification stage and engineered into the build.
Deployment
Typical use cases
GPU-dense AI training
A dedicated power and cooling module feeds a high-density rack row that a conventional building cannot support, sized to the kW-per-rack target.
Configure
Customisation options
Anything we can integrate, wire, test and seal inside the hall can be specified. Common deviation routes:
air-assisted, cold-plate, direct-liquid, or immersion-ready, with heat rejection to dry coolers or chillers to suit the site.
Lead time
Lead time and delivery
A power and cooling module runs on a 10 to 16 week custom schedule from signed-off drawings to ex works dispatch; simpler combined skids can follow the catalog 6 to 10 week track.
Specification and NDA
within 24 hours of first contact when the brief is sensitive.
Quotation
3 to 10 working
a fixed price and lead time, typically within 3 to 10 working days.
Engineering
power and thermal design to the kW-per-rack target, BOM, weld procedures, customer sign-off gate before steel is cut.
Production and integration
envelope, plant, pipework and controls built and wired indoors.
Factory acceptance
power and cooling tested together before dispatch.
Delivery
EXW Bánovce nad Bebravou as default, DAP and DDP on request, with the as-built dossier on handover.
Proven in service
Reference deployments
The deployments below are described by region and sector only, in line with our anonymised reference policy. They show the kind of work the line is built for; configuration detail is indicative.
Central Europe, AI compute operator
- Dedicated power and cooling module feeding a GPU-dense training rack row
- Cold-plate liquid cooling distribution with redundant pumps and leak detection
- UPS, switchgear and dual feed sized to the rack-row load, monitoring to the operator's DCIM
- Use case: high-density compute brought online without a new building's electrical capacity
The plant, not the compute, is usually the bottleneck. Delivering it as a proven module collapses the timeline.
Western Europe, research and HPC facility
- Cooling module built for direct-liquid loads alongside an existing compute hall
- Heat rejection to dry coolers, valved manifolds for maintenance without downtime
- Environmental and electrical monitoring with remote alerting
- Use case: thermal headroom added to a facility that had run out of cooling capacity
A cooling module lets a facility grow its thermal envelope without rebuilding the room around it.
Benelux, edge inference operator
- Combined power-and-cooling skid placing accelerator capacity close to the data
- Compact UPS and distribution, air-assisted cooling with a liquid-ready provision
- Hardened, unattended, remote monitoring
- Use case: inference at the edge, redeployable as the workload moves
Edge inference rewards a self-contained skid that runs unattended and can be relocated.
FAQ
FAQ
Send a specification. Get a fixed quote.
Engineering review included. Quotation turnaround is 3 to 5 working days for standard configurations.