The firm
Nourra Holdings is an AI data centre operator headquartered in Singapore. We take a facility through its whole life: securing land and power, integrating the mechanical and electrical supply chain, building it, and running and maintaining it over the long term.
- 01Land and powerPowered buildings · land · grid connection
- 02M&E supply chainPower · liquid cooling · modules
- 03ConstructionBuilding · M&E · racking
- 04Operations and maintenanceOn-site teams · inspection · spares
Value chain
Ten layers sit between a unit of electricity and a single model call.
Open each layer to see what it is and where Nourra sits.
Models and applications
Model training, inference services and enterprise AI applications. They decide how much compute is needed, where, and when.
The layer our clients and partners work in.
Compute services and cloud
Hardware turned into something you can rent: bare metal, GPU cloud, and model APIs billed by usage.
We arrange verified GPU capacity and enterprise cloud for partners.
Operations
Life after go-live: staffing, inspections and fault response, plus scheduled servicing, repair and spare parts for both facility and IT equipment. Availability is defended at this layer.
Our own on-site teams run operations and maintenance against a service level agreement.
Cluster deployment
Racking, cabling, power-up and commissioning once hardware arrives. A GPU cluster has thousands of cables, and one wrong run means rework.
Our own engineering crews rack, cable and commission.
Networking and storage
The high-speed fabric that makes hundreds of servers behave as one machine, and the storage that holds training data.
Specialist engineers are assigned per project and named in the plan.
Chips and servers
GPUs, servers and full-rack systems. Supply is tight and lead times are long. This is the scarcest link in the chain.
We verify supply and lead times. Hardware goes straight into our own facilities.
Cooling
Carrying heat out of the rack. The newest full-rack GPU systems produce more heat than air can remove, so liquid runs directly across the chips.
Designed and built to high-density standards in our own facilities.
M&E equipment and prefabricated modules
Transformers, switchgear, uninterruptible power, liquid-cooling equipment and prefabricated data hall modules. Their lead times and quality decide how soon a facility is built and how long it lasts.
We integrate manufacturers and local production, then procure, assemble and deliver to site as one package.
Building, mechanical and electrical
Structure, floor loading, transformers and switchgear, uninterruptible power and fire protection. A facility’s ceiling is set here.
We build our own data centres. Our crews handle electrical works and conversions.
Power and land
Substation capacity, power purchase and grid connection dates. How much power, and how soon, decides how large a facility can be and how fast it goes live.
We source powered buildings and land, and check power capacity, permits and power-on dates.
Inside the facility
An AI data centre does two things: it gets power into the rack and takes heat out.
Switch between the power and cooling paths, and select a component to read about it.
- Power
- Coolant supply
- Coolant return
Grid connection
The facility draws high-voltage power from the grid. How much capacity is available, and whether there are two independent feeds, is the first thing to confirm when choosing a site.
Transformers
They step high voltage down to what equipment can use. Total transformer capacity is, in practice, the ceiling of the facility.
Uninterruptible power
Batteries carry the load the instant the grid drops, until backup generators start. A GPU training run that loses power loses days of computation.
Power distribution
Power is split to every row and every rack. High-density racks need heavier busbars and a higher distribution rating.
Racks and GPU servers
Where the compute is. One current full-rack GPU system draws as much as a dozen or more conventional racks used to.
Coolant distribution unit
The heart of a liquid-cooled hall. It sends coolant to the cold plates in every server and brings the heated liquid back.
Heat rejection
Where the heat finally leaves the building. In a tropical climate like Singapore’s, this stage has a direct effect on the power bill.
Density
What sets an AI data centre apart is the power in a single rack.
Drag the slider to see how rack count and cooling change for the same 1 MW of power.
- Racks powered by 1 MW
- 25
- Type of facility
- High-density hall
- Cooling
- Enhanced air or rear-door heat exchange
Ranges are typical industry values, shown for illustration. Project figures follow the design documents.
Supply chain
How fast a facility gets built depends, as much as anything, on when the equipment arrives.
We procure and deliver power distribution, liquid cooling and prefabricated modules as one package.
- Power equipment
- Liquid cooling
- Prefab modules
Specification
Rack density and site conditions set the power architecture, cooling method and module specification. The equipment list is fixed alongside the engineering drawings.
We turn site conditions and client requirements into one complete equipment list.
Manufacture and prefabrication
Switchgear, liquid-cooling equipment and data hall modules are built, pre-assembled and factory-tested, which cuts the work left for site.
We integrate a global supply chain with local production in Southeast Asia, and manage schedule and quality as one.
Logistics and customs
Cross-border transport, customs clearance and arrival dates for heavy equipment have to line up with construction progress on site.
We coordinate shipping and arrival sequence, under applicable trade and export control rules.
Installation and commissioning
Modules are set in place, connected to power and pipework, and commissioned with the rest of the facility.
Installed by our own engineering crews and commissioned together with the manufacturer.
Maintenance and spares
Scheduled servicing, repair and spare parts once the equipment is in operation.
We set up spares stock and a maintenance plan, carried out by the on-site team.
Business
Six lines of business cover every stage of a facility, from site to operation.
Site and power
We source powered buildings and developable land, and check zoning, floor loading, power capacity, permits and power-on dates.
- Powered buildings
- Land assessment
- Grid connection
Development
Mainly small and mid-sized AI data centres of a few megawatts, expandable in phases and handed over ready for hardware to be racked.
- Building conversion
- Modular data halls
- High density
- Phased expansion
Supply chain
Power distribution, liquid cooling and prefabricated modules procured, integrated and delivered as one package, combining a global supply chain with local production.
- Power equipment
- Liquid cooling
- Prefabricated modules
- Spare parts
Engineering
Our own crews deliver electrical and low-voltage works, cabling, racking and facility conversion, and join larger projects as a specialist subcontractor.
- Electrical and low-voltage
- Cabling and racking
- Specialist subcontracting
Operations and maintenance
On-site teams run day-to-day operations, inspections and fault response, and handle scheduled servicing, repair and spare parts for facility and IT equipment.
- On-site staffing
- Inspection and fault response
- Scheduled servicing
- Repair and spares
Compute
Verified GPU capacity and enterprise cloud for our partners, deployed in our own facilities.
- GPU capacity
- Server procurement
- Enterprise cloud
Footprint
Based in Singapore, connected across Asia-Pacific, North America and Europe.
- Headquarters
- 1
- Project countries
- 2
- Partner countries
- 4
Lines show business links from our Singapore headquarters, not physical network routes. Markers sit on capitals or major cities.
Headquarters
Data centre projects
Partner network
Delivery
A cluster passes six gates between site selection and go-live.
- Gate 1
Site and power
Confirm land use, floor loading, ceiling height and available power. Until power is secured, nothing after it stands.
Nourra’s partSite survey, and checks on power capacity, permits and power-on dates.
- Gate 2
Design and approvals
Fix rack density, the power architecture and the cooling scheme, then complete drawings and approvals.
Nourra’s partWe organise design and coordinate approvals.
- Gate 3
Building and M&E
Structural strengthening, transformers and switchgear, uninterruptible power, cooling pipework and fire protection.
Nourra’s partBuilt by our own engineering crews and accepted at milestones.
- Gate 4
Arrival and racking
Servers, network equipment and full-rack systems arrive on site to be unpacked, racked, cabled and powered.
Nourra’s partWe verify supply and arrival dates. Our own crews rack and cable.
- Gate 5
Commissioning
Network connectivity, full-load running and thermal validation. Problems found here cost far less than problems found after go-live.
Nourra’s partWe commission alongside specialist engineers and issue test records.
- Gate 6
Handover and operations
The cluster is handed to its user and enters long-term operation.
Nourra’s partOur on-site team operates it against a service level agreement.
Principles
Three principles we work by.
End to end
From site selection to maintenance, the critical work is done by our own people. Responsibility stays in one place.
Compliance
All business follows applicable export control and sanctions law. We confirm end user, location and use before we proceed.
Long-term partners
We work with a limited number of partners over the long term, on written terms.