Data Center Construction in Saudi Arabia and the UAE: Cooling, Power, Tiers and MEP Coordination

Every gigawatt of AI compute announced across the Gulf is also a construction problem: a shell that must go up, a power system that must connect to a grid, and a mechanical system that must reject a heat load most buildings never see. The headlines cover the capital. This guide covers what actually gets engineered before any of it can run.

Why the Gulf Is Building Data Centers at This Scale

The scale of announced AI infrastructure in the Gulf is unusual even by regional giga-project standards. State-backed developers in the UAE are building toward a multi-gigawatt AI campus in Abu Dhabi, while Saudi Arabia’s HUMAIN initiative is pursuing several gigawatts of AI compute capacity through the early 2030s. Sovereign capital, abundant gas and solar resource, and governments able to move permitting and land allocation quickly are the commonly cited reasons the region has become a preferred site for hyperscale and AI-specific data center development.

For engineering and construction teams, the headline number that matters is not gigawatts of announced capacity — it is megawatts per building that must actually be designed, coordinated, permitted and commissioned. That is a fundamentally different exercise from a commercial tower, a hospital or a residential development, and it is the subject of the rest of this guide.

What Makes Data Center Construction Different

A data center is a mission-critical facility first and a building second. The distinction shows up in three ways that change how the project is engineered from day one.

• Continuous operation is the design brief. A data center is designed for zero planned downtime and a defined tolerance for unplanned downtime, not for occupancy comfort. Every system — power, cooling, fire, security — is specified against an availability target rather than a code minimum.
• Redundancy is structural, not incidental. Backup is not an add-on system; it is a parallel path designed in from the concept stage, sized so that a single failure — a transformer, a chiller, a generator — does not interrupt IT load.
• Density drives everything. Power and cooling loads per square metre are an order of magnitude higher than a typical commercial building, and AI-specific facilities running high-density GPU racks push those loads further still, which is reshaping cooling strategy across the industry.

These three factors are why a data center brief reads more like an industrial facility specification than a building specification, and why the engineering team assembled for one needs MEP, electrical and structural depth beyond what a standard commercial fit-out requires.

The Uptime Institute Tier System

Data center resilience is classified using the Uptime Institute’s four-tier system, and the tier target is set before design begins because it drives every downstream engineering decision — how many electrical paths, how much redundant cooling capacity, and how the facility is tested before handover.

Engineering for Extreme Climate

Every data center rejects enormous quantities of heat. In Riyadh, Dubai or Abu Dhabi, that heat is rejected into ambient air that regularly exceeds 45°C in summer — a condition most cooling equipment catalogued for temperate markets is not selected against.

Why Gulf cooling design differs from a standard commercial building

• Chiller selection against extreme ambient. Equipment must be selected and derated against the region’s actual peak dry-bulb and wet-bulb conditions, not a generic design temperature, or capacity shortfalls appear exactly when the grid and the ambient load peak together.
• Water availability constrains cooling strategy. Evaporative and water-cooled systems are efficient but compete with a genuinely scarce regional resource, which is pushing an increasing share of Gulf data center projects toward air-cooled chiller plant, direct liquid cooling at the rack, or hybrid systems that minimise water draw.
• Dust and sand loading affect filtration and airflow. Outdoor air handling and cooling tower components need filtration strategies sized for regional dust loading, which increases plant footprint and maintenance access requirements compared with a temperate-climate reference design.
• High-density AI racks push liquid cooling to the aisle. GPU-dense AI training racks generate heat densities that traditional air cooling struggles to reject efficiently, which is why direct-to-chip and rear-door heat exchanger liquid cooling are appearing more frequently in Gulf AI facility designs than in a conventional enterprise data hall.

This is the same underlying physics that makes ceiling plenums run tight on conventional Gulf towers — see our guide to MEP plans in construction for how cooling load reshapes routing on standard buildings — except in a data center the cooling load is not a comfort load, it is the entire IT load converted to heat, continuously, with no diversity factor to fall back on.

Power Infrastructure and Sequencing

Power is usually the true critical path on a Gulf data center project, not the building shell.

Grid connection lead time

Utility-scale grid connection for a multi-megawatt facility is rarely fast anywhere, and in a region where several giga-scale campuses are competing for the same transmission capacity and substation slots, connection studies and approvals can take longer than the building itself. Confirming the connection point, voltage level and available capacity with the utility is a first-quarter task, not something resolved once design is underway.

On-site substation and transformation

Most facilities at this scale require a dedicated on-site substation stepping utility supply down to distribution voltage before it reaches the data hall’s power distribution units. Our guide to electric substation construction covers the substation types, components and Gulf-specific factors — extreme heat derating, high soil resistivity affecting earthing design — that apply directly to a data center’s on-site power infrastructure.

Backup power and redundancy

• UPS systems bridge the gap between a utility interruption and generator start, sized and configured according to the target tier’s redundancy requirement (N, N+1, 2N).
• Standby generation — diesel or, increasingly, gas-fired — provides sustained backup, with fuel storage and run-time requirements set by the facility’s committed availability, not by code minimums.
• Electrical distribution redundancy mirrors the cooling redundancy strategy: a Tier III or Tier IV facility needs genuinely independent distribution paths, not a single path with a spare breaker.

MEP Coordination and the Role of BIM

If MEP coordination matters on a commercial tower, it is close to non-negotiable on a data center, because the consequence of an uncoordinated clash is not a lowered ceiling — it is a facility that cannot achieve its rated redundancy or, worse, a conflict discovered after the hall is energised.

Why data center coordination is harder than a typical building

• Density of services per square metre is far higher than a commercial floor plate — power distribution, structured cabling containment, fire suppression (often clean-agent rather than sprinkler in the data hall itself), and cooling distribution all compete for the same raised-floor void and overhead space.
• Redundant paths must stay physically separate. A concurrently maintainable or fault-tolerant design requires that parallel electrical and mechanical paths do not share a single point of physical vulnerability — a requirement that has to be verified in the coordinated model, not assumed from the single-line diagram.
• Clash consequences are operational, not cosmetic. A missed clash in an office building might mean a relocated diffuser. A missed clash in a data hall can mean a redundant cooling path that was never actually independent, discovered during commissioning or, worse, during an actual failure event.

The clash detection discipline covered in our guide to BIM clash detection — a defined test matrix, LOD 350 coordination sign-off, and classified, owned issue tracking — applies directly here, with the added requirement that redundancy separation itself becomes a modelled and tested condition, not just a spatial fit check.

Regional Codes and Compliance

A Gulf data center sits at the intersection of building code, electrical utility regulation, fire authority requirements and, increasingly, sustainability frameworks tied to the power draw these facilities represent.

Construction Sequencing and Commissioning

Data center delivery follows a sequence shaped by the fact that power and cooling infrastructure must be tested under full simulated load before the facility can go live — a step that has no real equivalent on a conventional building.

1. Shell and structure. Raised floor systems, structural loading for heavy mechanical and electrical plant, and seismic or wind bracing appropriate to equipment racks are established early, since floor loading capacity is far harder to retrofit than in a typical office building.
2. Power infrastructure. Utility connection, on-site substation and primary distribution installation typically begin as soon as the shell can support them, running in parallel with structural completion rather than waiting for it, given how long grid connection can take.
3. Mechanical plant installation. Chillers, cooling towers or air-cooled plant, and the distribution network to the data hall are installed and pressure-tested ahead of IT fit-out.
4. MEP fit-out and coordination sign-off. Data hall containment, cooling distribution and fire suppression are installed against the coordinated LOD 350 model, with redundancy separation verified physically, not just on paper.
5. Integrated systems testing and commissioning. Every redundant path is tested under simulated full load and simulated failure conditions — a generator start test with a live load bank, a chiller failover test, a UPS transfer test — before the facility is handed over. This phase alone can run for months on a large facility, and compressing it is one of the more common causes of early-life reliability problems.

Frequently Asked Questions

Conclusion

The Gulf’s AI infrastructure boom is, underneath the capital announcements, a construction and engineering programme of a scale and technical difficulty the region has not built at before — extreme-climate cooling design, power infrastructure that is frequently the real critical path, and MEP coordination where a missed clash can mean a redundancy path that was never actually independent. Facilities built well against these constraints will run for decades; facilities that treat a data center brief like a conventional commercial building tend to find out the difference during commissioning, when it is far more expensive to fix.