نمو بنية تحتية الذكاء الاصطناعي يعيد كتابة قواعد تصميم مراكز البيانات، حسب مدير Equinix في منطقة الشرق الأوسط وشمال أفريقيا

Equinix’s Managing Director for MENA, Kamel Al Tawil, warns that the rapid expansion of AI and GPU‑intensive workloads is forcing a fundamental redesign of data centre infrastructure across the UAE and the wider Middle East. He highlights stark figures — data centres in the UAE consumed 3 TWh in 2025, roughly 2% of the country’s 173 TWh total electricity demand — and says traditional air‑cooled architectures, built for 5–10 kilowatts per rack, are being outpaced by modern GPU clusters that can exceed 200 kilowatts per rack and are forecast to reach up to 1 megawatt.

“Systems must be designed with coolant distribution units that manage heat exchange between facility chillers and server racks while maintaining redundancy to mitigate unplanned cooling outages,” Al Tawil said, stressing that liquid cooling should be considered at the design stage rather than as a retrofit.

Why liquid cooling matters now

Al Tawil argues that direct‑to‑chip liquid cooling (DLC) has emerged as the dominant approach for supporting high‑density AI infrastructure because it delivers coolant directly to high‑heat components, improving thermal efficiency and reducing energy consumption tied to fan power. He notes that advances in piping technology have lowered the operational risk of liquid systems, making leaks a very low probability event, and enable consolidation of more compute into fewer racks — a factor that reduces capital and operational expenditure over lifecycle replacements.

Equinix has responded by designing new facilities with liquid cooling in mind. According to Al Tawil, integrating cooling strategy into the overall AI and infrastructure plan is critical because early design choices materially affect an organisation’s ability to scale high‑density AI workloads later. Retrofitting air‑cooled data centres for DLC, he adds, can be “complex, costly, and higher risk.”

Operational and regulatory considerations

• Power density and consumption: Modern GPUs drive a rapid increase in rack power, requiring facility power systems and distribution to be scaled accordingly.
• Cooling integration: Close alignment is needed across customers, OEMs and colocation providers on temperature, pressure and flow specifications.
• Redundancy and resilience: Coolant distribution units and redundancy planning are essential to mitigate unplanned cooling outages.
• Sustainability and regulation: Rising regulatory requirements around efficiency and emissions are forcing operators to design for compliance while remaining flexible to evolving workloads.

Al Tawil emphasises that liquid cooling not only addresses performance and energy concerns but also contributes to sustainability goals. He points out that DLC enables dematerialisation, wider operating temperature ranges, improved resource efficiency and heat reuse opportunities — all of which can lower total energy intensity and extend hardware lifespan by reducing thermal stress.

Outlook for the UAE and the region

Looking ahead, Al Tawil says advanced cooling technologies will be essential as the UAE scales high‑density compute to support national AI ambitions. Liquid cooling, he suggests, will reduce reliance on energy‑intensive air systems, enable more efficient heat management and support higher‑power AI deployments. For operators and colocation providers, the imperative is clear: design with high‑density cooling in mind from day one, align the full ecosystem on technical specifications, and build redundancy into coolant distribution to safeguard mission‑critical environments as AI workloads continue to grow.