![]() |
Data centres near cooling crisis as AI demand surgesArtificial intelligence is transforming industries, but it’s also testing the limits of the infrastructure that keeps the digital economy running. Across the world, and increasingly in Africa, data centres are consuming more electricity than ever before, and cooling has become the silent bottleneck. ![]() Source: Unsplash Inside a modern data centre, cooling can account for as little as 7% of power use in advanced hyperscale sites, but well over 30% in smaller or less-optimised facilities. As AI workloads drive unprecedented power densities, with racks drawing 20kW to 100kW or more, cooling is no longer a secondary concern; it’s the frontline constraint. Why this matters for AfricaIn Africa, the challenge is amplified by two factors: heat and grid instability. South Africa, the continent’s largest data centre market, already faces high ambient temperatures and a carbon-heavy, constrained electricity system. Add AI-scale computing to that mix, and the result is a perfect storm of energy intensity, cooling demand, and infrastructure strain. Per-capita data centre electricity use on the continent is still below 1kWh per person, but it is projected to double by the end of the decade. South Africa will lead that surge as the regional hub for cloud, hyperscaler, and sovereign AI projects. Yet, without intervention, cooling inefficiency could become a limiting factor in Africa’s digital growth story. Air cooling can’t keep upAir cooling has served the industry for decades because it’s simple and cost-effective at lower power densities. But it’s reaching its physical limits. The hotter the chips get, nd today’s AI-optimised processors exceed 600 watts each, the harder it becomes to remove heat using air alone. In practical terms, that means higher energy costs, greater risk of thermal throttling, and reduced operational reliability. For African operators already managing load-shedding or high electricity tariffs, this is not sustainable. Liquid cooling is no longer optionalTo support AI-driven workloads, the industry must pivot decisively toward liquid cooling, whether direct-to-chip or immersion-based. Liquid moves heat far more efficiently than air, enabling denser computing while lowering total energy use. It’s the technology that will allow Africa’s data centres to scale sustainably. At the same time, operators must rethink how power is sourced. Co-locating solar and battery systems with data centres can ease pressure on national grids, improve resilience, and lower emissions. South Africa’s solar boom creates a real opportunity here, if policy keeps pace. Collaboration and policy support are keyAddressing this challenge isn’t just a technical issue; it’s an ecosystem issue. Governments, utilities, and private operators need to work together to create clear frameworks for renewable integration, energy benchmarks, and data centre efficiency standards. Incentivising early adoption of liquid cooling and hybrid power solutions can de-risk the transition and attract investment. If we fail to act, Africa risks higher costs, reduced competitiveness, and missed opportunities in the global AI economy. But with foresight and coordination, we can turn this challenge into an advantage, positioning the continent as a leader in sustainable, next-generation digital infrastructure. About the authorGeneral manager - Africa - Lenovo Infrastructure |