This is a paid article which your subscription is allowing you to read.

Around the world, the demand for computer power is experiencing an exponential surge driven primarily by the new and growing applications opened up by artificial intelligence (AI). But this has sparked a backlash against the construction of AI data centres, and South Africa is not immune.

NGO activists, some of them based offshore, have appealed to Cape Town mayor Geordin Hill-Lewis to uphold their appeal against a plan to build a 160-megawatt (MW) AI-capable data centre close to the city’s international airport. The facility will be the biggest yet built in South Africa, although it is tiny next to the $500 billion, 10 000MW giant planned by a consortium of hyper-scalers in the United States. The activists complain that the issue is not “an imported moral panic”. But that is exactly the dead horse they are trying to flog.

The Cape Town project is led by the world’s largest digital infrastructure developer, California-based Equinix. South Africa currently has about 1% of global AI data centre capacity, with only five out of its 56 operational data centres explicitly equipped to handle high-density AI workloads. South Africa's 56 data centres account for about 0.5% of total global data centres.

The chart below shows the proportion of data centres that South Africa has compared to selected countries.

The activists zero in on AI data centres’ hunger for both energy and water. AI-capable data centres represent a step change in intensity. More traditional centres, used for cloud storage, are little more than large warehouses. But some processing units in AI factories consume huge volumes of electricity during their high-speed operations. They also generate considerable heat, which can require both air conditioning and liquid cooling.

But activist opposition is misplaced when it refers to South Africa’s shortage of energy, with load-shedding now definitively behind the country, largely because South Africa could afford to throw a fortune at the problem. The very different problem that has emerged over the past two years is what to do with the country’s surplus generation capacity. AI-capable data centres seem like a rather neat solution.

Water use seems, intuitively, to be a more realistic problem in a water-scarce country. But there is every reason to use closed-loop water cooling technology where water is not “lost” but recovered and re-used. In any case, South Africa has an extensive environmental impact assessment system (some say too extensive), which would apply to any development on the scale of a data centre and protect sensitive water sources.

The heart of the environmental activists’ argument is that resources used by data centres will deprive poorer urban residents of much-needed access to water and electricity. This zero-sum argument is as old as the hills and is no more correct now than it ever was. Urban service delivery has to confront all too many (financial, technical, and social) hurdles, and the competing demand of industry is a vanishingly small factor in the overall picture. It is absurd to argue that energy or water used by a data centre in Cape Town has been re-routed from the homes of those living on the urban periphery.

South Africa cannot afford to be left out of the data-centre race. The case for applying AI to business, especially in the services and retail sectors, has emerged with remarkable swiftness. By automating inefficiencies, cutting structural costs, and uncovering new revenue streams, AI allows local enterprises to thrive even in the face of macroeconomic headwinds. In the absence of AI development, business activity and jobs throughout the economy will be adversely affected.

Furthermore, the socioeconomic upside of AI is massive. AI has the potential to act as an equaliser that bypasses legacy infrastructure bottlenecks across critical sectors such as healthcare, education, and public service. For example, AI-powered chest X-ray screening tools are being deployed in rural clinics to detect lung ailments, such as tuberculosis and pneumonia, in seconds, allowing immediate treatment before patients return to their remote villages.

AI excites both our best hopes and deepest fears. The Common Sense leans strongly towards the more optimistic perspective. But The Common Sense will also insist on calling out the precious middle-class concerns that, if acted on, will ultimately harm the poorer elements in society.