In May this year, the South African Human Rights Commission (SAHRC) opened submissions on the question of building data centres in the country. On the face of it, much of the SAHRC’s interest falls under what might be called “slow-growth” politics: discussions around the localised environmental, social, and economic impacts that arise whenever a large-scale industrial development is proposed.

However, submissions to the commission reveal intense interest from non-governmental organisations (NGOs) and activist groups with broader, even international, agendas. These include Islamist organisations as well as groups funded by large progressive United States (US)-based foundations – part of what is sometimes called “Big Philanthropy”. Their presence points to a wider politics playing out in the debate.

In part, the South African debate is a product of what is rapidly emerging as something between the space race and the programme to build the first atomic bomb. This competition is taking place primarily between the US and China as each tries to gain the advantage in artificial intelligence (AI). These technologies are likely to be major economic drivers of the future, but will also be central to the military strength of countries seeking to challenge the US for global hegemony. There have, for instance, already been allegations of China using AI technology to surveil domestic populations, including the Uyghurs, and against foreign adversaries.

Amid public discussions about how AI will affect the workforce, intellectual property, and whether AI videos of US President Donald Trump in episodes of Seinfeld are actually funny sits the physical backbone of the technology: AI data centres containing thousands of processors used to store and process information and train AI models.

Big Business

In both the US and China, data centres have become big business, with hundreds of facilities built or proposed in recent years. This has sparked debates about where they should be built and their costs and benefits. In the US, despite early enthusiasm among states to attract them, public opinion has increasingly soured. Even typically pro-development Republicans in states such as Texas have begun scaling back their open-ended endorsement.

There have also been claims that China has sought to influence the American data-centre debate, particularly online, an allegation considered serious enough to attract congressional attention. Such interference would make strategic sense: if public pressure prevents the infrastructure required for AI from being built in America, China would be an obvious beneficiary. However, any Chinese propaganda is likely about exploiting existing fissures in American society, including traditional left/right divides such as progressive concerns about climate change.

Whether by design or simply because they are plugged into progressive American debates, it is these US-funded NGOs that have been important in discussing AI data centres in South Africa using substantially the same arguments as their American counterparts.

Underlying this is a deep public anxiety about the philosophical, economic, and ethical implications of AI. This has turned what should also be a practical discussion about infrastructure into a political and geopolitical one. However, once the rhetoric is stripped away, the objections boil down to three main areas: governance and oversight; pressure on resources, including land, water, and electricity; and economic benefits. Simply transplanting the American debate into South Africa does not translate neatly into a developing world context.

Electricity

Take electricity. In the US, one objection is that enormous data-centre electricity consumption can drive up local energy prices. That is not how South Africa’s electricity system works, since we have set tariffs. The concern here is instead that Eskom, having secured an extended period without load-shedding, could be pushed back into shortages by rapidly growing data-centre demand. The related, more ideological concern is that Eskom might start once again relying on coal-fired generation to supply these facilities, undermining a vision of a renewables-based energy future.

The opposing view is that data centres could help get Eskom back on its feet. The recent absence of load-shedding arguably reflects not only improved generation performance, but also weak economic growth, increased private generation, and electricity prices that have made some large industrial customers uncompetitive. Data centres could offer Eskom a new source of large, reliable industrial demand.

Water consumption is probably the strongest resource-related objection. A large data centre can use quantities of water comparable to a golf course, depending on its cooling technology and operating conditions.

South African policymakers have never taken seriously enough the fact that the country is, in many respects, a “desert in disguise”. But our water crisis is also about inefficient agricultural use, enormous losses through leaking infrastructure, and the widespread use of drinking-grade potable water for industrial purposes, while municipalities fail to maintain sewage and grey-water systems.

Data centres can recycle water and reduce their consumption. More importantly, their infrastructure investment could provide an opportunity to expand grey-water systems so that drinking water is not unnecessarily used for industrial purposes. Policymakers should ask whether this investment can be leveraged to improve the system itself.

Cape Town-centric

Lastly, there is the Cape Town-centric discussion around land use and affordable housing. Most South African data centres, however, are located in Gauteng, with a major cluster around Midrand and others in outlying urban areas, central business districts, and older manufacturing districts. Much of this land is well suited to data centres: it is relatively inexpensive, often close to power and fibre infrastructure, and its redevelopment could help revitalise neglected urban and industrial spaces near economically depressed communities.

NGOs are also importing a particularly Western argument that data centres do not create the number of jobs their proponents promise. They argue that the thousands of jobs associated with these projects are largely temporary construction positions, while permanent employment may number only in the hundreds.

That argument has considerably more force in an American context, where governments may offer substantial tax incentives to attract data centres partly on the promise of employment creation. It does not hold easily in South Africa.

Given our catastrophically high unemployment rate, I would venture to argue that there is no such thing as “just” a temporary construction, artisan, or electrician job in South Africa. There is instead a worker and a fed family for the roughly 18 to 36 months it can take to build one of these facilities. A sustained pipeline of data-centre construction could turn supposedly temporary employment into years of demand for skilled artisan trades near communities that need the work.

As for permanent employment, the industry is already complaining that South Africa does not have enough skilled people to fill existing data-centre positions, with companies reportedly poaching staff from one another. At least one company has responded by establishing its own training facility to bridge the skills gap. The employment question should therefore not simply be about the number of permanent jobs, but whether South Africa can use the industry to build a larger pool of technical skills and make South Africa into the continent's AI hub.

Resource constraints associated with large industrial projects are real. South Africa already has development protocols, such as environmental impact assessments, that should be used to scrutinise how these projects are undertaken.

The suggestion by some NGOs that they are not opposed to data centres but merely want a “pause” while a new regulatory framework is developed seems questionable. Given South Africa’s well-known regulatory lethargy, a temporary pause could easily become a permanent moratorium.

The better approach is not to stop data-centre development, but to use the mechanisms we already have to address legitimate concerns around electricity, water, land, and infrastructure, while allowing the government to come to grips with this new industry and allowing South Africans to participate in one of the most important technological and economic developments of the coming decades.