World

Johannesburg is becoming one of Africa’s most important data-centre hubs. That is good news for South Africa’s digital economy. But billions in investment should not end the conversation. They should start one: how much electricity, water, land and public infrastructure will these facilities consume — and who ultimately pays?

Johannesburg is fast becoming one of Africa’s digital infrastructure capitals. The city and wider Gauteng region are attracting some of the world’s biggest cloud, technology and infrastructure companies. The reason is straightforward: Johannesburg is South Africa’s commercial and financial centre, with extensive fibre networks, major internet exchange infrastructure and a concentration of banks, telecoms operators, businesses and public institutions that increasingly depend on cloud services. That is an economic opportunity South Africa should welcome.

Data centres underpin digital banking, e-commerce, artificial intelligence, government systems, streaming, cybersecurity and telecommunications. Hosting more computing infrastructure locally can improve resilience, reduce latency and strengthen South Africa’s position as a digital gateway into Africa.

But data centres are not just buildings full of computers. They are large, concentrated infrastructure consumers. They need electricity around the clock. They need high-capacity connections, substations, backup generation and well-serviced land. Depending on how they are designed, they may also require substantial water for cooling.

And AI is raising the stakes.

AI is changing the power equation

The new generation of AI infrastructure packs enormous computing power into increasingly dense facilities. That means more electricity.

Global demand from data centres is expected to rise sharply this decade as cloud computing and AI expand. For Johannesburg, the question is not simply whether South Africa has enough generating capacity today. These facilities may operate for decades. The real test is whether the electricity network — generation, transmission, substations and municipal distribution — can absorb large new users without pushing costs onto everyone else.

If a hyperscale facility requires a new substation, who pays for it? If the surrounding electricity network must be strengthened, does the developer cover the cost or does it eventually appear in tariffs? And if additional generation or storage is needed to support the load, should that form part of the investment obligation?

These are not hypothetical questions. Other countries have already discovered what happens when data-centre expansion runs ahead of infrastructure planning.

Ireland learned the hard way

Ireland enthusiastically attracted the world’s largest technology companies and became a major European data-centre hub. Then electricity demand became impossible to ignore. Data centres grew into a significant part of the country’s power consumption, creating grid pressures and forcing regulators to rethink how new facilities should be connected. Ireland did not reject the industry. It changed the rules.

New developments face greater scrutiny over where they are built, their effect on the grid, whether additional generation or storage accompanies them, and the wider economic value they create. That is the lesson South Africa should absorb early. Once data centres reach a certain scale, they stop being ordinary property developments. They become energy-policy decisions.

Singapore asks: value per megawatt

Singapore went further. With scarce land and limited energy resources, it temporarily constrained data-centre development and later reopened the market more selectively. The crucial shift was philosophical. The question was no longer simply: How much money is being invested? It became: How much economic and strategic value does the country receive for the energy and land being allocated? That is a better question for Johannesburg too.

A billion-rand project sounds impressive. But if it consumes enormous amounts of scarce infrastructure while creating relatively few permanent jobs and importing most of its equipment, the headline number tells only part of the story. Investment should be measured against what remains in the local economy.

Land is not free simply because an investor buys it

Data centres also compete for prime industrial land. The best sites tend to have exactly the infrastructure cities struggle to provide: strong electricity connections, fibre, water, roads and proximity to economic centres. That land has alternatives. It could support manufacturing, logistics, housing or other commercial activity. The Netherlands eventually restricted the location of very large hyperscale facilities partly because government concluded that land and energy were too important to allocate without a wider strategic test.

Johannesburg need not copy the Dutch approach. But it should copy the question: Is this the highest-value use of scarce, infrastructure-rich land?

Then there is water

For Johannesburg, water deserves particular scrutiny. Residents already live with outages, ageing infrastructure, losses and supply constraints. In that environment, the water demands of large new developments cannot be treated as an afterthought.

Not every data centre is a water guzzler. Cooling designs differ significantly. Some use evaporative cooling; others rely on closed-loop or air-cooled systems with far lower water consumption. That is precisely why disclosure matters. For every major facility, the public should know how much water will be required, whether it will be potable water, what cooling technology will be used and what happens during droughts or supply interruptions. A promise that a facility is “water efficient” is not enough. Show the numbers.

What about the jobs?

The same principle should apply to employment. Data centres can create large numbers of jobs during construction. Engineers, electricians, builders, fibre contractors, security companies and equipment suppliers all benefit. But once construction ends, the permanent workforce can be relatively small compared with the billions invested. That does not make the investment undesirable. Data centres can support broader digital ecosystems and create valuable technical jobs. But the numbers should not be blurred.

How many jobs are temporary? How many are permanent? How many are South African? How many apprentices, graduates and technicians will be trained? And how much of the procurement actually goes to local companies rather than imported servers, cooling systems and electrical equipment?

A project should not qualify as transformative simply because the capital expenditure is large. The better question is: How much capability does it leave behind?

South Africa absolutely needs data centres

This is not an argument against the industry. Quite the opposite. South Africa cannot build a serious AI, cloud, fintech or digital-services economy without world-class data-centre infrastructure. 

Johannesburg should compete aggressively for that investment. But competing for investment does not mean accepting it on any terms. South Africa should want the most efficient facilities, the strongest skills commitments, the deepest local supply chains and the greatest economic value per megawatt consumed. That is what strategic investment policy looks like.

Publish the numbers

Large data-centre developments should therefore come with a basic public-interest disclosure.

A national register of major data centres would help too. One project may look manageable on its own. Ten large projects drawing electricity and water from the same constrained system are a different matter. Infrastructure planning must consider the cumulative load.

Follow the infrastructure bill

The most important question may ultimately be the simplest: Who pays?

When a private development requires major public infrastructure, somebody carries the cost. If the investor pays, say so. If a municipality, electricity utility or water provider is expected to contribute, disclose that too. Because there is a real danger in celebrating a R10 billion private investment while quietly socialising part of the infrastructure bill.

Johannesburg already has communities waiting for reliable electricity, water, roads and basic municipal services. Those residents should not unknowingly subsidise infrastructure for some of the world's richest technology companies.

Make public value part of the deal

Johannesburg should want to become Africa’s leading data-centre city. But leadership should mean more than hosting the most servers. It should mean building the continent’s smartest framework for digital infrastructure. Every major project should face a simple test: What does South Africa get for the megawatts, water, land and infrastructure it gives up?

Are local companies gaining business? Are South Africans gaining skills? Are permanent jobs being created? Is new energy capacity being added? Are infrastructure costs being carried fairly? And can the public see the answers? The choice is not between data centres and development.