AI data centers: what Africa should learn from the water crisis in the US and Europe
September 23rd 2026
The meteoric rise of artificial intelligence has a hidden cost: water. While digital giants are multiplying mega-data centers against a backdrop of record droughts in America and Europe, Africa, also courted by hyperscalers, is faced with a choice: repeat the same mistakes or anticipate them.
In Cape Town, South Africa, a hyperscale data center project led by Equinix is currently suspended following an appeal against its approval. The Housing Assembly, a South African organization specializing in housing rights, and the British group Foxglove are challenging the decision of the Municipal Planning Tribunal, which had authorized the use of the King Air Industria site for the project. The appeal was filed on August 6th, with the support of the Legal Resources Centre.
The applicant organizations argue, in particular, that the municipal authorities did not have sufficient information at the time of approval regarding the future site’s water and electricity needs, as well as its environmental and socio-economic impacts. They also criticize the process for not adequately considering the cumulative effect of the two data centers planned for the same area.
The issue of water is particularly sensitive. The project is slated to be built near several residential areas, whose communities already face difficulties accessing regular water and electricity. Furthermore, the project comes at a time when the city narrowly avoided a “Day Zero” crisis in 2018, when Cape Town’s water reserves reached critical levels.
This controversy is part of a debate that has been intensifying for several years in the major global data center markets, particularly in the United States and Europe, where the pressure exerted on water and electricity is now fueling questions that go beyond purely technological issues.
A global concern
On June 3, the United Nations University and the Institute for Water, Environment and Health published a report entitled ” Environmental cost of AI’s energy use: Carbon, Water and Land Footprints “, warning that water and energy consumption, as well as pollution from data centers, will double in the next four years, due to the growth in the use of artificial intelligence (AI).
A data center consumes water in two ways. Either directly, to cool the servers that heat up under the AI computing load, or indirectly, through the water consumed by the power plants required to run them. The United Nations estimates that global electricity consumption by data centers will reach 448 terawatt-hours in 2025 (1 terawatt is equivalent to 1 trillion watts). Producing this energy requires approximately 4.5 trillion liters of water, the equivalent of 1.8 million Olympic-sized swimming pools.
The report predicts that by 2030, global electricity demand from data centers will exceed 945 terawatt-hours, with associated water consumption reaching 9,300 billion liters, equivalent to the annual essential drinking water needs of 1.3 billion people in sub-Saharan Africa.
The five largest hyperscalers—Amazon, Microsoft, Google, Meta, and Oracle—are preparing to allocate over $750 billion to investments in 2026, a 67% year-over-year increase, with the bulk of it earmarked for AI-related infrastructure. This rush is transforming entire regions, sometimes at the expense of communities whose groundwater and rivers are being put under pressure by industrial neighbors whose true consumption they are often unaware of.
United States: Energy-hungry data centers
The American case clearly illustrates the difficulty of regulating such a rapidly expanding sector. In 2025 in Georgia, and in 2026 in Arizona, data center projects were accused of having drawn far more water than they had declared or paid for.
The figures are staggering. US data centers directly consumed approximately 17.4 billion gallons of water in 2023 for cooling, a figure expected to increase significantly by 2028. Indirect water consumption related to electricity production was much higher, at approximately 211 billion additional gallons.
This pressure comes on top of the droughts that are already severely impacting American agriculture. The 2026-2027 winter wheat harvest is expected to be the smallest since 1965, while Lakes Mead and Powell reached historically low levels this summer.
Faced with the controversy, some operators are responding. Amazon Web Services developed a liquid cooling system adaptable to existing facilities in just eleven months and already uses treated wastewater rather than potable water to cool several of its sites. Meta, for its part, claims a 60% improvement in water efficiency at its Mesa, Arizona site and funds restoration projects that deliver more than 200 million gallons of water annually to the Colorado and Salt River basins.
Europe: Moratoriums and contested confidentiality
The Old Continent is not immune to this tension, under the combined pressure of water and electricity shortages. Since early 2024, the Netherlands has banned the construction of new hyperscale data centers, with the exception of a few sites. In Ireland, where sixteen of the world’s twenty largest technology companies have facilities, a moratorium introduced in 2021 suspended the granting of grid connection permits for new data centers in the Greater Dublin area. This was replaced in 2025 by the stricter Large Energy Users (LEU) Connection Policy.
In Ireland, data centers were already consuming 22% of the country’s electricity in 2024, which increased the cumulative bill for households by 360 euros between 2015 and 2023.
The issue of transparency has also become explosive. A 2026 investigation by a consortium of journalists revealed that Microsoft and the tech industry lobby had obtained the right from the European Union to keep environmental data from their data centers confidential, fueling distrust among citizens and local regulators.
Africa: The boom begins, but the same warning signs are already appearing.
This is where the African trajectory becomes instructive. The continent, and particularly sub-Saharan Africa, faces a structural water crisis. More than a third of African countries, representing a population of over half a billion people, are considered water insecure by the United Nations. In sub-Saharan Africa, the scarcity is primarily “economic”: a lack of water supply and treatment infrastructure, inadequate management, and limited funding, despite sometimes abundant natural resources.
It is in this context that investments aimed at making Africa a new data center hub are accelerating. According to the consulting and market research firm Arizton Advisory & Intelligence, investments in data centers in Africa are expected to reach $8.76 billion by 2031, driven by improved connectivity, the adoption of cloud and artificial intelligence (AI) in various sectors, increased demand for local data processing, and growing investments in renewable energy and submarine cables.
The threat to water is therefore no longer a distant problem. In South Africa, which alone has 55 existing data centers, voices are now calling for a moratorium on the construction of new sites until the use of water, land and electricity resources has been seriously assessed.
The South African scenario is not unique. In Nigeria, the data center market is expected to more than double by 2031, driven by cloud adoption and regulatory requirements pushing for local data hosting. But local observers are clear: water and electricity, not just financial investment, will determine which projects can actually thrive.
However, some signs are encouraging. In Kenya, the G42 group and Microsoft announced an investment in the construction of a data center powered by renewable geothermal energy and incorporating, from its design stage, water conservation technologies touted as being at the forefront of the industry. This approach is still too rarely seen in the first generations of Western data centers, built before the issue of water became a prominent topic in public debate.
Three lessons to avoid repeating history
From this parallel between two continents mature in their digitization and a continent that is just beginning its own, three lessons clearly emerge.
First, regulation must precede construction, not follow it. In the United States, authorities have had to intervene after the fact in response to poorly designed permits and undeclared extractions. South Africa, on the other hand, is now testing a more cautious approach before granting new permits.
Furthermore, transparency must be non-negotiable. The case of the confidentiality obtained by Microsoft from the European Union demonstrates that, in the absence of a legal obligation to report on water consumption, operators do not spontaneously publish this data. This is a trap that African regulators would be wise to avoid from the outset by directly including this obligation in the conditions for granting permits.
Finally, water-efficient technologies must be required from the design stage, not added as an afterthought in a rush, as Amazon had to do when it developed an adaptable liquid cooling system in just a few months. Kenya’s choice, relying on geothermal energy and an architecture designed to conserve water from day one, outlines a possible path: that of a continent negotiating its entry into the AI era not by hosting the cheapest infrastructure, but by setting its own conditions.
Back