Concerns and questions are arising after the Cape Town Municipal Planning Tribunal (MPT) approved a zoning change application for the construction of a hyperscale data center in King Air Industria.
As the use of artificial intelligence rapidly grows worldwide, its environmental cost is causing serious alarm. The public is questioning how much water these centers consume and whether this warrants concern.
Approximately 519 milliliters of water are consumed when processing a single AI query, such as interacting with ChatGPT. Although this seems like a negligible amount at first glance, considering the billions of queries processed every minute globally, it amounts to billions of liters per minute.
In 2020, data centers accounted for about 1–2% of the total global electricity demand. Since advanced processing chips generate enormous amounts of heat, facilities are forced to use powerful cooling systems to prevent equipment damage.
According to Uptime Institute data, a relatively small 1-megawatt data center can consume 26 million liters of water annually solely for cooling purposes. Meanwhile, the number of massive 'hyperscale' centers is rapidly increasing. By 2020, there were nearly 600 such facilities worldwide, with China accounting for 10% of the global volume and Japan for 6%.
The actual water footprint of data centers extends far beyond the water supplied to local cooling towers. The overall water footprint of a facility also includes 'indirect' water consumed by power plants generating electricity, as well as water needed for hardware manufacturing. Power plants running on fossil fuels and nuclear energy use large boilers with water that is superheated into steam to drive turbines, leading to the evaporation of huge volumes of water just to keep servers running.
Furthermore, microchips inside servers require highly purified 'ultra-pure' water during production to remove silicon residues. According to the World Economic Forum, an average chip fabrication plant consumes about 37.8 million liters of ultra-pure water daily.
The significant strain on local resources is causing infrastructure crises and negative reactions in several countries. In Ireland, data centers consumed approximately the same amount of electricity as the entire country in 2023. Because the national grid is nearly exhausted, some facilities in Ireland are forced to use high-polluting off-grid generators to maintain operation.
In the UK, Thames Water noted that 80% of the country's data centers are concentrated in the London area, which has unexpectedly low annual rainfall. Since data center water consumption peaks during the hottest and driest periods of the year, the utility company has been forced to restrict or object to new data center projects to protect regional water reserves.
Similar concerns over resource scarcity led to a temporary ban on new data center construction in the Netherlands, while France is in the process of adopting new laws requiring greater transparency regarding water consumption.
Since freshwater makes up only 3% of Earth's water, the technology industry is actively seeking ways to mitigate this crisis. Regarding hardware, Microsoft has conducted experiments placing data centers in cold waters off the coast of Scotland for natural reduction in cooling requirements. Other facilities are transitioning to 'immersion cooling,' where servers are submerged in specialized synthetic, non-conductive fluids instead of using evaporative water cooling.
Technology companies are also investing heavily in international water management projects to replenish the basins from which they draw water. For example, Google funds wetland restoration in the Andes in Chile, supports rainwater harvesting and the construction of dual dams in the Netherlands, and improves water quality at the Singapore Botanic Gardens.
However, experts warn that technological solutions alone will not be enough. Global communities will need to implement integrated land and water resource planning to carefully weigh the economic benefits of the AI boom against its profound impact on our most valuable and finite resource.
In a statement published on Thursday, the City of Cape Town announced that the MPT approved the application, but this decision is subject to several conditions. The City added that official notification of the decision will be issued as soon as possible.
The City stated that anyone who submitted a valid objection during the established public participation process will receive information about the decision and how to appeal. After an appeal is filed, it will be reviewed according to the City's established appeals process before a final decision is made. If no appeal is filed, development cannot commence immediately.
The City emphasized that the applicant must still obtain additional permits, including the submission and approval of a Site Development Plan. As part of this process, relevant City departments must assess detailed technical issues, including the provision of engineering services and infrastructure, before further permits are issued. The City acknowledged concerns regarding water consumption, electricity demand, and environmental impact, stating that all of this 'is part of the planning and regulatory processes and is considered in accordance with applicable legislation and technical requirements.'
The escalating water crisis in South Africa is attributed not so much to a lack of infrastructure, but rather to weak governance, inefficient financial management, and the inability of municipalities to generate revenue.
Shirley Robinson, head of DNA Economics' practice on climate and public economics, emphasized that financing municipal water infrastructure is primarily an issue of municipal finance. She shared these thoughts during the webinar 'Beyond Infrastructure: Financing South Africa's Water Future,' which gathered experts to shift the discussion focus from infrastructure deficit to financial capacity.
The webinar was moderated by Dr. Ochiocheoya Omiunu, a research fellow at the British Academy, and participants included the South African Institute of International Affairs (SAIIA), DNA Economics, GreenCape, and the Infrastructure Finance and Implementation Support Agency (IFISA). Robinson explained that discussions about financing always begin with how the necessary investment will be executed, but they must start with an assessment of the municipality's ability to sustain long-term infrastructure investments.
She clarified that the country's water supply system consists of several interconnected components. The national government is responsible for strategic water infrastructure, including dams and main pumping schemes, which are designed and operated by water boards. Municipalities, on the other hand, are responsible for the infrastructure that directly provides water and sanitation services to end consumers.
According to Robinson, each part of the water sector has its own institutional duties, risk profiles, and financing. She stated that financing municipal water infrastructure is fundamentally a municipal finance problem, but more deeply, it is a systemic problem. In her words, funding failures rarely occur due to a lack of capital; they arise because the institutional conditions necessary to absorb, manage, and maintain these investments have not been created.
The expert noted that the success or failure of projects depends not on isolated factors, but on the functioning of planning, budgeting, revenue collection, procurement, asset management, and governance systems. When these systems operate effectively, finances can flow; otherwise, a market failure occurs, and even well-designed projects face difficulties attracting funding.
Robinson added that the main obstacle to financing municipal water infrastructure is not the availability of capital, but the financial health and institutional capacity of the municipality or sponsor to plan, finance, procure, operate, maintain, and upgrade this infrastructure over the next twenty to thirty years.
She also drew attention to the National Treasury's recent decision to temporarily suspend 'Fair Share' transfers for July for 69 municipalities, which serves as a reminder of the direct link between municipal financial management, institutional trust, service delivery, and willingness to invest. This creates a paradox: despite the acute need for significant investment to restore aging assets, reduce water losses, expand services, and increase climate resilience, many municipalities are simultaneously experiencing decline, revenue collection problems, growing debt obligations, weak liquidity, outdated infrastructure, and limited fiscal capacity. Paradoxically, municipalities with the greatest infrastructure needs are often the least able to attract funding.
Robinson stressed that a project's attractiveness for financing is a characteristic of the environment in which the investment takes place, not an inherent characteristic of the financial instrument itself. She noted that building financially sustainable institutions, reliable management systems, and well-prepared projects begins long before a municipality approaches a lender.
Raldo Kruger, a technical specialist at GreenCape for green finance and the water sector, added that any discussion of municipal creditworthiness must begin with an analysis of the current state of investments and highlighting historical underfunding of infrastructure and the huge need to accelerate investment, especially in the context of increasing climate resilience. He reported that municipalities lose about 35% of the water entering the system due to leaks and losses, which in 2019 was equivalent to a loss of about 10 billion rand in revenue for municipalities across the country.
Kruger defined creditworthiness as the confidence of capital providers that their investments are safe and that they will be able to earn a return. He pointed out that key factors depend on management, revenue management, and expenditure management, all three pillars resting on good governance, leadership, and future planning. Without creditworthiness, municipalities will face serious difficulties in stimulating economic growth through infrastructure investment.
Data centers used for artificial intelligence are consuming significantly more water than disclosed by major technology companies, according to a report by The Wall Street Journal.
The core issue is not just the server cooling process, but also the water required to generate the electricity that keeps all this infrastructure operational. With the rapid expansion of AI, corporations like Microsoft, Google, and Amazon have expanded their global network of data centers. However, a significant portion of the environmental cost remains outside official reports.
These documents tend to account only for the water used directly within the data centers themselves. Nevertheless, there is a secondary consumption associated with the power plants that support these operations, and this can far exceed direct usage. Studies conducted in the United States suggest that this indirect consumption can reach up to twelve times the declared value, depending on the composition of the energy matrix used.
The magnitude of this impact varies drastically depending on the energy source. Plants powered by coal and nuclear energy demand large volumes of water for cooling purposes. In contrast, the use of natural gas reduces this need, while solar and wind sources have virtually no water dependency.
A study mentioned in the article indicates that, specifically in the case of Google, the indirect consumption can be approximately three times greater than the direct consumption. It is important to note that server cooling constitutes the most evident expense, but energy generation has the potential to substantially increase the total impact. Fossil fuels intensify the demand for water, while renewable sources minimize this consumption almost completely. Furthermore, a considerable share of the real impact is not included in corporate statements.
Currently, there is no legal requirement forcing companies to disclose all water consumption related to AI. This lack of regulation allows criticism from experts and environmental agencies, who frequently point to underreporting of data. Meta, for example, has already admitted that its indirect consumption can exceed twenty times the direct consumption. Despite this, most companies continue to focus their reports only on the internal use of data centers.
To mitigate this impact, companies are investing in closed-loop cooling systems, which allow for water reuse instead of continuous discharge. Nvidia claims that this technology has the potential to eliminate almost all direct water consumption in new installations, and Microsoft also foresees the adoption of similar methods in the coming years. Even with these innovations, much of the existing infrastructure still relies on evaporative systems, which, although energy efficient, consume more water.
In summary, the exponential growth of artificial intelligence establishes a clear dilemma: the greater the computational capacity, the greater the pressure exerted on already scarce natural resources.
