In the ongoing discourse around technology and its environmental costs, a concerning trend has emerged. As awareness grows about the outsourcing of data center costs to developing countries, it becomes clear that many citizens globally are increasingly opposed to this practice. High energy demands linked to artificial intelligence are often passed on to consumers, creating both financial and environmental burdens that are disproportionately shouldered by those in less affluent regions. The need for transparency and responsibility from tech giants has never been more critical.
Yves here. It seems that citizens in various countries are rejecting the Big Tech strategy that shifts data center environmental and energy expenses onto so-called third-world nations. It strikes me as ironic that the increasing energy demands tied to AI technology are burdens that all customers must bear. Those who advocate for higher energy prices should face the costs themselves. Instead, the general population ends up subsidizing AI through elevated energy expenses while also coping with the added pollution.
Citizens in advanced economies have displayed enough resistance to prompt tech leaders to seek out other regions as potential targets, following the assumption proposed by Larry Summers that poorer countries would readily accept the burdens created by energy generation. However, Jomo Kwame Sundaram highlights the growing awareness among many in these nations regarding this neocolonial maneuvering, and they are not willing to comply.
By Jomo Kwame Sundaram, former UN Assistant Secretary General for Economic Development. Originally published at Jomo’s website
Data centers have been expanding rapidly, driven largely by the rise of artificial intelligence.
Who Are Data Centers For?
The current boom in AI has eclipsed other cloud-related applications, leading to a swift increase in data centers (DCs) and placing immense pressure on resources. This trend has sparked a bipartisan backlash in the U.S., as concerns grow over heightened energy, water, and land usage, along with rising costs.
In an October 2024 report, McKinsey projected that global energy demand from data centers would surge by 19% to 22% annually through 2030, reaching a staggering 171 to 219 gigawatts. This demand far exceeds the current requirement of 60 GW, making it clear that to avoid shortages, at least double the capacity built since 2000 must be constructed in a fraction of the time.
Unfortunately, tech companies are not shouldering the additional expenses tied to increased energy generation; instead, consumers and local governments are left to bear these costs, regardless of whether they use AI’s benefits.
Faced with growing resistance from wealthier nations, developers of data centers have turned their sights toward developing countries, effectively outsourcing the environmental and economic challenges to those with limited resources.
Understanding the extensive energy and water needs of these facilities is crucial for safeguarding economies, societies, communities, and ecosystems.
Energy Needs
The increasing corporate and consumer appetite for AI ensures that data center growth will not only continue but may also accelerate.
The rising demand for AI will dramatically elevate both energy and water consumption, exacerbating climate change both directly and indirectly.
As the demand for AI and data centers escalates, the supporting infrastructure will need substantially more electricity, leading to an increase in heat generation. This, in turn, necessitates more energy and water for cooling purposes. Notably, between 38% and 50% of energy consumed by data centers is dedicated to cooling.
Electricity generation from fossil fuels or nuclear fission requires more cooling than energy produced through renewable sources, such as solar panels or wind turbines.
A small-scale data center containing 500 to 2,000 servers consumes between one and five megawatts (MW). In contrast, a ‘hyperscale’ data center, which houses tens of thousands of servers, can consume 20 to over 100 MW, akin to a small city!
Data Centers Not Cool
While much attention is given to the significant energy needs of data centers, their extensive water requirements for cooling often go unnoticed or are understated.
Locating new data centers in developing areas can exacerbate local climate conditions and contribute to global warming. Heat becomes even more problematic in tropical regions, where baseline temperatures are already elevated.
Furthermore, the establishment of additional data centers will inevitably lead to the depletion of freshwater supplies essential for local communities, as well as the reduction of groundwater reserves.
It’s no surprise that investors in data centers rarely inform host governments about the substantial quantities of energy and water these facilities demand.
Data centers require a significant amount of freshwater for cooling servers and routers. For instance, Google alone utilized nearly 23 billion liters of water for cooling in 2023. In cooling systems that rely on evaporation, cold water absorbs excess heat and releases steam into the air.
Closed-loop systems manage heat with piped water, while air-cooled chiller units work to cool hot water. Although recirculated cold water requires less overall water, it demands more energy to chill effectively.
Investors Expect Subsidies
Data centers, like other types of investors, often relocate to regions where government officials are eager to provide financial incentives and impose fewer demands.
Championed by U.S. President Trump and his allies in the tech industry, many foreign investors have benefited from subsidized energy, affordable land, and special perks.
Host countries are often in a competitive race to offer tax breaks and other incentives, like subsidized energy and water, to attract foreign direct investment in data centers.
The U.S. has pressured nations like Malaysia and Thailand to block Chinese firms from using their territories to evade export controls for AI chips. Washington alleges that data centers outside China acquire chips for military AI training, though Malaysia is currently the only country that has complied.
This situation restricts Chinese companies’ access to such chips, as the U.S. asserts that substitutes from China fall short of the quality of American products, aiming to safeguard U.S. technology from competition.
High-Tech DC Jobs?
While data centers are appearing across the globe, the potential for job creation is minimal. Advocates argue that these centers will generate high-tech employment opportunities.
However, most data centers operate autonomously, necessitating minimal human involvement, mainly for maintenance tasks determined automatically. This results in limited job openings.
Construction and installation jobs will only be temporary, while managerial responsibilities tend to be carried out from remote locations. According to a Georgetown University report, merely 27% of jobs in data centers are considered ‘technical’.
Although the conversation primarily revolves around foreign investments, there is a notable lack of discussion on the rising national interest in data sovereignty.
Compromising national capabilities to satisfy such foreign demands could hinder efforts to cultivate end-to-end data center capabilities rather than merely hosting them.
Additionally, there seems to be little thought given to the ‘afterlife’ of data centers—what occurs once they have fulfilled their function and how waste materials are disposed of.
Ultimately, the high costs of energy and water, along with various incentives and complications linked to data centers, are hardly justified by the modest employment opportunities and benefits they may provide.
