Rising Environmental Impact of Data Centers: Should We Abandon AI?

In just a few short years since the emergence of ChatGPT, the integration of artificial intelligence across various sectors has led to an exponential rise in the number of datacentres. This rapid growth is accompanied by increasing environmental repercussions that cannot be overlooked.

Worldwide, the power requirements for datacentres are escalating at a rate fourfold compared to all other industries, as noted by the International Energy Agency. This trend suggests that, by 2030, the energy consumption of datacentres could surpass that of Japan’s entire electricity use.

In Australia, the outlook is similarly concerning. The national energy market operator projects that the energy demand from datacentres is set to triple within the next five years, exceeding the electricity consumed by the entire fleet of electric vehicles by 2030. Such a surge is also expected to place tremendous pressure on drinking water supplies.

In light of movements like the QuitGPT initiative, which advocates for a boycott against AI due to its applications in surveillance and weaponry, it raises a crucial question: should individuals concerned about the environmental impact of AI also consider stepping back?

How Bad Is AI for the Environment?

While estimates vary, most studies indicate that generative AI models—which create text, images, and videos—utilize “orders of magnitude” more energy compared to traditional computing.

Some analyses suggest the energy consumption could be as much as five times greater, whereas others predict the figures may be significantly more. A lot of this depends on the model being used and the complexity of the query presented.

Prof. Jeannie Paterson, co-director of the Centre for AI and Digital Ethics at the University of Melbourne, states that one major issue is the opacity from tech companies regarding the energy, water, and emissions implications relating to AI and datacentres. “It’s clear that training models and running datacentres is an energy-intensive task,” she emphasizes.

Ketan Joshi, a climate analyst affiliated with the Australia Institute, explains that consumer software generating text and images is particularly energy inefficient. This inefficiency arises from the immense datasets and computational demands involved in underlying processes. When people ask AI chatbots questions, the energy consumed is far greater than using simple web searches or calculators. This resembles choosing to drive an SUV to the store rather than pedaling a bike.

“One could complete the shopping trip with either method, and the standalone energy costs may not seem alarming, but when this behavior becomes normalized across society, that’s where the true impact manifests,” he comments.

A study published in the journal Patterns estimates the global carbon emissions from AI to be between 32.6 to 79.7 million tonnes of CO2 in 2025 and predicts a water usage of around 312.5 to 764.6 billion liters—equivalent to global bottled water consumption.

In Australia, the increasing reliance on datacentres for processing and storing AI data is expected to hinder the renewable energy transition, bolster emissions, and elevate power costs for consumers.

“That’s a substantial energy requirement for unclear or minimal societal benefit,” Joshi remarks, contrasting the situation with that of video-calling technologies, which have allowed for reduced flights and facilitated communication during the pandemic.

AI Is Everywhere. Is It Possible to Opt Out?

AI tools have infiltrated various sectors, becoming integral to workplace software, educational platforms, and even in chatbots utilized by banks and local governments. The rollout of generative AI is also making its way into supermarket self-checkouts, facial recognition at stores, and even for transcribing doctors’ notes.

“We’re becoming immersed in this technology,” Paterson observes, emphasizing that it’s increasingly challenging to escape. Nonetheless, she states, “We still have a window to voice our opinions about how we prefer AI to be implemented.”

Individual actions can help mitigate this reliance—such as saving energy by turning off lights or appliances. Users can opt-out of AI platforms, exclude AI-generated results in searches (by appending “-AI” to their queries), and avoid utilizing AI for tasks that are both unnecessary and energy-intensive, such as generating images for celebrations or work presentations.

Joshi expresses concern that major corporations like Meta, Google, and Microsoft have embedded generative AI deep into their systems. He likens this effort to a tactical shift aiming to instill dependence on such technology, much like the rise of single-use plastics in the 1970s.

According to Joshi, opting out can be a “meaningful act of resistance.” It not only addresses energy demand but is also a part of broader collective action against a potentially detrimental industry. While he acknowledges the strength of consumer boycotts, he is alarmed that QuitGPT is encouraging users to shift from one AI platform to another instead of promoting a complete withdrawal from AI.

What About the Impacts of Datacentres on Local Communities?

Datacentres are the tangible manifestations of the AI expansion, rapidly increasing in both number and capacity. As this trend continues, calls for accountability regarding their environmental impacts are becoming louder.

A coalition of environmental and energy organizations, including the Clean Energy Council, Electrical Trades Union, Australian Conservation Foundation (ACF), and Climate Energy Finance, has put forward a series of “public interest principles for datacentres.” These principles emphasize the importance of investing in renewable energy and responsible water usage.

“If you want to establish a datacentre, it is essential to also invest in renewable energy and water recycling,” states Adam Bandt, CEO of ACF, urging that large tech companies participate in their fair share to prevent draining resources.

Beyond energy, water, and emissions, there are local ramifications that can affect communities and wildlife residing near these datacentres—large warehouse-like structures characterized by constant lighting and the continuous hum of air conditioning units.

In some regions, communities have stepped up to protest against the establishment of massive datacentres in their vicinity, taking matters into their own hands by actively campaigning against such developments.

Dr. Bronwyn Cumbo, a transdisciplinary social researcher at the University of Technology Sydney, notes that these datacentres are often located in clusters, creating an industrial complex rather than operating as standalone entities. “While it is in their best interest to communicate and engage with the community, the incentive to be a good neighbour often hinges on the individual company,” she explains.

Cumbo asserts that discussions surrounding AI and its interactions with the physical environment and its social, political, and economic implications are reaching a critical point. Raising awareness is vital for communities to think critically and identify relevant questions to raise.

“Though AI’s integration into our lives seems inevitable, we still have agency over how it integrates,” she concludes.