AI Sustainability: Use of an Energy Hungry System in the Name of Efficiency

Photo by Chris LeBoutillier on Unsplash

Technological advances paint a promising picture: a smarter, cleaner and more efficient future, but what if the very thing supposedly changing the world for the ‘better’ is draining the lifeblood of the climate fight – energy itself. Artificial intelligence is neither clean nor sustainable. It is not a weightless, frictionless entity but a costly development, paid through damage to the environment. Behind every ‘intelligent’ model is an energy hungry infrastructure of data centres, servers, cooling systems and electricity fields, and they are all largely fossil fuelled. What does this mean? The same technology positioned as a climate solution is the very thing amplifying the problem.

So, are we fuelling salvation or reckoning disaster?

In an analysis of 237 papers between 2010 and 2024, researchers highlight both the benefits and significant negative effects on energy consumption of AI across the world. It is not false that algorithms can predict demand, calibrate renewable integration or expose inefficiencies missed through human error, but the very processes used to create and maintain AI models consumes absurd quantities of energy, resulting in more damage done, despite their use in this context being primarily to benefit.

AI, therefore, is not neutral; it has a carbon signature. People are susceptible to the narrative that ‘AI will save us’, in the same way that burning fossil fuels used to be the way of the future; a large change in infrastructure (and a lucrative one, at that). The training of large language models, neural networks and deep learning architectures is energy intensive and relies heavily on power drawn from carbon heavy grids. Sustainability and ‘net-zero’ are not possible in the hands of AI models draining tens of megawatts of electricity just during the model training. The obsession with computational capability undermines climate change goals.

Yes, AI is capable of improving efficiency…and dramatically, from buildings, to grids, to transport networks. It can integrate renewables more smoothly into power systems and it can forecast consumption and optimise peak loads. But it is not what AI is capable of, it is what we allow it to do. AI use does not end with optimising systems to be more sustainable, it is integrated into daily life. While it decarbonises a small part of the system, it’s increased use is not reducing energy consumption and demands at all. The question is not whether we should use it, but why  we choose to.

There is an ethical divide, AI is not only a technological challenge, but a moral one. Every organisation, developer and policy maker that promotes AI use is not promoting technology, it’s making an ethical choice – or perhaps, ignoring that there is an ethical choice to be made. A machine that learns faster and deeper isn’t automatically serving humanity, it may just be learning how to consume more. That’s not progress, it’s indulgence.

This research is not just for engineers and climate scientists, the philosophical challenge is for us all, everyone immersed in the age of AI. As AI use shapes markets, workplaces, policies and even personal choices, the residual energy footprints are at the responsibility of all who participate. AI is not invisible, its impact on energy consumption is measured both in kilowatts and in our own values, priorities and consequences. With any system, there should be consideration both of its output and of the cost to achieve it, is this a worthy cost? Perhaps at its best, AI is only a tool, requiring wisdom to use it, and reflection is the key distinction between acceleration and real progress.

From a depth psychology perspective, the revelations of our relationship with AI is far more about us than it is about the machine. AI has become a projection screen for the modern psyche, humanity remains desperate for control, spurred by fear of uncertainty, outsourcing decisions due to the repeated cycle of placing responsibility onto something intangible or omniscient. We attribute algorithms with ‘mythical powers’ while disowning the shadow they carry, which is, in this case, the energy they consume, the emissions they generate, the consequences we choose not to see. In Jungian terms, AI becomes the trickster archetype: it is brilliant, transformative, and deeply destabilising when left unchecked.

The facets we do not consciously integrate, like AI’s limits, the physical and psychological trade offs and the ecological costs, returns unconsciously as climate debt. The machine itself is not reckless; our unconscious worship of limitless intelligence is. If AI’s shadow is its energy consumption, then the change does not lie in rejection but in integration; instead of outsourcing responsibility to the external (like technologies), humanity can reclaim autonomy of choice. An integrated relationship with AI would not heavily rely on its output, but be designed to respect limits that value sufficiency over ‘efficiency’ and recognising that not every problem requires maximum intelligence at a maximum scale.

For the full research paper published in Energies, Volume17, 2024, find the link here:

Challenges of Artificial Intelligence Development in the Context of Energy Consumption and Impact on Climate Change

Follow The Heretic for reflections on humanity's connection to technological advances and the consequences of relationships with AI.