Climatic Footprint of Artificial Intelligence: The Hidden Environmental Cost

INTRODUCTION

The human brain, being the epitome of the evolutionary wonder of Mother Nature, and its ability for logical reasoning, problem-solving, and learning skills, is the cornerstone of human intelligence. It is unique in the context of active perception of surrounding environmental stimuli and subsequent adaptive response.¹ The first attempt to simulate this “Biological intelligence” into an artificial one was made by English mathematician Alan Turing, who in 1950 introduced the “Turing test” to determine whether a machine is capable of thinking or not. This artificial intelligence (AI) though fundamentally different in structure, tries to mimic the cognitive processes using algorithmic logic, pattern recognition, and decision-making capability, thus, can be described as a “mechanical system” that can interpret and learn from external sources or cues, and subsequently apply the same to solve specific problems with greater flexibility in a more efficient manner.² The 21^st century has seen an overwhelming development in the field of AI, which, like other general-purpose technology, i.e., steam engines, revolutionizes human civilization. In a spectrum of supervised or “trained” AI to unsupervised or self-aware AI, it has penetrated deep into our daily life, from automation to communication.³ With its ever-expanding horizon, AI has a clear prospect in solving many of the problems currently faced by humanity, often without foresight.

CLIMATE CHANGE AND AI

Climate change refers to the “long-term shifts in the temperature and weather patterns. Though nature had its mechanisms to regulate and change climatic conditions, human activities played a major role since the inception of the Industrial Revolution in the 18^th century. In comparison to the pre-industrial era, the Earth’s surface is 1.2°C warmer owing to the greenhouse effect, such that 2024 was the warmest year on record with an average surface temperature of 1.29°C above the 20^th century average.^4,5 This resulted in the melting of polar ice and rising sea levels, frequent and extreme climatic events and disasters, ecological disruption and declining biodiversity, water and food insecurity, and lastly emergence and re-emergence of pathogens.⁶ While it is a well-known fact that the burning of fossil fuels such as coal and oil is the major cause of emission of greenhouse gases, lately, there has been a new character silently promoting climate change, which is AI.⁷

Apart from the ethical issues related to AI use, the environmental impact of AI is seldom discussed, but with each passing day, they are becoming prominent. For programming, training, and running large language models, there is a need for “data centers” that house and manage the computing equipment, along with servers and data storage.⁸ However, as a byproduct, they generate heat, and there is a need for a proper cooling system to maintain their function. Thus, water, specifically clean water, is being used as a cooling agent in these data centers, and with the advent and popularization of AI, water demand for these data centers skyrocketed as in 2021 a data center in Dallas, USA, consumed water equivalent to 29% of the city’s water demand. This highlights the artificial water scarcity created as an indirect result of AI, and more so in the areas that are already prone to drought, also hampering the response made during disasters.^6,9 The latest example is the California wildfire, where it was suspected that, due to the presence of massive data centers, there was a lesser amount of water available, needed for firefighting.^10,11 Sometimes, the data centers are located in third-world countries, and in the absence of a strong regulatory mechanism, they deprive the local population of their access to clean drinking water.^12,13 On the other hand, the massive amount of energy consumed by these data centers mainly comes from burning fossil fuels, and they also need water for cooling purposes.⁹ Thus, directly and indirectly, AI is contributing to global warming and the ever-increasing need for clean drinking water. Apart from water, data centers heavily rely on semiconductors, which require rare earth minerals such as yttrium and neodymium. Their mining again is related to increased emissions of greenhouse gases.¹⁴ Moreover, the E-waste generated from these data centers is hazardous to human health and the environment.¹⁵

CONCLUSION

Although slowly but steadily this environmental problem associated with AI, is being recognized early on and initiatives are taken like the concept of “Water Positive” AI with recycling and reducing the water demand by shifting to alternative methods like air-based cooling, etc., and shifting the data centers in cooler regions which significantly reduce the water demand. Shifting from fossil fuel to renewable energy to power these data centers also facilitates sustainable growth and development to fulfill the Paris Agreement to limit global temperature rise below 2°C. Finally, to navigate the world through this “AI storm,” we need global administrative action and political commitment to create a regulatory framework for the mitigation and management of the adverse environmental impact of AI, making it sustainable and climate-friendly.