Smart Tech, Dirty Footprint: Why AI Is an Environmental Issue
Hi, I’m Mina. I’m a student, a writer, and someone who, more often than not, wonders about the true cost of the services I use daily, such as apps. Like, how much water does it really take to power a chatbot? Or what’s the carbon footprint of me sending a small message (yes, it has a carbon footprint, look it up)? Admittedly, I’m not perfect — I still binge-watch shows and order takeout — but I’m trying to be more conscious. That’s why I started this blog: to explore sustainability beyond surface-level green choices and think about what it really means to live responsibly in a world that’s rushing forward faster than most of us can keep up with.
For my first post, I want to dive into something that’s becoming impossible to ignore — Artificial Intelligence. You’ve probably seen headlines like “AI will change everything,” or maybe you’ve tried one of the new chatbots yourself. However, beneath the glossy ads and sleek interfaces is a story about resources, environmental impact, and justice that we don’t talk about enough.
I’m not here to say “the robots are coming,” or to scare you with doomsday tech talk. Instead, I want to ask a simple question: why does running a chatbot sometimes use the energy of a small country? What’s really going on behind the scenes, and what does it mean for our planet?
The Invisible Cost of Smart Tech
When we talk about sustainability, images of plastic bottles or car emissions often come to mind. But digital technology? It feels weightless, intangible. A conversation with an AI doesn’t leave trash in a landfill or smoke in the air. It’s all data, right? So it must be “clean.”
Unfortunately, that’s not how it works.
Training a large AI model is an incredibly energy-intensive process. These models don’t just pop out of nowhere — they’re created by feeding enormous amounts of data through algorithms that run on thousands of powerful graphics processing units (GPUs) in data centers. This process can take weeks or months and consumes massive amounts of electricity.
For example, training GPT-3 — the AI behind many chatbots today — required roughly 1,200 megawatt-hours of electricity and emitted about 550 metric tons of CO₂ (Patterson et al., 2021). To put that in perspective, that’s approximately the same amount of carbon emitted by an average American’s roundtrip flights from New York to London 550 times. That’s one model, trained once. But these models are not static. They get retrained, fine-tuned, and scaled constantly, multiplying that footprint many times over.
Why AI Feels So Clean — But Isn’t
AI feels sleek, modern, and even futuristic. Its white-and-blue interfaces, gentle voices, and minimalist design make it seem harmless — or even helpful. But behind that is a vast infrastructure that’s anything but ethereal.
These AI systems run on physical servers housed in data centers — huge warehouses packed with rows of machines running 24/7. To keep the servers cool and prevent overheating, data centers use massive amounts of water, often in areas already suffering from water scarcity.
Take Microsoft’s data centers, for example. They reported a 34% increase in water use within a single year, much of it tied to AI expansion (Hao, 2023). Some centers consume millions of gallons of potable water daily — water that could otherwise be used for drinking, farming, or sustaining local ecosystems (Li et al., 2023).
This means that the digital convenience we enjoy isn’t free. It comes at the cost of electricity often generated from fossil fuels and water pulled from drought-stricken regions. That’s the messy, complicated reality hidden behind your “smart” assistant.
The Environmental Footprint Beyond Energy
Energy consumption is just the tip of the iceberg.
The hardware powering AI — GPUs, CPUs, memory chips — requires rare minerals like cobalt, lithium, and nickel, which are mined in environmentally and socially sensitive regions. Mining these materials causes habitat destruction, pollution, and human rights issues (Earth.Org, 2023). And the rapid turnover of tech means that these precious materials often end up as electronic waste after only a few years.
Plus, the physical infrastructure itself — data centers and networking equipment — needs to be built, maintained, and eventually replaced, all of which adds to the overall environmental impact.
Is AI Always Bad for the Environment?
It’s important to stress that AI is not inherently bad. In fact, AI holds significant promise for addressing some of the biggest sustainability challenges.
Researchers are using AI to model climate change with precision, optimize renewable energy grids, reduce waste in supply chains, and detect illegal deforestation or overfishing. In these cases, the environmental costs of AI may be outweighed by the benefits of smarter resource use and better decision-making.
But here’s the rub: much of AI’s environmental impact depends on how it’s used. Today, a significant amount of AI-powered computing is devoted to consumer apps, ads, entertainment, and novelty tools. Whether the use case is serious or frivolous, the energy consumption is similar and that makes the environmental cost a question of choice by individuals, companies, and governments.
Who Bears the Environmental Cost?
Another layer of this issue is environmental justice. One key lesson I've learned during my time at university is that marginalized communities often end up bearing the brunt of the world's burdens.
Data centers are often built in regions where energy is cheap and regulations are lax, rural areas in the U.S., parts of South America, or countries in the Global South (Earth.Org, 2023). These communities rarely benefit directly from the AI-powered apps their infrastructure supports. Instead, they face higher energy costs, water shortages, and environmental degradation.
This pattern is similar to what we see in many extractive industries — those who profit most from a resource or technology often live far from where the environmental costs are borne.
This raises tough questions: if AI is the future, whose future are we building? And who gets to decide?
What Can We Do?
If you’re feeling overwhelmed or powerless right now, I get it. These problems are big, systemic, and complex. However, recognizing the hidden costs is a first step.
Here are a few ways we can start shifting the narrative:
Demand transparency. Companies developing and deploying AI should be open about the environmental impacts of their technologies.
Support greener tech. Some companies are investing in renewable energy, waterless cooling, and more efficient hardware. Supporting them!
Use AI mindfully. As users, we can reflect on whether we really need to ask an AI to write that email or generate that image, especially if it’s just for fun.
Push for policy.
and most importantly…
STAY CURIOUS AND CRITICAL
Why This Matters to Me
When I first learned about AI’s hidden environmental costs, I felt a bit betrayed. I love technology and what it can do for us but I also want to live in a world where “progress” doesn’t mean sacrificing the planet or leaving others behind.
Sustainability can’t just be about individual choices — reusable bags, electric cars, or cutting down on plastic. It has to be about looking deeper at the structures, habits, and systems that define how we live and that includes digital technology.
So, if you’re like me — curious, worried, hopeful — I hope you’ll stick around. This blog isn’t about perfection. It’s about learning together, asking questions, and trying to do better.
At the end of the day, sustainability isn’t a destination. It’s a way of paying attention — to what we use, who we affect, and the world we want to build.
Thanks for reading my first post. I hope you stay for more.
References:
Earth.Org. (2023). The real environmental impact of AI. https://earth.org/the-green-dilemma-can-ai-fulfil-its-potential-without-harming-the-environment/
Hao, K. (2023, August 9). The AI industry is secretly using a ton of water. MIT Technology Review. https://www.technologyreview.com/2023/08/09/1077471/ai-industry-water-usage/
Li, P., Yang, J., Islam, M. A., & Ren, S. (2023). Making AI less “thirsty”: Uncovering and addressing the secret water footprint of AI models. arXiv. https://arxiv.org/abs/2304.03271
Patterson, D., Gonzalez, J., Le, Q. V., Liang, C., Munguia, L. M., Rothchild, D., So, D. R., Texier, M., & Dean, J. (2021). Carbon emissions and large neural network training. arXiv.https://arxiv.org/abs/2104.10350
