The Hidden Environmental Costs of AI

There is a hidden cost to
artificial intelligence.

When we think about the costs of using AI, we often think about time or money – like paying for a subscription or trying to save time on an assignment. But there’s another cost most people don’t see: the environmental impact.

You may heard that AI is in “the cloud”, and not think anything of it. It’s not a machine that you have to plug in, it’s just there. However, anytime you hear “the cloud” think “other people’s servers.” And while the cloud is more efficient than us all having our own computers, that doesn’t mean it doesn’t need power… in fact, they need a lot of power to operate.

If you’re curious as to how much, several companies have been looking at building/using nuclear power plants to meet their needs.

This section explores how artificial intelligence uses energy, why that matters, and what role we play as students and future professionals in making tech choices that are more sustainable.

Why AI Needs So Much Energy

When you use an AI tool, whether it’s generating a paragraph of text, creating an image, or answering a question, your device sends that request to a massive data center. These centers contain thousands of powerful computers, known as servers, that run AI models and return a result to your screen in seconds.

Behind that quick response is a huge amount of processing power, especially with modern AI models like ChatGPT or DALL·E. These tools are built using deep learning, which means the models are trained by running billions (or even trillions) of calculations over massive amounts of data.

This process requires:

• High-performance hardware (GPUs and TPUs)
• Powerful cooling systems to prevent overheating
• Constant electricity to keep everything running 24/7

That all adds up to significant energy use, both during the training phase (when the AI model is first being built) and the inference phase (when the model is responding to users like you).

Key Terms to Know

Here are a few technical terms that come up when talking about AI and energy:

• Training: The initial phase where an AI model learns patterns from a large dataset (can take weeks or months and use massive resources).
• Inference: The part where the model uses what it learned to respond to a prompt or generate content (what happens every time you ask a question).
• Data Center: A facility that houses many servers, often used by cloud services like Microsoft Azure, Google Cloud, or Amazon Web Services.
• Carbon Footprint: The amount of carbon dioxide (CO₂) released as a result of energy use. This is a common way to measure environmental impact.

---

How Much Energy Are We Talking About?

The numbers vary depending on the size and type of AI model, but here are a few examples:

• Training GPT-3, a predecessor to today’s more advanced models, was estimated to use the same amount of electricity as a small town for several weeks.
• Running millions of daily user prompts through large AI models adds up quickly — some researchers estimate each prompt may use as much energy as a Google search times 10.
• AI image generation tools, which create complex visuals from text, often require even more computational power than language models.

And don’t forget — this doesn’t include the energy used by your own device (laptop, phone) or the network infrastructure that sends your request to the AI servers.

Why This Matters

At first glance, using AI tools may seem harmless. There’s no smoke, no pollution, no noise. But just because something happens in the cloud doesn’t mean it’s impact-free.

Here’s why it matters:

1. Climate Change
Electricity production is still a major source of carbon emissions globally, especially in areas that rely on coal or gas. More energy use from AI means more emissions, unless the data centers are powered by renewable energy. Due to their 24/7 nature, renewables can only contribute a small fraction of the energy needs of a data center.

2. Sustainability and Equity
Running AI takes money, power, and specialized hardware. That means only well-funded companies or countries can train and maintain large models. This can lead to inequalities in who controls AI tools, and who benefits from them. This may also directly lead to certain biases, and factual errors if a host countries laws prohibit certain knowledge. DeepSeek, built in China, will not give accurate information about certain topics such as the Tiananmen Square – https://www.theguardian.com/technology/2025/jan/28/we-tried-out-deepseek-it-works-well-until-we-asked-it-about-tiananmen-square-and-taiwan.

3. Growing Demand
AI is spreading fast into apps, search engines, social media, and productivity tools. If the energy demand of AI keeps growing at its current rate, it could become a major part of the tech sector’s environmental impact in just a few years.

Should You Feel Guilty Using AI?

This doesn’t mean you should stop using AI altogether. Afterall, do you only walk because cars create pollution? But it does mean you should think critically about when and how you use it.

Consider these two approaches:

1. Mindless use:

• Prompting an AI 20 times just to write something you could do yourself
• Generating images “just for fun” without realizing the power it takes
• Rewriting the same output over and over without a clear goal

2. Mindful use:

• Using AI to brainstorm ideas when you’re stuck
• Asking for feedback on something you’ve already written
• Editing or refining work based on your own learning process

The second approach respects both your time and the environment. It makes AI a support tool, not a default or a shortcut for everything.

What Tech Companies Are Doing About It

Some large tech companies are trying to address AI’s energy demands:

• Google and Microsoft are investing in renewable-powered data centers and improving the efficiency of their AI systems.
• Meta and OpenAI have released papers on making AI models smaller and more efficient.
• Some companies are now labeling their tools with energy use estimates or offering eco-friendly modes for AI use.

But progress is uneven, and it’s still hard for the average user to tell how “green” a specific AI tool really is.

What You Can Do as a Student

While you can’t control how AI companies build their tools, you can control how you interact with them. Here are a few ideas:

• Use AI purposefully, not passively. Don’t run dozens of prompts just for fun if you don’t need them.
• Choose tools backed by responsible companies. Look for information on how they power their data centers.
• Turn off auto-suggestions or AI assistants when you don’t need them.
• Talk about these issues in class or with friends. Raising awareness can lead to more mindful habits across your campus or community.
• Support local or open-source alternatives that focus on sustainability or efficiency.

Final Thoughts

Artificial intelligence doesn’t run on magic, it runs on electricity. A lot of it. And the energy required to train and operate large AI models has real-world consequences for the environment.

As a student, you have a choice: use AI tools thoughtfully, understanding that your decisions, even digital ones, leave a footprint. With a little more awareness, you can balance the benefits of AI with your values around sustainability, responsibility, and long-term impact.

Next up: Bias in AI and Training Data Transparency – exploring how AI models learn from the past, and why that can lead to some unfair or unexpected outcomes.

The Hidden Environmental Costs of AI was originally found on Access 2 Learn