AI is helping charities do more with less. But what is the hidden cost?

AI is becoming part of everyday work across the sector. It is helping teams move faster and do more with less. But behind these benefits sits a growing environmental cost that is often overlooked.

From drafting emails, content or funding applications to analysing data or summarising reports, AI tools are helping teams save time and increase productivity. In a context where resources are limited and demand is growing, that matters.

But there is a side of AI that we talk about far less (me included): its environmental impact.

This article is inspired by recent research from Hive Mind and Connection Technologies, alongside additional insights on water usage and data centre design. Together, they help paint a more complete and balanced picture.

The scale problem most of us overlook

A single digital action often feels insignificant. For example, a ChatGPT (or any other LLM) query uses roughly 0.2-0.5 watt-hours of electricity, and a small amount of water for cooling. On its own, that is negligible, but at scale, it becomes meaningful.

With around 1 billion AI queries per day globally, this adds up to energy consumption comparable to tens of thousands of homes each year, alongside significant water use.

This reflects a familiar pattern: when technology becomes cheaper and easier to use, we tend to use it more, not less.

Where the environmental cost comes from

As highlighted in the Hive Mind article, AI’s footprint goes far beyond what happens on our screens. It spans the full lifecycle:

Hardware and infrastructure

AI relies on physical systems. Manufacturing chips, requires large volumes of water and rare minerals, while data centres require significant energy and land.

Model training

Training large models can consume energy equivalent to thousands of households over several months.

Everyday use

According to Connection Technologies, this is now the dominant cost. Every prompt, every generated output contributes to cumulative demand.

E-waste

The push for more powerful hardware shortens device lifespans, contributing to growing electronic waste.

Human labour

Behind AI systems are workers labelling data and moderating harmful content, often under challenging conditions.

A quick note on water: not all metrics mean the same

Water use is often cited in discussions about AI, but it is easy to misinterpret.

There is an important distinction between water withdrawal (water taken from a source) and water consumption (water that is not returned and is no longer available). This matters because some systems withdraw large volumes but return most of it, while others consume smaller amounts but with greater long-term impact.

Understanding both gives a clearer picture of real environmental pressure, especially in water-stressed regions.

Are data centres as “thirsty” as headlines suggest?

Some reports highlight the growing water demand of AI infrastructure, and those concerns are valid. However, not all data centres rely on water-intensive cooling, and some systems use closed-loop or air-cooled designs, where water is reused or not consumed at all.

This does not remove the environmental impact, but it shows the picture is more nuanced than it sometimes appears. Technology choices, design decisions and location all play a role.

What this means for our sector

Most charities are not involved in building AI models themselves, but we are increasingly becoming regular users of these tools, which means that while we may not control how they are developed, we do have influence over how and when we use them. In many organisations, AI is becoming a default solution, even in situations where simpler tools or more traditional approaches might be just as effective, and this can lead to patterns like generating content that is not really necessary, repeating similar queries instead of building on existing outputs, or relying on AI out of habit rather than intention (I will admit I am guilty of the latter).

Practical ways to reduce impact without losing value

This is not about avoiding AI altogether, but about using it more deliberately and thoughtfully, making conscious decisions about when it genuinely adds value and when it might not. In practice, this can mean taking a moment to match the tool to the task rather than defaulting to the most powerful option available, avoiding more resource-intensive uses such as image or video generation unless they are truly needed, and making better use of existing outputs instead of starting from scratch each time. It can also involve paying closer attention to the platforms and providers we rely on, particularly as some are making more progress than others in improving the efficiency and sustainability of their infrastructure.

A balanced way forward

AI is already transforming how charities operate, offering clear benefits in terms of efficiency, quality of work, and the ability to do more with less, which is especially valuable in a sector where time and funding are often constrained. At the same time, its environmental footprint is real and continues to grow as adoption increases, which makes it important not to ignore this aspect but to understand it well enough to make informed choices. Taking a more conscious approach doesn’t mean losing the advantages that AI brings, but ensuring that we use those advantages in a way that is both effective and responsible over the long term.

Want to explore AI in your charity in a more conscious and practical way?

If you are at the stage of “we know we need to start, but we are not sure how”, here are a few resources that might help: