Picture of an illustration of a mann standing in front of many screens

Navigating The Environmental Challenges Of AI

As a digital agency, Luminary recognises that its use of AI – a non-negotiable in today’s competitive technology landscape – is inevitably going to have an impact on the environment. We reflect on our journey towards offsetting that impact.

Emma Andrews

15 June 2026

3 minute read

This article was originally published in SMBtech.

A little more than 12 months ago, our CEO and I sat down and asked ourselves “How do we stay true to our commitment to environmental sustainability now that we’re using AI not only to do our work, but with the knowledge that when our work is complete, the websites we’ve built are feeding the system where customers are using AI to find the information within them?”

We started investigating how we would balance our need to use AI with our responsibility to the environment and to our values as a certified B Corp (a movement where companies commit to balancing profits with people and the planet).

We’d already implemented an AI usage policy and across our team of 90 people, some were adopting AI tools at pace. We encouraged this, too, by providing the structure for people to use AI to provide the best value for our clients and the most fulfilling work for our teams.

But what was the impact?

There were now many facets to consider. There was a growing recognition (including from sustainability bodies) that AI was having an impact across several areas:

  • Energy consumption – high energy required for training and inference
  • Water – significant amounts of fresh water is used for cooling data centres
  • E-waste – implications for obsolete hardware being replaced with highly specialised equipment in data centres).

We weren’t training the models like the large tech companies in Silicon Valley. So we were limited as to how we could make a difference there. Nor could we do much about the e-waste generated from data centres. We also weren’t going to stop building websites anytime soon. What was left? Water. Water was the area where we could have some impact.

AI has a water footprint

You’re familiar with carbon footprints: the total amount of greenhouse gases (including carbon dioxide and methane) that is generated by our actions. But you might not be so familiar with a water footprint…

AI needs water; a lot of it. More specifically, the data centres that are needed to power the AI need a lot of water for cooling. But there have always been data centres, right? So what’s the problem now?

Well, older data centres tended to rely on air cooling. Demand for more computing power means higher server rack density so the output has become warmer. As a result, data centres have turned to water for cooling.

As well as rising temperatures, data centre numbers have also surged. Australia’s data centre market has expanded by 40 times in 20 years, with two-thirds of this growth coming since 2020. There are now 260 data centres in Australia – and growing.

Sydney Water has estimated up to 250 megalitres a day will be needed to service the data centre industry in Australia by 2035. Globally, the UN reports that AI-related infrastructure may soon consume six times more water than Denmark, a country of six million. These numbers are hard to fathom. So let’s distill it down to a relative amount. ABC’s Matt Bevan reports that we each use one fifteenth of a teaspoon of water per prompt.

We knew what problem we wanted to address now. And we were left with the question: how do we ‘create more water’ to offset what is used?

It wasn’t easy

We turned to our industry peers and it seems no-one else really had solved it either. Many reported having the same challenges:

  • They want to strive to be competitive by taking advantage of the latest tech.
  • They recognise both positive and negative impacts – on the positive side, helping staff get time back and being more creative or enhancing their outputs; while on the negative side, acknowledging the environmental impacts that are still being realised.
  • They recognise that we are only half the equation and need transparent reporting from the major AI companies.

Despite not necessarily finding the answers we were hoping to uncover, we were still as determined as ever to stay true to our values and mitigate our own impact.

Keep Reading

Want more? Here are some other blog posts you might be interested in.