by Andrew Johnson, Training and Development Lead, Okana. Women in BIM AI Ambassador

Every few weeks, someone sends me the same article. AI is an environmental disaster, it drinks water, it burns electricity and we should all feel guilty for using it. As one of Women in BIM new AI Ambassadors, I hear that guilt most from the very people best placed to do something useful with AI: the digital construction professionals carrying the sector forward.

So let’s start honestly. The world is consuming more of everything, energy, water, raw materials and AI is adding to that demand. Less than the loudest headlines claim, as I’ll come to, but adding to it all the same. I won’t pretend otherwise.

Here’s the part the debate keeps missing. If you work in BIM and digital construction, you are not a passenger in this. You sit on one of the largest levers we have for cutting the built environment’s emissions and AI, aimed at the right problems, is how you pull it. The question worth your energy isn’t whether AI has a footprint. It’s what you choose to point it at.

The scare stats, in context

In August 2025, Google published the first detailed per-use figures for its Gemini model.¹ A median text prompt used about 0.24 watt-hours of electricity and roughly 0.26 millilitres of water, around five drops. That energy is less than nine seconds of watching television.

More telling is the direction of travel. Over the previous 12 months, Google’s figures show energy per prompt fell 33 times over, and carbon 44 times, while the answers improved. It’s the pattern Moore’s law taught us to expect, capability rising as cost falls, now showing up in energy too. That is the pace of change I keep pointing to, not wishful thinking.

This does not make the impact trivial. The International Energy Agency reports data centre demand grew around 17% in 2025 and could nearly double by 2030, reaching close to 3% of global electricity.²

Adding to a wasteful system

The waste we already tolerate is the part that gets ignored. The International Energy Agency notes that if every air conditioner bought worldwide since 2019 had been the most efficient model available, the energy saved would have offset much of the growth in data centre demand.² So the honest question is simple: does AI save more than it costs?

For our industry, the numbers point one way. The built environment accounts for around 37% of global carbon emissions, counting both operational and embodied carbon.³ Almost none of that is data centres. It’s the concrete, steel, glass, heating, cooling, and the material we specify and then throw away.

Used well, AI goes to work on exactly that. On our BIM models it can test low-carbon material options at the early design stage, where embodied carbon is locked in, model retrofit scenarios, balance heating and cooling, and cut the rework that wastes time and material on site. It’s through applications like these that Boston Consulting Group estimates AI could help mitigate 5 to 10% of global greenhouse-gas emissions by 2030, though it notes the figure rests on limited data.⁴

Who aims it?

I’m not waving the problems away. Training a frontier model is costly in a way a single prompt is not. Water and land footprints are real, and where you site a data centre matters.

A UN University report warns that by 2030 the water used by AI could meet the annual domestic needs of the 1.3 billion people in Sub-Saharan Africa.⁵ In June 2026 the UN Secretary-General told AI firms it was “time to come clean” on their full environmental impact.⁶ Anyone claiming AI is simply green is selling something.

So the honest position avoids both slogans. AI is not simply green, and it is not a disaster. The technology is moving fast, the cost per use is small and falling, and what counts is whether we point it at cutting emissions and resource use faster than it adds them.

The myth I’d retire first is the idea that using AI and taking sustainability seriously are opposing choices. For an industry responsible for more than a third of global emissions, the real risk is moving too slowly to aim it at what matters. And AI is something we choose how to use, which is why it matters who is in the room making those choices.

So here’s my question, as part of this Women in BIM community. If AI can help cut carbon and resource use in the built environment, and the early evidence suggests it can, which BIM or digital problem would you point it at first, on your projects or in your teams?

References

Google, Measuring the environmental impact of AI inference, August 2025.
International Energy Agency, Energy and AI, 2025.
UNEP and the Global Alliance for Buildings and Construction, Global Status Report for Buildings and Construction, 2023.
Boston Consulting Group, Accelerating Climate Action with AI, 2023.
UN University (UNU-INWEH), Environmental Cost of AI’s Energy Use, 2026.
UN Secretary-General, AI Environmental Transparency Initiative, June 2026.