Every AI prompt has an environmental cost. Are workplaces ready to talk about it?

For years, workplaces have been told to think before they print. Now they may need to think before they prompt.

That might sound dramatic. After all, asking an AI assistant to summarise a meeting or rewrite an email hardly feels like an environmental issue. There is no paper involved. No delivery truck. No factory smokestack.

Just a blinking cursor and a response that arrives in seconds. Yet as artificial intelligence becomes embedded in everyday work, a new reality is emerging. Every AI prompt requires electricity. Every chatbot response relies on data centres. 

Every image generated, document analysed and presentation drafted taps into a vast network of servers that consume power, water and computing resources.

The workplace productivity revolution has an environmental footprint. Most employees simply cannot see it.

The colleague that never sleeps

AI has quickly become the newest member of the workforce. It writes first drafts. It summarises calls. It analyses spreadsheets. It helps recruiters screen candidates, marketers brainstorm campaigns and managers prepare presentations. Unlike most colleagues, it never takes a lunch break.

Businesses are embracing these capabilities at remarkable speed because the benefits are tangible. Employees save time. Teams automate repetitive work. Organisations improve efficiency.

What receives far less attention is the infrastructure making those gains possible.

According to the International Energy Agency (IEA), data centres consumed around 415 terawatt-hours (TWh) of electricity globally in 2024, accounting for roughly 1.5% of worldwide electricity consumption. The agency projects that figure will more than double to approximately 945 TWh by 2030, with AI emerging as the single biggest driver of growth.

That is slightly more electricity than Japan consumes today. Not bad for a technology most people access through a chat box.

The cloud is not floating in the sky

One reason this conversation feels distant is because AI feels invisible. The language of technology does not help.

"Cloud computing" sounds light and weightless. Almost magical. In reality, the cloud is an enormous physical system. It consists of warehouses filled with servers, networking equipment, storage devices, cooling infrastructure and backup power systems.

These facilities work continuously to process and store information. The rise of generative AI is making those facilities significantly more energy intensive. According to researchers at MIT, training and operating large AI models requires far more computing power than traditional digital workloads. A generative AI training cluster can consume seven to eight times more energy than a typical computing workload.

Even after a model has been trained, the energy demand does not disappear. Every prompt requires fresh computation. Every answer requires processing power. Every interaction adds to the load.

Researchers cited by MIT estimate that a ChatGPT query may consume roughly five times more electricity than a standard web search.

One query is hardly a crisis. Several billion queries a day start looking very different.

AI has a water footprint too

Electricity is only part of the story. Data centres generate enormous amounts of heat. Keeping servers operational requires extensive cooling systems, and those systems often depend on water.

MIT researchers note that for every kilowatt-hour consumed by a data centre, approximately two litres of water may be required for cooling.

That means the environmental footprint of AI extends beyond electricity grids and into local water systems.

There is also the hardware itself. The rapid growth of generative AI has increased demand for specialised chips and graphics processing units, or GPUs. Manufacturing these components requires raw materials, energy-intensive production processes and global supply chains.

The AI boom, in other words, is not just a software story. It is increasingly an infrastructure story.

A sustainability blind spot in the workplace?

What makes AI particularly interesting from a sustainability perspective is that it remains largely invisible to employees.

Workers can see office lighting. They can see waste bins. They can see flights listed in travel expense reports. They cannot see the servers processing an AI request hundreds or thousands of kilometres away.

As a result, AI often escapes the sustainability conversations taking place inside organisations. Many companies now have governance frameworks covering AI ethics, privacy and cybersecurity. Relatively few have started discussing AI's environmental footprint.

That is understandable.

The technology is evolving quickly. Measurement standards are still developing. And for most organisations, the immediate focus remains productivity rather than resource consumption.

But as AI becomes woven into everyday work, the environmental conversation is unlikely to remain on the sidelines.

The irony at the heart of the debate

There is another twist. AI is not just consuming energy. It is also helping organisations use energy more efficiently.

According to the IEA, AI applications are already being used to optimise electricity grids, improve renewable energy forecasting, identify equipment faults and reduce industrial energy consumption.

Some of the potential benefits are significant:

• AI-based fault detection can reduce electricity outage durations by 30% to 50%.
• AI-enabled grid management could unlock up to 175 gigawatts of transmission capacity without building new power lines.
• AI-driven optimisation in industry could generate energy savings equivalent to more than the total energy consumption of Mexico today.
• AI-enabled building systems could deliver global electricity savings of around 300 TWh.

That creates an unusual paradox.

The technology is increasing electricity demand while simultaneously offering tools to reduce energy waste.

It is both part of the challenge and part of the solution.

The next chapter of responsible AI

The workplace debate around AI has largely focused on productivity, jobs, ethics and governance.

Environmental impact may be the next chapter.

That does not mean organisations should slow their AI ambitions. The benefits are too significant and the technology is becoming too deeply embedded in business operations.

It does mean companies may need a broader understanding of what responsible AI adoption looks like.

The future conversation may not simply be about how often employees use AI.

It may also be about where that AI runs, how efficiently it operates, what energy sources power it and whether organisations understand the hidden resources behind every prompt.

For now, AI remains the office's most enthusiastic new assistant.

But on this World Environment Day, it is worth remembering that every seemingly effortless answer comes from somewhere.

The prompt may feel virtual. The electricity, water and infrastructure behind it are anything but.