The Environmental Cost of AI: What Businesses Need to Know
Artificial intelligence has quickly become part of everyday business. It can write content, analyse information, generate images, summarise documents, answer questions and help businesses automate repetitive tasks. For many organisations, AI is already changing the way work gets done.
But there's something worth remembering whenever we open an AI tool and type a prompt; AI has an environmental cost.
That doesn't mean we should stop using it. It does mean we should understand what sits behind the technology and think about when using AI is genuinely worthwhile.
There is also another side to the story. AI has the potential to help reduce environmental damage in areas such as energy management, agriculture, transport, climate modelling and environmental monitoring.
So it's not a question of whether AI is good or bad for the environment. The reality is more complicated than that.
AI is digital, but its environmental impact is physical
It's easy to think of AI as something that exists somewhere in the cloud. In reality, the cloud is made up of physical infrastructure.
AI models are trained and operated in data centres containing large numbers of servers, networking equipment, storage systems and cooling infrastructure. These facilities require electricity to operate and, depending on their design and location, can also require significant amounts of water for cooling.
The environmental impact does not begin and end with the electricity used when you type a prompt.
There are also impacts associated with manufacturing computer chips and other hardware, extracting the raw materials needed to make them, transporting equipment and eventually disposing of it.
The OECD describes the environmental footprint of AI as something that needs to be considered across its entire life cycle, including energy, water, greenhouse gas emissions, critical minerals and end-of-life impacts. That is an important distinction.
When we talk about the environmental impact of AI, we are talking about more than electricity.
How much electricity does AI use?
This is where things can become confusing.
You'll often see claims online about how much electricity is used by a single AI prompt. The problem is that there is no single figure that applies to every prompt. Different AI models require different amounts of computing power.
The type of task also matters. A simple text question is very different from generating a high-resolution image, creating video or asking an AI system to perform a complex multi-step task.
The good news is that AI is becoming considerably more efficient.
The International Energy Agency's 2026 analysis found that the energy required for an individual AI task has been falling rapidly as hardware and software improve. It reports that simple text queries now typically use less electricity than running a television over the same period, which is encouraging news for anyone who cares about the environment.
But there's a catch 😏.
More people are using AI, more frequently, while increasingly sophisticated applications can require considerably more computing power. The IEA reports that energy-intensive applications such as video generation, reasoning and AI agents can consume hundreds or thousands of times more energy than simple text generation.
So greater efficiency does not automatically mean lower overall energy use.
If each task becomes cheaper to run but the number and complexity of tasks increase significantly, total demand can still rise.
Data centres are becoming a bigger part of the energy picture
AI isn't the only thing using data centres. They also support cloud computing, websites, streaming, online services and many other digital activities. However, AI is becoming an important driver of growth.
The latest IEA analysis estimates that global data centre electricity consumption was around 485 TWh in 2025. It projects that this could roughly double to 950 TWh by 2030, equivalent to around 3% of global electricity demand. Electricity consumption from AI-focused data centres is expected to grow even faster, tripling over the same period.
The IEA also reports that global data centre electricity consumption increased by 17% in 2025, while electricity consumption from AI-focused data centres increased by around 50%.
These figures refer to data centres as a whole and to AI-focused data centres respectively. They should not be interpreted as meaning that AI is responsible for all data centre electricity consumption; an important distinction.
And then there's water
Electricity is only part of the story. Data centres generate a lot of heat and need cooling systems to keep their equipment operating safely. Some cooling systems use water.
The amount varies considerably depending on the technology, climate, data centre design and source of electricity.
This makes broad claims such as "one AI query uses X amount of water" difficult to apply universally.
It's better to understand the principle: AI infrastructure can have a water footprint as well as an energy footprint.
Water can also be involved further up the supply chain, including in the manufacturing of semiconductors and other components.
In areas where water resources are already under pressure, the local impact of a new data centre can therefore be much more significant than a global average might suggest. The OECD identifies freshwater use as one of the environmental factors that should be considered when assessing the full lifecycle impact of AI compute.
Hardware has an environmental cost too
It's tempting to focus entirely on what happens when we use AI. But before an AI model can answer a question, someone has had to build the hardware that runs it. Advanced processors require raw materials, manufacturing facilities, energy and complex global supply chains. Eventually that equipment needs to be replaced.
Electronic waste is already a significant environmental issue and the rapid development of computing hardware adds another challenge.
The OECD therefore recommends looking beyond operational electricity consumption when assessing AI's environmental footprint. Critical mineral extraction, freshwater use and electronic waste all form part of the bigger picture.
But AI could also help the environment
In the interests of balance, it's important to look at the positive potential impacts of AI on the environment.
The same technology that requires significant computing resources can also be used to make other systems more efficient.
AI can analyse huge amounts of information quickly and identify patterns that would be difficult for people to find manually. That has applications across many areas of environmental management.
Energy
AI can help forecast electricity demand, optimise power generation and improve the operation of energy systems.
It can also help integrate renewable energy sources by improving forecasting and managing fluctuations in supply and demand.
The IEA estimates that widespread adoption of existing AI applications in the energy sector could lead to emissions reductions that are significantly larger than the emissions associated with data centres themselves, although there are important barriers and uncertainties around achieving those gains.
Buildings
Buildings consume significant amounts of energy. AI can help analyse patterns in heating, cooling and electricity use and adjust systems according to demand.
That could reduce wasted energy while maintaining comfortable conditions for occupants.
Transport
AI can be used to optimise routes, manage traffic and improve logistics.
For businesses operating fleets or delivering goods, better route planning can potentially reduce unnecessary mileage, fuel consumption and associated emissions.
There are also potential applications in public transport and infrastructure management.
However, the environmental outcome is not guaranteed. If AI makes certain forms of transport cheaper or more convenient and this leads to increased overall consumption, some of the potential environmental benefit could be lost. This is known as a rebound effect.
Agriculture
AI can help farmers analyse weather, soil conditions, crop health and other information.
This can support more targeted use of water, fertiliser and pesticides. Precision agriculture is one example of how analysing more information could help reduce waste while maintaining productivity.
Environmental monitoring
AI can also be used to analyse satellite imagery, weather data and other environmental information.
Applications include monitoring forests, identifying changes in ecosystems, predicting weather and improving climate modelling.
AI is not automatically environmentally friendly
This is an important distinction.
It would be just as misleading to describe AI as an environmental solution as it would be to describe it as purely an environmental problem. An AI system can create environmental costs and environmental benefits at the same time.
What matters is what the technology is being used for and what happens as a result.
If an AI system consumes resources to produce something of little value, the environmental cost may be difficult to justify.
If the same technology helps reduce substantial energy use, waste or emissions elsewhere, the calculation looks very different.
The OECD describes this as a distinction between the direct impacts of AI compute and the indirect effects of AI applications. The direct impacts are generally associated with resource consumption, while applications can have either positive or negative environmental effects.
The IEA similarly concludes that AI could make a significant contribution to reducing emissions, but that this potential depends on successful adoption and overcoming practical barriers.
So what does this mean for businesses?
It doesn't mean businesses should stop using AI; it means businesses should use it thoughtfully.
Before introducing an AI tool, ask what problem you're trying to solve:
Will it save a significant amount of time?
Will it improve the quality of your work?
Will it help your customers?
Will it reduce waste or unnecessary work?
Could you achieve the same result using a simpler method?
These questions are useful from both a business and environmental perspective.
There is little value in generating hundreds of AI images simply because you can. There is more value in using AI to help produce a useful campaign that would otherwise require significantly more time and resources.
Likewise, if a simple piece of work can be completed effectively without AI, there may be no reason to use a more resource-intensive technology.
The most sustainable approach is not necessarily to use less AI
It is to use AI more deliberately. That means understanding that every AI request involves physical infrastructure somewhere.
It means recognising that some tasks are relatively lightweight while others require considerably more computing power.
It means questioning whether the output is actually useful rather than generating content simply because it is easy to produce.
And it means looking beyond the environmental cost of AI itself to consider what the technology enables.
If AI helps a business reduce unnecessary travel, improve energy efficiency, reduce material waste or optimise logistics, its wider environmental impact may be positive.
If it simply encourages us to generate more content, consume more resources and automate tasks that did not need automating in the first place, the picture is very different.
AI is a tool. How we use it matters.
The environmental debate around AI is likely to continue as the technology develops.
Models are becoming more efficient. Hardware is changing. Data centres are expanding. New applications are emerging and the energy mix used to power them is changing.
That means there is no single number that can tell us whether AI is environmentally good or bad.
What we can say with confidence is that AI has an environmental footprint. Every use of AI depends on physical infrastructure and resources.
At the same time, AI has genuine potential to help us use other resources more efficiently and address some environmental challenges.
The sensible approach is neither to ignore the environmental cost nor to assume that using AI is inherently harmful.
It is to understand the trade-offs and use the technology where it provides genuine value.
AI can be a useful business tool. But like every other technology we use, it is worth understanding what sits behind it.
Our approach to AI
AI is a useful tool, but we don’t believe everything needs to be done with AI simply because it can be.
At Black Rocks Marketing, we use AI thoughtfully for relatively lightweight tasks, practical problem-solving and logical automations where it can genuinely improve the way we work.
As environmental champions, we’re also conscious that AI has an environmental cost. We avoid using it for unnecessary content or images for the sake of it (no one really needs an image of their cat as Henry VIII or to see what they'd look like if they were bred from a pineapple and a grater). Our mantra is that images shouldn't be generated simply because they can be, rather than because they serve a useful purpose.
We dislike the 'AI made this' aesthetic of corporate social media posts (we're not artists, but we appreciate art). These things need soul!
We’ll continue to explore where AI can add real value while remaining mindful of its wider environmental impact.
Sources cited in this article
IEA, Key Questions on Energy and AI (2026)
IEA, Key Questions on Energy and AI: Executive Summary (2026)
IEA, Energy and AI (2025)
OECD, Measuring the Environmental Impacts of Artificial Intelligence Compute and Applications: The AI Footprint (2022)
OECD, Digitalisation and the Environment
