AI is the environmental bill events haven't budgeted for
Ellie Ashton-Melia takes a look at the environmental impact of AI in events.
The industry is finally starting to feel the full effects of climate change. You only need to read the recent AMI feature on the impacts of heatwaves on events to recognise that.
Climate change is no longer a future risk for events - it's already affecting when, where and how we meet. So you may well ask yourself, "Why are we accelerating a technology that produces soaring emissions and demands huge amounts of energy and water?"
Like many technologies, AI in events isn't just a productivity story, it's also an environmental one. And while the economic incentives far outweigh the unpredictable ecological and social cost right now, almost no one is considering the long term repercussions of its use.
According to a 2025 study by Forrester the industry is slow to adopt the technology, but the use-case is endless. In our rush to power venue sourcing, live translation, automating registrations and generating booth designs with AI, it's simultaneously drawing on enormous quantities of electricity, water and raw materials most of us never consider and few of us are questioning.
The environmental cost is invisible to every day users. Estimates suggest the carbon footprint for a simple text-based query is anything between 0.03g - 1.5g CO2e. Slightly less than sending the average email. Like a grain of sand on a beach, it is meaningless on its own, it's only when we consider the culmination of all those daily questions to an entire operation running on automation that we see the problem.
To understand the impact of AI we need to look at it in its entirety. We need to see the beach.
Ellie Ashton-Melia
Ellie Ashton-Melia
What does it actually take to make AI work?
AI is both a digital technology and a material system that requires resources. The recent surge in data centres expansion is largely down to AI. McKinsey reported that global data centre infrastructure expenditure is expected to reach $7 trillion by 2030. That equivalent investment could finance the entire global transport decarbonisation transition for three years. We take for granted that the “cloud” is tangible, it needs steel and concrete, it needs the mining of raw critical materials for the chips and these massive material inputs are neither resource-efficient nor infinite.
It currently takes 1.5 per cent of all electricity used worldwide to power AI, and this figure is expected to double by 2030, even as global emissions need to halve by then. As electricity infrastructure struggles to keep pace with the uninterrupted power demand, water-use only worsens the environmental footprint. Cooling these mega-units requires freshwater at a time when the globe is facing severe freshwater stress.
To put this in perspective, GPT-3's LLM training alone used an estimated 700,000 litres of water, while Google's data centres pulled 29 billion litres of fresh water for cooling in 2023. Meanwhile the companies running this infrastructure are the same ones with public net zero pledges. Microsoft's own 2026 environmental report shows its emissions climbed 25 per cent in a single year, driven largely by data centre expansion. Google and Amazon have disclosed similar upward trends.
A bigger problem at play is the often overlooked unintended consequences. This surge in infrastructure demand is straining power grids, draining water local supplies and squeezing out necessary housing, the extent to which that 41 mayors across six continents, including London, signed the Global Urban Data Centres Pact at London Climate Action Week to establish a framework ensuring data centres integrate sustainably into cities while safeguarding local climates and communities.
The human cost hiding in the numbers
In 2024, 78 per cent of organisations reported using AI in their work and 40 per cent of employers expect to reduce their workforce for task automation. With every role AI replaces even partially, shrinks the workforce. JP Morgan projects tens of millions of people could be unemployed within the next five years. That’s someone's income gone, a real livelihood affected, with a knock-on cost to public spending on unemployment that rarely makes anyone's ROI calculation.
You may think events, being in-person, physical experiences, are largely safe from the headlines about job displacement. That may be true, but the real concern is homogeneity. A recent UK survey by Digital Labs found that one in five designers say AI dulls creativity.
Events are built on connection, imagination and passion, fundamentally human traits. If we cut corners with copyrighting, design, or finding that unusual location to transform into an event space, events will start to look, sound and feel the same because the people who used to bring the difference aren't part of production anymore.
Can AI ever be “sustainable”?
AI should be built to meet real social and ecological needs, not just commercial benefit. This starts with disclosure. Currently there is no consistent agreed framework for measuring AI's emissions and resource use, an excuse technology firms use to under-report. Even if reporting is inherently complex with conflicting methodologies, disclosing honestly, even at the risk of over-reporting or double counting drives systemic accountability and exposes hidden ecological impact to mitigate.
Then there’s the question of infrastructure. If data centre growth is inevitable, developers have a responsibility to avoid water-stressed regions, construct fully renewable facilities that capture and reuse waste heat rather than pumping it straight into the atmosphere. Some of the engineering being developed to tackle this is inventive, from submerging data centres in deep seawater for passive cooling to utility-scale battery storage, but none of it scales fast enough to match the pace AI is rolling out.
Investing in this infrastructure now is fundamental to sustainable development as data centres become the backbone of our digital world. It shouldn’t be left to slower moving governments to develop legislation for faster-moving tech firms to take action.
Where does the events industry go from here?
Will we hit a point when enough is actually too much? I don't think a full reversal is likely, whatever the pockets of AI-rebellion suggest. If anything, we're seeing an appetite for in-person experience surge because people are craving something AI can't replicate. On a business level, that means asking a few uncomfortable questions.
Where can AI genuinely reduce impact?
While this won’t offset the ecological cost, AI can be used for positive environmental benefit. Venue energy management, automated waste tracking in commercial kitchens, smarter material use in graphic production, logistics route optimisation are real applications that meaningfully cut waste and emissions across the industry.
Is the time-saving exercise worth the trade-off?
If you can do the simple task or protect a skill by not using AI, then do it. There are so many cases where using a human is simply better. If you want something genuinely distinctive, hire the skillset that brings that value - otherwise we’re sliding toward homogenised mediocrity.
Be honest about how you use AI
Start to disclose publicly how AI is deployed across the business, how frequently, and what it's doing to roles within it. That transparency, multiplied across an industry, could reveal the true scale of adoption and its real impact on jobs and the environment, something no single disclosure can show alone.
Somewhere in all of this, businesses need to draw a line. AI may not be the most pressing environmental challenge our industry faces, audience travel emissions and event waste trump it all. This shouldn’t stop us from considering how or why we’re using artificial intelligence.
But consider this question: how much of an industry built on real experiences, made by real people, are we willing to run with fewer of those people behind it?
Ellie Ashton-Melia is a sustainability consultant at Not All Green (NAG) - a sustainability consultancy for the events industry.