Like most of my clients, I work in a values-driven sector. And also like many of my clients, my mind is constantly alert to our ever-deepening environmental crisis. 

I’ve sat with one particular concern recently, an environmental and moral concern borne out of my recent use, both in work and in my personal life, of AI.

Numerous of my charity clients are environmental, justice-oriented, mission-led. They exist to serve a better world, not just a more efficient or profitable one. And even in these spaces, like in mine, the quiet march of AI is undeniable. 

The Charity Digital Skills Report 2024 found that 61% of charities were using AI daily. 

AI has many obvious benefits for fundraising teams, bids teams and policy teams. ‘Speech to text’ tools can ensure every meeting is properly minuted. There are also countless opportunities to use AI directly in service delivery and the creation of charitable projects - Surrey Wildlife Trust, for example, has undertaken a three-year Space4Nature programme that uses satellite earth observation imagery, combined with volunteers’ observations and AI, to help map and assess habitats across Surrey.

No one’s under illusion - we know AI isn’t neutral. But most conversations currently happening in the workplace are concerned with the question of bias, or automation or the potential impact on human jobs. 

The question I am most concerned with is: what powers the machine? And how can I become comfortable with any use of AI when I know the environmental damage any use will cause? 

Artificial intelligence runs on electricity, in amounts that grow with every new model, every new user, and every new functionality added. And I cannot ignore that serious ethical questions are raised by the fact that the tools we have built to serve humanity may end up deepening the harm we meant to heal.

So here I am, someone shaped by climate activism, committed to social good, and working daily with organisations trying to hold both. Writing an article, many of my activist friends would wholly disapprove of being written – a defence of nuclear power.

The hidden carbon cost of AI

For all its abstraction, AI is not ethereal. It’s physical. It’s powered by vast data centres that draw electricity on an industrial scale, constantly. Every moment of the day. In vast quantities. It is drawn with every prompt, every question, every response.

Right now, most of that electricity comes from grids still soaked in fossil fuels. Countries heralding AI as the future are quietly fuelling it with gas, coal, and oil. The result is a strange moral contradiction: we turn to AI to solve complex human problems - health, justice, climate resilience - while letting it accelerate the environmental crisis we claim to care about.

Energy use by artificial intelligence currently only represents a fraction of the technology sector’s power consumption and is estimated to be around 2 to 3% of total global emissions. If AI application use continues to expand at a rapid rate, the sector could account for 32% of a country's total electricity demand by 2026. And without a radical shift in how we think about energy, not just in volume, but in ethics, AI risks becoming a beautifully coded hypocrisy.

There are already some who are refusing to engage with AI for this very reason, but I think it unlikely that the freedom to refuse will long be maintained if our employers believe (rightly) that the use of AI is an advantage and, eventually, a necessity.

The charity sector prides itself on transparency, integrity, and justice. But if we ignore the carbon cost of the tools we adopt, those values start to ring hollow. And the sector will lose its power to hold others accountable, because it too will be compromised.

Why solar and wind alone aren't enough

In no way would I ever wish to dismiss renewables. I want them to succeed. But whilst solar and wind are essential, I can see they are not sufficient. Not for the scale and speed we now face. For one, these sources are intermittent by nature. The sun doesn’t always shine, the wind doesn’t always blow. For two, the storage technology we need to bridge those gaps isn’t reliable or scalable enough (yet). Meanwhile, AI systems don’t politely pause when clouds roll in. They demand constant, high-voltage energy 24/7, 365 days a year.

Some argue we just need more time - more investment, more panels, more batteries. But time is the one thing the climate doesn’t have. And when demand surges faster than capacity, the fallback is never idealism. It’s fossil fuels.

That fall-back is already happening.

Tech companies are signing deals for energy in regions where grids are carbon-heavy. Data centres are being located close to cheap gas. And because no one wants to talk about it, it is easy to let oneself believe a fiction: that AI will somehow run clean by default. 

But hope is not a power source, would that it were. Pretending solar can do it all is not a strategy; it’s a way to stay comfortable while emissions rise.

If we want to power AI responsibly, we need to tell the whole truth. And the truth is, we need a stable, zero-carbon backbone to the energy system. And nuclear is the only option available to us for that.

Keir Starmer has said that he wants the UK to be a world leader in the use of AI. Labour has announced its plans to use AI to ‘turbo-charge’ national renewal. Essential that the UK’s plans to vastly expand its nuclear power capabilities receive the necessary support to see the plans come to fruition. Because without it, I and many others will continue to hold AI at arm’s length - not from fear, but because ethical progress demands ethical power.