During Sustainability Week 2nd – 4th March 2026, David Harrison, Head of Sustainability at Rathbones Asset Managment says the sustainability debate around AI is often framed in extremes, yet AI is not inherently good or bad. Like any tool, its impact depends on how it’s deployed, regulated and improved over time. Used well, it has the potential to help us do more with less.
The development of AI has been so swift in the past couple of years that it has pushed most of us off balance.
Many people describe AI as a revolution, which is exactly what it is. The technology is so pervasive, accessible and jarringly new that it forces people, businesses and nations to completely rethink how they work.
AI’s impact on people, supply chains and our most precious resources is highly complex and rapidly evolving. Because of this, I think that anyone with a neat answer and a final destination is making it up.
AI is so polarising because of the big changes it’s making to the workplace and our everyday lives. Yet AI is a tool, plain and simple. And from that flows all the complexity you already know and feel.
AI is neither good nor bad on its face. Instead, as with all tools, how we use it determines whether it’s a force for good or bad. That is the result of millions of individual decisions, big and small, made by businesses, governments and individuals.
AI is a tool, not a moral actor
Like everyone else, companies themselves are still coming to grips with making the best use of AI, yet they are so excited about using it because they can see the huge potential it offers.
We’ve met with healthcare companies that are using these next-generation tools right now.
If they can be more accurate in the design stage, and can use more simulation enabled by AI, that means better medicine without as much waste in the lab, with fewer and better trials. That should bring down the cost of drugs, too.
Finance businesses are far along with developing better fraud detection which could go a long way to reducing financial fraud.
In the UK it’s the single-largest type of crime, making up 40% of the total, and when it happens it can often involve life-altering sums of money.
I’ve seen cybersecurity programs tested that can intelligently spot suspicious transactions based on location, time-stamps and logic. Use your card to buy lunch in London, then, five hours later a fraudster tries their luck with your details in New York.
AI will reject it, knowing you can’t have travelled so far in that time. Try 10 hours later (with other, blisteringly fast, checks) the payment will go through. This could save countless millions of pounds and prevent untold emotional harm.
If these products and many more like them, continue to sprout from this era-defining investment in AI, it should dramatically improve people’s lives. It would make them healthier, safer and give them more free time from the drudgery of everyday tasks that can be outsourced to an AI agent.
Of course, all these changes come with trade-offs and problems that need to be worked through, minimised and offset where possible. Some people will worry that badly programmed AI tools could shut them off from credit.
That the same technology that can make developing cures easier brings with it privacy concerns for our most precious data. Where some people see freedom from work, others will see a loss of the identity that their careers bring them.
Why the data centre story is more complicated than it looks
Perhaps the most visible drawback of the AI rollout is the insatiable appetite for data centres, the chips that fill them, the power that fuels them, and the water that cools them.
There have already been countless reports of water stress, rising power prices and housing projects delayed because infrastructure capacity has been gobbled up by data centres.
Yet the greatest wave of investment is still ahead of us: You can see from our chart that, over the next five years, the four largest tech hyperscalers plan to spend $2.5 trillion more on AI capital expenditure than they did in the last five.
That’s roughly equivalent to the GDP of Italy, the world’s eighth-largest economy.
Source: FactSet, Rathbones; 2021-2025 actual capex, 2026-2030 are company forecasts, Microsoft’s 2030 capex is assumed same capex as 2029 (its farthest-out forecast)
But this isn’t the whole story. AI is also being deployed to more efficiently manage those resources in the wider economy. Like the electrical engineering companies are rolling out smartgrid technology that reduces wasted energy by 15-30%.
Or sensors and software that help water companies locate and fix leakages that litter our pipe networks. Or tractor and GPS companies which help farmers apply nutrients and water with pinpoint precision, delivering greater yields with far fewer inputs.
And while data centres have sucked up a lot of power, tech companies have added a huge amount to the grid as well. Absolute demand is still rising fast, yet AI itself has helped reduce the energy intensity of each unit of compute by improving the design of chips and cooling systems.
These improvements are in the realm of 30-50% reductions in power use. About half of the 56 gigawatts of clean energy global corporate long-term power agreements in 2025 were signed by the big the four AI hyperscalers (Microsoft, which we own, and Alphabet, Meta and Apple, which we don’t).
These multi-decade agreements are crucial for backing new power projects, as they lock in sales that can be leveraged to gain financing to install new plants.
Last year, the US built more power generators and storage than in any other of the past 20 years. Renewables accounted for about 60% of that, mostly solar.
As for water, yes, data centres need a whole lot. And their usage is still increasing fast. But again, they have become ever better at minimising usage for each unit of computing power they deliver, through using a mixture of better technology, non-potable water, reusing old water and more careful site selection.
One area where big tech companies fall down, however, is in transparency of water usage.
Data centre sustainability is one of our priorities for company engagement in 2026. Some politicians are already highlighting the problems that the huge expansion in data centres is having in some areas on power prices and water usage.
Companies need to get ahead of this and show how they are mitigating their impacts, while also showing how they are some of the largest backers of new clean energy projects out there, which is good for all of us and the planet to boot.
Main image: AI, green, zach-m-hVEx0yZSQWY-unsplash
