Is AI Sustainable? Perhaps That's the Wrong Question?
When I agreed to oversee an intern, I was genuinely excited about the opportunity. Sustainable investment has been my career for nearly 30 years and I was looking forward to sharing some of that experience, helping someone develop their knowledge and hopefully encouraging a passion for responsible investing.
What I wasn't expecting was how much it would challenge me in return, prompting me to revisit assumptions and reflect on how my own thinking has evolved over the years.
My intern is bright, thoughtful, values-driven and, perhaps most importantly, willing to challenge things. Rather than accepting information at face value, he asks questions. Good questions. The sort of questions that make you stop, think and occasionally realise you've become so familiar with a topic that you've stopped looking at it with fresh eyes. In an increasingly complex sustainability landscape, and one subject to growing scrutiny around greenwashing, that ability to challenge claims, interrogate evidence and look beyond the headline has never been more important.
His approach reminded me a lot of myself when I first started out.
Back then, sustainable investing was generally referred to as ethical investing. The debates felt simpler. Tobacco, gambling and armaments were often the focus. Nobody was discussing artificial intelligence, data centre energy consumption or the sustainability implications of semiconductor manufacturing.
The world has changed, and sustainable investing has evolved alongside it.
Alongside my investment career, I also spent many years working in social work. One thing both professions taught me is that people often begin their careers as idealists. They want to make a positive difference and can see issues in fairly clear terms. Over time, however, experience introduces complexity. Compromise becomes necessary. Grey areas emerge. If we're not careful, a little cynicism can creep in too.
Spending time with someone at the beginning of their professional journey reminded me how valuable that early optimism can be.
As part of his internship, I asked him to review the funds within our responsible portfolios and identify sustainability themes, impact stories and areas worthy of further exploration.
When he came back, one observation stood out.
Many of the funds had significant exposure to artificial intelligence, particularly through holdings such as NVIDIA.
His conclusion was straightforward:
"AI is unsustainable."
To be fair, there was a lot in his argument that was difficult to disagree with. Data centres consume significant energy. AI training models are resource intensive. Cooling systems can require large volumes of water. Semiconductor manufacturing has environmental impacts. These are all legitimate concerns and exactly the sort of questions sustainability professionals should be asking.
Rather than answering the question directly, however, I encouraged him to reframe it.
Instead of asking:
"Is AI sustainable?"
Perhaps the better question is:
"Under what circumstances does AI create a net sustainability benefit?"
That subtle shift changes the whole discussion. One of the biggest lessons I've learned over the years is that sustainability is rarely black and white. Progress in one area can create challenges elsewhere.
Renewable energy provides a useful example.
Solar panels require mining, manufacturing, transport and land use. Wind turbines require steel, concrete and rare earth materials. Neither technology is impact-free. Yet most sustainability professionals would conclude that the long-term benefits outweigh those impacts.
The question is not whether something has an environmental footprint.
The question is whether that footprint is justified by the value it creates.
The same principle can be applied to AI.
The more we talked, the more I encouraged him to look beyond just the environmental dimension of ESG. Coming from a social work background, the social element has always been particularly important to me and ESG was never intended to be solely about environmental outcomes.
What if AI helps identify cancers earlier?
What if it accelerates drug discovery? What if it improves accessibility for people with disabilities?
What if it creates educational opportunities through personalised learning?
What if it helps improve workplace safety or advances scientific research?
What if it improves access to mental health support for people who might otherwise struggle to receive help?
This is one area that particularly interests me, given my social work background and the reality that demand for mental health support increasingly outstrips the services available to provide it. Could AI play a role in improving access to support and information? Whilst AI is certainly not a replacement for trained professionals, it may help people access guidance, monitor their wellbeing, seek help sooner or find support outside traditional service hours. For someone struggling at two o'clock in the morning, immediate access to support may be far more valuable than waiting days or weeks for an appointment.
Suddenly the conversation becomes less straightforward.
Rather than asking whether AI consumes resources, we start asking whether the benefits it creates justify those resources.
How many additional cancer diagnoses?
How many medical breakthroughs?
How many accessibility improvements?
How many people receiving timely mental health support?
How many lives improved?
These are not easy questions to answer, but they are often more useful questions to ask.
The irony, of course, is that AI may also become part of the solution to some of the sustainability challenges associated with its own growth. It is already being used to optimise electricity grids, improve energy efficiency, support climate modelling, enhance battery storage and reduce waste across numerous industries.
Multiple things can be true at the same time. AI can consume significant resources whilst also helping to solve significant sustainability challenges.
Our conversation eventually moved beyond sustainability theory and into the practical realities of investment management. This is often where idealism encounters complexity.
In practice, sustainability considerations need to be balanced alongside fiduciary duty, diversification, risk management, value for money and, ultimately, delivering good outcomes for clients.
Every additional exclusion reduces the available investment universe. Whilst exclusions can absolutely be appropriate, they also have consequences. NVIDIA became a useful example.
Even if concerns exist around the environmental impacts of AI, excluding one of the most influential companies in global equity markets has implications for diversification, portfolio construction and long-term returns.
I found myself in the slightly uncomfortable position of defending NVIDIA, something that the younger version of me may have found quite surprising. I reassured myself that experience had made me more nuanced rather than simply selling out. I couldn't quite bring myself to apply the same defence to Meta, however. Even after all these years, there are limits to my professional maturity.
Jokes aside, this is where stewardship and engagement become particularly important. Rather than automatically excluding companies, investors can consider whether improvements are more effectively achieved through active ownership and engagement whilst continuing to access the benefits those companies may bring.
One of the biggest lessons sustainable investing has taught me is that it is often less about finding perfect companies and more about managing imperfect trade-offs.
That's not always comfortable, particularly for people entering the profession with strong values and a genuine desire to do the right thing. In truth, it's not always comfortable for those of us who have been doing it for years either.
The transition from asking "What do I think about this company?" to "What is in the best interests of thousands of underlying clients?" is often where idealism meets the realities of professional responsibility.
Looking back, I realised this was never really a conversation about AI. It was a conversation about curiosity, critical thinking and being willing to challenge assumptions. It was about exploring multiple perspectives before reaching conclusions.
My intern's concerns weren't wrong. In fact, they were exactly the sort of concerns sustainability professionals should be asking about. The challenge is making sure we apply the same scrutiny to all sides of an argument and remain open to the possibility that several things can be true at once.
The longer I work in sustainable investing, the more I find myself moving away from absolute answers and towards better questions.
Looking back, I started the internship thinking I would be helping someone learn about sustainable investing. Instead, I was reminded why I became passionate about it in the first place. I may have been the mentor, but on this occasion, I suspect I learned just as much as, hopefully, my intern did. In many ways, the lesson was my own: never become so experienced that you stop asking the questions that mattered when you first started out.
