Generative AI is everywhere; shoehorned into every digital nook and cranny by companies desperate to demonstrate value from it. From web searches “enhanced” by Google’s Gemini, Microsoft Copilot’s insistence to rewrite your emails in Outlook, to LinkedIn offering to pen your latest industry insights – it’s difficult to escape the explosion of AI tools readily available to everyday consumers.
AI is notoriously resource-intensive, and energy politics is at boiling point with catastrophic flooding in Spain and apocalyptic wildfires in California, set against Donald Trump’s recent promise to “drill, baby, drill” and executive order to withdraw the USA from the Paris Agreement.
With many impact-conscious companies going to great lengths to track their Environmental, Social and Governance (ESG) metrics, is the proliferation of generative AI tools a risk to their climate goals?
AI’s insatiable appetite
It’s common knowledge that AI is power-hungry. Exactly how much, however, is something that is difficult for a consumer to find out. Providers are unwilling to divulge this information and estimates can vary widely.
And it’s not just power, AI is thirsty; the largest models run on entire data centres which often require evaporative cooling on an industrial scale to run effectively. This typically involves the use of potable water which can stress already-stretched water networks, with the water itself requiring energy to clean and transport.
In short, it’s relatively easy to power a data centre efficiently; just throw water at it! Yet the efficient use of water is harder to quantify. As Simon Ponsford at Tailpipe.ai, a specialist in cloud computing emission reporting, noted:
“Water Usage Effectiveness (WUE) is becoming an increasingly important statistic to measure. But at this time, the majority of data centres do not report their WUE figures.”
The lack of transparency from service providers is a nightmare for those interested in tracking their Scope 2 and 3 emissions but limiting access to the incredible potential of AI tooling could frustrate innovation in your organisation. Defining an acceptable use policy for public AI tools is a useful first step; exploring alternatives could be transformational.
DeepSeek R1 – a model model?
There are reasons to be optimistic. The enormous upfront cost for training a model has decreased dramatically over the past couple of years. Hardware efficiency gains continue to soar whilst challengers like DeepSeek claim to provide cutting-edge performance for a fraction of the investment of larger rivals. Distilled versions of these models could deliver workable results for even less capital.
Crucially, this can feasibly be done on consumer hardware. Running top-tier models – previously the exclusive domain of big tech companies – is now in reach of individual financial services firms. The future is pointing to smaller, specialised models (or “Small LLMs”) owned privately over larger, generic models owned by tech giants.
A brighter future
AI itself is also increasingly being used to manage data centre operations. Firms like Fluix use demand management models to responsively run cooling, lighting, and other utility systems at top efficiency which can reduce power and water impact. Eventually, residual heat from data centres could also be harnessed for uses ranging from heated swimming pools to water purification, turning waste into utility.
So, has the use of Google’s Gemini, Microsoft’s Copilot or any other AI tool dented your ESG credentials? Unfortunately, it’s hard to tell for certain. However, the future is pointing towards a democratisation of AI which will lead to lessened impact, greater transparency, and more comprehensive ESG reporting.
We can help
At Altus, we can work with you to produce a bespoke AI strategy for your business. We’re experts in financial services and can cut through the hype to assess the true value of AI.
Curious about AI but don’t know where to start? Using real-world use cases, separating fact from fiction, and providing actionable insights, our AI Bootcamp for Senior Leaders empowers senior Financial Services professionals with the necessary knowledge and confidence to leverage AI effectively and responsibly in your organisation.
