Artificial Intelligence (AI) is rapidly transforming various aspects of our lives, and the fight against climate change is no exception. However, the potential impact of AI on our planet remains a topic of debate, and last week was no exception with the signing of Microsoft and Occidental’s carbon credit deal. As reported, Occidental, one of the biggest US oil producers, will sell 500,000 carbon credits for an undisclosed amount to Microsoft over six years in a bid to help offset the firm’s AI energy surge.
As we know, AI excels at processing vast amounts of data. It can improve weather forecasting, optimise resource use in agriculture and identify pollution sources with great accuracy – just to name a few ways in which it may impact the planet. But on the other side of the coin, AI requires large amounts of energy and capital investments. Training and running an AI model requires significant computing power, leading to higher energy consumption and water usage – potentially negating its environmental benefits.
As a case in point, Google’s greenhouse gas emissions have surged 48% since 2019 due to the expansion of its data centres with AI systems. On top of this, Microsoft’s have risen 30% since 2020. This leaves both companies commitments to get to net zero by 2030 in jeopardy. By 2027, it is predicted that AI will be using as much energy as the whole of Argentina or Sweden.
See also: Regulatory burden must drive responsible investment industry to look further afield
Now, greenhouse gas (GHG) emissions are not the most material sustainability issue for large tech companies, which generally have very low emissions intensity compared to other industries. But companies like Google and Microsoft can more easily transition to renewable sources because software development and data warehousing is much easier to decouple from carbon emissions, compared to steel or concrete production. So, if a commercial opportunity like AI is causing these high-profit and easy-to-decarbonise companies to abandon their emission reduction trajectories, why should we expect the industries that need much more radical change to stay committed?
This should not be a finger-pointing exercise, but Microsoft and Google are examples of a dynamic that challenges the overall transition to a sustainable economy: we’ll only do it, as long it doesn’t get in the way of profitability.
For sustainability-minded investors, this has serious implications. Those who track Paris-Aligned Benchmarks or Climate Transition Benchmarks in their investment strategies commit to ensuring an overall emissions reduction of their portfolio companies year over year. This has already tilted many of these portfolios towards tech and finance. As tech companies increase their emissions, investors run out of targets that fulfil the requirements for these strategies. Moreover, investors are left with a more fundamental question, which is how to weigh the pros and cons of new solutions that may benefit society, but initially increase the emissions of the companies that develop them.
When it comes to steel or electric vehicle production, investors may have had a model for assessing contributions and harm towards sustainability objectives, enabling them to weigh operational issues against product benefits. But when it comes to AI model development, everyone is currently in the dark: how do we weigh an explosion in energy consumption with the potential benefits of better weather forecasts, or productivity gains in white-collar work?
The answer 10 years ago would have been to develop yet another ESG rating methodology that attempts to balance these concerns against each other. Today, the financial industry relies on multi-stakeholder and regulatory frameworks to categorise and assess trade-offs. But it takes time to develop such frameworks, and until then, investors are left to decide for themselves just how much ‘bad’ they can accept from portfolio companies in their search for the next good. Their existing commitments to clients and regulators might already hold some of that answer.
