The paradox of decarbonising AI

ARTIFICIAL intelligence (AI) holds immense potential to drive decarbonisation, by offering revolutionary benefits such as optimised energy consumption and speeding up the deployment of renewable energy.

However, this potential is not without its challenges, because running AI operations entails the substantial use of energy and water – a paradox which puts investors in sustainable practices in a difficult position, given that they must balance the push for AI-driven transformation with the risk of increased carbon emissions if AI systems are not sustainably powered.

In the realm of investing, this issue has intensified. Investors face the dual burden of decarbonising investment portfolios while ensuring profitability. Explicit emissions targets or “budgets” often require rebalancing portfolios to stay within set carbon limits. Yet, as investments in AI technologies grow, so will the complexity of managing their environmental impact.

Environmental impact

Training and operating advanced AI models consume enormous amounts of energy. The International Energy Agency (IEA) has noted that a single Google search consumes 0.3 watt-hours of electricity, but a ChatGPT request requires 2.9 watt-hours – pointing to the large difference in computational expense.

But while both figures appear small in isolation, there are profound environmental implications at scale.

Companies across industries are rapidly integrating AI for diverse applications, ranging from customer-service chatbots to AI-powered analytics tools that boost efficiency and decision-making. For instance, in Singapore, an AI utility was launched to combat greenwashing by using nationwide green building to support decisions on sustainability-linked loans.

As the deployment of AI increases, so does the global demand for energy, leading to burgeoning electricity consumption. This demand has begun influencing policy decisions; the US, for example, has delayed the retirement of coal plants so it can meet rising electricity needs.

Mitigating AI’s own carbon footprint

To effectively address the AI-decarbonisation paradox, an all-encompassing approach is crucial – by aligning the development and deployment of AI with sustainable practices.

One logical strategy is to power AI infrastructure with renewable energy. Notably, the global green data-centre market is predicted to grow significantly to US$279.5 billion by 2032, from US$81.1 billion in 2024, largely driven by regulations encouraging renewable energy use in data centres.

SEE ALSO

Can Singapore balance its ambitious AI goals with sustainability?

S’pore’s climate targets still ‘highly insufficient’, even after clean energy push: Research group

Singapore to study viable decarbonisation strategies in push for more sustainable urban environment

Interestingly, AI itself could mitigate its environmental repercussions. Although it may seem counterintuitive, advancements in AI and machine learning might offer solutions to their own energy demands.

Reports project that AI and machine learning will play a vital role in efficiently powering and cooling data centres worldwide. For example, Google’s collaboration with DeepMind on algorithms and predictive modelling to reduce cooling power needs is a step towards sustainable solutions. Enhancing AI’s energy efficiency, particularly in generative AI that requires substantial computational power, is critical for energy security and the scaling of AI applications.

AI’s potential also extends to climate intelligence by optimising energy grids, amplifying industrial efficiency and bolstering the deployment of renewable energy sources.

Synchronising wind turbines using AI, for example, can ensure optimal performance by aligning them with wind conditions – thus integrating technological advancements with environmental sustainability. Ultimately, the development of nuclear fusion reactors will almost certainly rely heavily on AI to maintain the stability of the fusion plasma.

To support decarbonisation solutions, the European Union (EU) has introduced a series of incentives to champion AI-powered decarbonisation initiatives under the EU Green Deal and Energy Efficiency Directive. Among these incentives are substantial financial grants from programmes such as Horizon Europe, which has a budget of 95.5 billion euros (S$137.8 billion), targeting AI-driven projects that contribute to the EU’s climate objectives.

Driving sustainable outcomes

From an investment perspective, capitalising on AI for decarbonisation presents a significant opportunity in the transition to a low-carbon economy, assuming that the underlying AI systems are sustainably powered.

Shareholder proposals focused on responsible AI have gained traction, particularly at industry giants such as Microsoft and Alphabet. Following a shareholder proposal filed in 2023, Microsoft published its inaugural Responsible AI Transparency Report this year and expanded the team responsible for ensuring its AI products are safe. Investors can similarly engage with companies to drive substantive change, advocating for commitments to renewable energy, improvements in energy efficiency, and transparent carbon footprint reporting.

Investing in AI companies that prioritise transparency and climate intelligence ensures these innovations are responsible and sustainable. Impact investments are key to advancing AI-powered decarbonisation initiatives. By directing funds into advanced technologies across diverse sectors such as agriculture, utilities and energy, these investments are driving AI solutions that address critical issues such as biodiversity conservation, climate-change modelling, transparency in AI model management and the circular economy.

While the inevitable rise in AI investments continues, a balanced approach is vital. Hyperscalers will be challenged to adopt energy-efficiency and renewable-energy measures to keep their place in low-carbon investment portfolios. Simultaneously, we may see investors and policymakers take an active role in promoting responsible and sustainable AI applications soon. We remain optimistic that a good balance between advancement in AI and sustainable energy usage can be achieved.

Mika Kastenholz is head of investment solutions for Asia-Pacific at LGT Private Banking