WEF charts roadmap to sustainable AI – and calls for collaboration

In a new paper series released at its Davos conference, the World Economic Forum suggests a collaborative roadmap to ensure AI’s large-scale deployment doesn’t cause collateral damage to the environment.

Artificial intelligence is one of the most talked about topics at this year’s WEF conference in Davos, Switzerland. Speakers and delegates are particularly interested in how to solve AI’s energy conundrum: while the technology can help reduce energy consumption by up to 60% in certain use cases, the electricity needed to power it is projected to grow by 50% every year until 2030.

Read also: Generative AI - Sustainability game changer or ticking time bomb?

“Partnerships are critical to unlock AI’s potential to drive transformative innovation, while responsibly addressing its environmental impacts,” noted Google’s Chief Sustainability Officer Kate Brandt upon announcing that she will be speaking on several panels at the conference.

Balancing AI’s energy challenges and opportunities

In a paper launched at Davos in collaboration with Accenture, WEF explores what it calls “AI’s energy paradox”: AI applications can help optimise energy efficiency and reduce power consumption, but its own electricity demand is rising sharply – with data centres expected to use 12% of US electricity generation within four years.

AI growth is already jeopardising climate targets for Big Tech: Microsoft’s carbon footprint jumped by nearly 30% in 2023 due to data centre expansion, and Google, Meta, Microsoft and Salesforce are now investing heavily in carbon credits to help offset this impact.

To help understand AI’s evolving energy impact, WEF recommends monitoring the ways AI is deployed for decarbonisation, but also transparently communicating on its electricity use and associated emissions – something that is currently not a given.

Five challenges to sustainable AI infrastructure

In its Blueprint for Intelligent Economies, WEF outlines five key challenges in building sustainable AI infrastructure: namely its high energy consumption and environmental impact, the scale of investments required, the lack of resilience in AI supply chains, the growing digital gap and the currently high cost of internet services.

“To minimise the environmental impact of AI, it is crucial that the rapid expansion in data centres is powered by sustainable and responsible energy sources,” the Forum notes, warning that the infrastructure investment required to make that a reality “will not be affordable for most countries”.

The report shares several examples of successful initiatives to overcome these challenges. For example, regional sharing of data centre capacity via clusters and arrays, as well as tax incentives and financial grants for private sector investment can help reduce the investment barrier for green AI infrastructure.

Similarly, collaborative agreements between data centre firms and energy providers can support the deployment of zero-emissions energy such as wind, solar or nuclear to power AI’s massive compute demand.