Enhancing AI digital infrastructure within planetary boundaries

Every time you ask an AI system a question, a data centre somewhere draws electricity and water to compute your answer. As artificial intelligence becomes woven into the fabric of daily life – from healthcare diagnostics to financial planning to creative work – we as a global society face the challenge of making sure we can meet the energy and water demands of the technological revolution while simultaneously protecting the world’s precious resources.

Data centres are the beating heart of AI, and they consume up to 1.5% of global electricity and vast volumes of water for cooling. According to the International Energy Agency (IEA), global electricity demand from data centres is projected to double in the next five years, driven largely by AI.

Water use tells the same story: US data centres consumed 17.4 billion gallons of water for cooling in 2023 and this is projected to quadruple by 2028.

Water scarcity affects 2.1 billion people globally, making resource-conscious AI development and reusing water resources not just an environmental issue, but a social and economic imperative.

To enhance the value of digital infrastructure investments, and to address the full range of community concerns related to these projects, we believe there are five priorities:

With strong payback periods – one study citing less than two years for facilities beyond 7MW – waste heat recovery offers the most attractive financial incentives. There are many uses for waste heat recovery across communities and industry, and it is clear the industry is actively embracing the topic, as best outlined by Vertiv’s guidance on redefining efficiency through heat reuse.

Data centres must reduce dependency on stressed grids and municipal water supplies. On-site renewable generation, paired with energy storage and microgrid capabilities, ensures resilience and cost control. Similarly, integrating advanced water treatment and water reuse systems, such as membrane bioreactors and closed-loop cooling, can cut freshwater withdrawal by up to 70%, while safeguarding operations in water-scarce regions.

Circular design principles should extend beyond building materials to include IT hardware and cooling infrastructure. Lifecycle tracking of pumps, heat exchangers and server components enables refurbishment and recycling, reducing embodied carbon and mitigating supply chain risks. Hyperscalers can lead by creating closed-loop ecosystems for rare earth metals and critical components.

Cooling alone represents approximately 20% of capital and 50% of operating costs for data centres. Air-only cooling consumes more electricity; water-based systems consume less energy but more water. Direct-to-chip liquid cooling can reduce energy use by up to 25-30% compared to conventional methods. Distributed pumping design can further add energy and cost efficiencies. Each site must pursue a portfolio of cooling technologies to meet their unique needs, and more innovation is to come as cooling keeps up with the exponential growth in rack densities.

Beyond serving consumer AI workloads, data centres should also be seen as a powerful tool for solving real-world infrastructure challenges. AI-driven digital twins can optimize grid performance, move and manage the grids in smarter and more efficient ways, detect leaks, and prevent flooding, reducing water losses by 25%, according to industry reports, and improving energy efficiency across municipal systems. This is not philanthropy; it’s a strategic investment that strengthens community resilience and lowers long-term risk.

Worldwide, public and private sectors are collaborating on solutions to reduce data centre resource use, and the results are promising. Two examples from the Nordic countries illustrate this progress:

Finland is pioneering heat recovery at scale. Microsoft's new data centres in the Helsinki region will supply up to 40% of district heating needs in Espoo, Kauniainen and Kirkkonummi – enough to heat tens of thousands of homes while cutting emissions.

Denmark offers another compelling example. Meta's data centre campus in Odense already channels 100,000 MWh of recovered heat annually into municipal heating networks, warming approximately 11,000 homes. The facility operates on 100% renewable energy while contributing to local decarbonization goals.

Building resilient AI infrastructure that serves as a local community partner requires collaboration across sectors. When data centres and utilities collaborate in designing thermal networks, sharing grid planning data, and developing flexible rate structures, they unlock benefits for both parties. We need regulatory frameworks at the national level that incentivize and streamline permitting across heat reuse and water systems.

With billions of people worldwide facing water stress, responsible management of water resources in digital infrastructure is not only an environmental concern – it is essential for social equity and economic stability.

The future of AI infrastructure is being written now, and we stand at a decisive moment in time. Much of the technology needed exists already. The business case is proven. What we need now is increased collaboration and coordination to leverage the undeniable potential of AI – while protecting the world’s most precious resources.

The World Economic Forum has long emphasized that breakthrough technologies succeed when they serve broad societal interests. As we convene leaders from technology, government and civil society, the joint collaboration on the integration of AI infrastructure within resource boundaries must be a central conversation. It builds on the Forum’s call for responsible innovation that drives impact, helps de-risk investments and transforms communities – a perspective that resonates with our founder's enduring belief that "the world is full of problems that can be solved in a better way".

Today, that better way means ensuring AI infrastructure evolves within planetary limits, shaping solutions that are both transformative and responsible.