Artificial intelligence’s environmental impact is emerging as a growing area of focus for insurers, as increased use of AI‑driven services lifts energy and water consumption in data centres worldwide.
Insurance group Everywhen is among those warning that unchecked growth in AI workloads could put pressure on natural resources.
"We need to be asking ourselves an important question," the company's spokesperson said. "Can we keep using AI the way we currently are without harming the planet? Many people aren’t aware of the environmental impact that comes with a single AI query but, the truth is, if we continue to use AI at the current rate, it will put a significant strain on our natural resources.”
Generative AI tools such as ChatGPT now process billions of prompts a day worldwide. Each request draws on large language models running in energy-intensive data centres, meaning even a single query has a measurable carbon footprint. Studies suggest an AI chatbot request can use several times more electricity than a traditional web search, particularly where queries are complex or require multi-step reasoning.
This surge in AI-related computing is changing the risk profile of the technology sector. Large, high-value data centre campuses are becoming more critical to AI workloads, often clustered around a limited number of power grids and fibre routes. Property and business interruption underwriters are being asked to assess not only traditional perils such as fire, equipment breakdown and natural catastrophe, but also how AI-driven utilisation affects load profiles, cooling requirements and resilience.
Growing scrutiny of the carbon intensity of insureds’ power mix is also feeding into wider environmental, social and governance (ESG) discussions, as carriers weigh their own net-zero commitments against the expansion of AI capacity in their portfolios.
Alongside electricity demand, AI’s water footprint is emerging as a material concern. Modern data centres depend on intensive cooling systems to prevent hardware from overheating; many of these designs rely on significant volumes of freshwater, both directly on site and indirectly through power generation.
Academic research indicates that training and operating large language models can consume millions of litres of water a year at a single facility, depending on cooling technology and local climate conditions. A substantial share of global data centre capacity has been built in water-stressed regions, raising the risk of conflict with municipal, agricultural and industrial users during droughts or heatwaves.
This creates additional dimensions of risk for insurers. Physical exposures include potential constraints on operations or forced shutdowns where water use is curtailed. Regulatory and social pressures may follow if operators are perceived to be worsening local shortages.
At board level, there is a growing risk of directors’ and officers’ claims if disclosures around water use, cooling strategies or “green AI” marketing are challenged.
AI is not considered sustainable in its current form, but governments and regulators are moving to tighten oversight.
In the European Union, the AI Act and parallel measures on data centre reporting and energy efficiency are expected to increase transparency around power use, cooling methods and emissions.
In the US and other major markets, securities and prudential regulators are moving towards more detailed climate and technology risk disclosures for large corporates and financial institutions, capturing cloud and AI infrastructure.
These developments influence liability and regulatory investigation exposure for clients, particularly in directors’ and officers’ and professional indemnity lines. At the same time, more standardised data on energy and water use can be fed into underwriting, risk engineering and the design of ESG-linked covers.
Most organisations today use AI in at least one business function, with global surveys showing a clear majority of large companies deploying AI tools for tasks ranging from customer service to supply chain optimisation. Adoption is also climbing among small and mid-sized enterprises as models become easier to access via cloud platforms.
As AI becomes embedded in operations across sectors, its environmental impacts are no longer confined to a handful of technology giants. Manufacturing, financial services, logistics, healthcare and retail clients are all contributing to higher demand for compute.
As AI becomes an integral part of day-to-day life, many countries are exploring the idea of ‘sovereign AI,’ or AI that is designed, built, deployed and governed by a specific country or organisation. This allows them to set their own laws and regulations, while retaining greater control over security and data handling. It also means that they can decide how to source their energy and make greener choices.
In practice, sovereign AI strategies can encourage the development of national or sector-specific cloud regions, often backed by government support and subject to local environmental standards. However, they may also concentrate large amounts of compute capacity, and therefore energy and water demand, in particular jurisdictions or metropolitan areas.
AI is now deeply embedded in the modern world and has become part of the infrastructure shaping how organisations operate, how economies grow and how people interact with technology.
Across the market, carriers are beginning to update proposal forms and risk surveys for data centres, cloud providers and AI-intensive corporates to capture more detail on energy mix, cooling technology, water sources and local climate exposures. Climate and AI regulation is being monitored more closely, as new disclosure requirements could expose boards to claims if AI-related environmental impacts are under-reported or mismanaged.
Some insurers are also looking at how to support risk-reducing investments – such as waste-heat reuse, advanced cooling or relocation to less water-stressed regions – that can strengthen the insurability of AI infrastructure.
Everywhen’s spokesperson concludes: “Looking ahead, we can only hope that sustainability will be a priority, with more conscious choices about how and when AI is being used. Even something as simple as turning off our Google AI Overviews when carrying out traditional searches can help reduce unnecessary demand and help protect our planet.”
