Environmental groups balance AI benefits against its high energy and water costs

A tool for urgent conservation

Environmental and conservation organizations are increasingly turning to artificial intelligence to address the escalating climate crisis. These groups use the technology to analyze complex climate data, track forest fires in real-time, and count endangered wildlife populations. By leveraging machine learning, they aim to move faster and make better use of the information they collect from around the globe.

One specific example of this work involves monitoring remote areas in Cambodia. Conservation International recently began using AI to assess hotspots detected by satellites to send alerts to forest patrols about potential wildfires. This allows for a more rapid response in environmentally sensitive regions where manual monitoring would be nearly impossible.

The goal for many of these organizations is to make nature measurable and therefore investable. By providing more accurate data on biodiversity, these groups can secure the financing needed to protect natural resources at a larger scale. They view the technology as a necessary evolution in how they fulfill their duty to protect the planet.

Speeding up scientific research

AI is significantly reducing the time required for massive data processing tasks that were previously considered too expensive or difficult. For example, the Jane Goodall Institute used a tool developed with Amazon Web Services to digitize decades of handwritten field notes about chimpanzees. This project, which was expected to take up to 15 years to complete, was finished in about two years because the AI could scan, translate, and structure the notes in minutes.

Similarly, Kenya's Save the Elephants is using machine learning to improve population surveys and study how elephants communicate. An aerial survey of a large protected area generated roughly 200,000 images that would have taken nine people nine months to review by hand. AI models trained on Nvidia graphics platforms can now complete these types of analyses in just a few hours.

These automated systems have already revealed significant gaps in previous human-led research. The survey in Kenya found that human researchers were missing about a third of the elephant population, roughly half of the zebras and giraffes, and three-quarters of the oryx populations. This level of precision helps conservationists understand the true state of wildlife and where intervention is most needed.

The environmental cost of data

Despite the benefits, these same organizations are raising alarms about the environmental footprint of the data centers that power AI. These facilities require vast amounts of electricity and water for cooling systems. The International Energy Agency estimates that global data centers will produce the equivalent of more than 300 million metric tons of carbon dioxide annually by 2030.

The physical scale of this growth is substantial, with researchers estimating the global data-center footprint could more than double by 2030. This expansion is causing rising emissions for major tech giants like Microsoft, Amazon, and Alphabet's Google. These figures represent a significant challenge for companies that have publicly committed to long-term climate goals while expanding their AI infrastructure.

Water consumption is another major concern for the conservation community. Projections suggest that global data center water use will increase from about 560 billion liters in 2023 to roughly 1.2 trillion liters annually by 2030. This creates a dilemma for groups who are trying to protect water resources while using technologies that may deplete them.

Strategies for responsible use

To mitigate these impacts, some environmental groups are experimenting with specific technical strategies to limit their resource consumption. The World Resources Institute limits resource-heavy tasks like generating images or videos and caches query results so that identical searches do not require reprocessing. These steps help ensure that the AI tools they use remain as efficient as possible.

The Nature Conservancy has taken a different approach by moving computation to the 'edge.' Instead of uploading massive amounts of video from fishing vessels to a central data center, they installed AI-powered cameras that process information directly on the boats. This method reduces both the energy needed for computation and the requirements for data transfer across networks.

While some groups like Greenpeace International express a preference to sidestep AI entirely due to its footprint, others argue that the technology is a vital tool. They are currently establishing internal guidelines that emphasize transparency and human oversight. The overarching goal is to ensure that any AI deployment is intentional and serves a clear purpose in the fight against climate change.

Looking toward global policy

The tension between AI's utility and its environmental impact is expected to be a major topic at upcoming international climate events. At the UN's COP31 climate conference this month, host Turkey will call on various institutions to sign a pledge regarding sustainable AI development. This move aims to establish clear principles for how the technology should be deployed without harming the planet.

UN Secretary-General Antonio Guterres has called on major AI companies to be more transparent about the environmental impacts of their systems. He wants public disclosure of the energy and water costs associated with these large-scale models. This pressure on tech companies comes as they face increasing scrutiny over their ability to balance innovation with sustainability.

As the global data-center footprint continues to expand, the conversation will likely shift toward standardizing how we measure these costs. Organizations are now looking for a way to harness the speed of AI while demanding that the infrastructure powering it does not undermine the very goals they are trying to achieve. The coming months will be critical for determining how these trade-offs are managed on a global scale.