The increasing reliance on artificial intelligence (AI) and data management has raised questions about their environmental impact, an issue that is proving to be more significant than many initially believed. The environmental cost of AI is not solely linked to actions like querying a chatbot or streaming a video, but rather stems largely from the extensive infrastructure required to support these technologies—primarily, the data centres that operate continuously around the clock.
According to Professor Mark Gahegan from the University of Auckland, data centres serve as the backbone of the digital economy. “Data centres are a place where the computers are, and because you often need the computers to be next to the data, that’s also where the data lives,” he said. These facilities require substantial energy for both computing power and cooling systems to prevent overheating. Current estimates indicate that data centres account for approximately 0.5 to 1 percent of global energy-related emissions and are one of the fastest-growing sources of electricity demand.
AI exacerbates this issue, although its energy consumption varies greatly. While the initial training of AI models is energy-intensive, the operational costs of running these models are relatively low. For instance, OpenAI’s ChatGPT consumes about 0.34 watt-hours of energy per query, similar to the energy used by a high-efficiency light bulb in a few minutes. However, the training of large models like GPT-4 requires between 52 million and 62 million kilowatt-hours of electricity, highlighting where the substantial costs lie.
The demand for computing power has surged since the 1990s across various sectors such as medicine, genomics, and climate science. While improvements in cooling technologies and chip design may help, they are outweighed by the rapid growth in energy needs. The International Energy Agency projects that global data centre electricity consumption will reach around 415 terawatt-hours in 2024, accounting for 1.5 percent of global electricity use, with an annual increase of 12 percent. Research firms like Gartner predict even steeper growth, estimating that consumption will rise to 448 terawatt-hours by 2025.
This escalating demand places data centres in a unique position; unlike many industries, their emissions are expected to rise over the next decade. Gahegan points out that there is currently no viable alternative to large data centres. “I don’t think we’ll be moving away from them any time soon,” he stated, emphasizing the challenges inherent in a modern digital economy.
Interestingly, while AI poses a climate challenge, it also has the potential to mitigate other environmental issues if deployed thoughtfully. For example, Microsoft’s AI climate forecasting system, known as Aurora, can produce weather forecasts in just one minute after an energy-intensive training phase, compared to traditional models that take hours. Gahegan notes, “There’s a big upfront [energy] cost, but once the model exists, it’s much cheaper to run.” AI also shows promise in enhancing system efficiencies, detecting methane leaks, optimizing power plants, and improving transport efficiency, among other applications.
In New Zealand, the situation is further complicated by reliance on overseas data centres. Much of the cloud computing used in the country depends on infrastructure located in Australia and beyond, where electricity sources may be less sustainable. Gahegan highlights the ethical tension this creates: “Asking ChatGPT a question in New Zealand means relying on overseas data centres,” he said, noting that the environmental costs are often borne elsewhere.
Despite the challenges, individuals can still make a meaningful difference. Gahegan argues that while home users are not the primary drivers of the problem, they should focus on energy-efficient choices. “People tend to buy the most powerful computer they can,” he noted, suggesting users consider the energy draw of devices before purchasing. Practical steps include opting for minimum-spec devices, extending the lifespan of current equipment, and being mindful of data storage practices.
Businesses are also encouraged to ensure that their computing equipment is recycled responsibly. Gahegan points out that running heavy computing tasks on personal laptops may often be less efficient than using a well-managed data centre. He encourages better data habits, advocating for the archiving and deletion of unnecessary information to mitigate energy consumption.
The intersection of AI and environmental sustainability raises urgent questions about how society will navigate these challenges in the coming years. The ongoing evolution of technology necessitates a serious dialogue about its broader implications for our planet.
See also
Tesseract Launches Site Manager and PRISM Vision Badge for Job Site Clarity
Affordable Android Smartwatches That Offer Great Value and Features
Russia”s AIDOL Robot Stumbles During Debut in Moscow
AI Technology Revolutionizes Meat Processing at Cargill Slaughterhouse
Seagate Unveils Exos 4U100: 3.2PB AI-Ready Storage with Advanced HAMR Tech
