By now, a lot of people have seen enough of AI to understand both its potential advantages and disadvantages for humans. What may not be widely known is how much of a natural resource guzzler it is.
Big tech corporations are using an increasing amount of water to cool their energy-intensive data centres, as the Financial Times reported this week. The server farms that are required to run generative AI have an even greater need for water and power.
The importance of AI’s energy dilemma
The introduction of AI hasn’t made sustainability concerns go away. The CEO of OpenAI, Sam Altman, has already issued a warning regarding the looming energy crisis that may impede the advancement of AI. He made this statement during the World Economic Forum conference in Davos. It may be interpreted as a continuation of his practice of forewarning people about the unsettling realities of the technology he represents. By doing this, he takes the lead in shaping the story and presents OpenAI—as well as himself—as a visionary leader.
However, his remarks regarding the necessity of a “breakthrough” in energy technology to power AI highlight another major problem with the technology: it is extremely expensive and will only become more so due to supply constraints, with the few companies producing chips capable of supporting AI reporting skyrocketing profits.
Not to mention the energy expenses, which were “eye-watering” when ChatGPT was launched in the fall of 2022. In summary, AI will significantly progress sectors like marketing, but at a cost.
The “elephant in the room”—a new academic study that has been quoted in the scientific magazine Nature—indicates that AI is significantly increasing environmental costs, particularly concerning the quantity of water needed:
- Water use increased by 20% and 34%, respectively, while Google and Microsoft educated their LLMs, according to their environmental reports.
- An additional study indicates that by 2027, the amount of water used globally by AI systems may only account for half of what the UK uses.
Users, as well as AI development companies, suffer from expensive and environmentally harmful systems. For the benefit of all, businesses should strive for AI to be both powerful and ecologically friendly.
Tech’s environmental paradox: balancing innovation with energy consumption
The image of Internet businesses as sleek, modern alternatives to traditional smokestack sectors is ironic, especially in light of Microsoft’s and other corporations’ commitments to being good climate citizens.
According to the International Energy Agency, data centres, cryptocurrency, and AI accounted for roughly 2% of the world’s power demand in 2022; by 2026, this percentage might increase to almost match Japan’s electricity usage. Ireland, a popular destination for server farms, is now restricting the number of new data centre connections to the electrical grid and other countries are considering following suit.
AI, though, makes a better case than cryptocurrency, which is already well-known for being an energy-hungry industry, that the power it uses may be justified by its potential for societal good.
However, tech businesses should expedite their use of AI to support the green transition in exchange for access to the finite water resources, energy, and grids that they require. It has been claimed that the system enables more accurate emissions tracking and weather forecasts, which helps balance power demand.
Finding creative solutions to reduce energy use in manufacturing, supply networks, office buildings, and farming could be beneficial.
At one of its data centres, Google cut down on cooling energy use by 40% by using machine learning from DeepMind, its AI division, to estimate weather and demand. Additionally, data centre activities are being sited by employing software to find regions of the world with surplus solar and wind power.
Tech giants investing in renewable energy for climate neutrality
The internet giants – Microsoft, Google, and Amazon – who are responsible for many of the biggest data centres in the world, have already made investments in solar and wind energy as well as water-saving data centre redesign.
However, they will need to take further steps to make sure they can harness power in a climate-neutral way, particularly as generative AI gains traction. This entails funding energy innovation as well. Microsoft has committed to be a future customer of Helion, and OpenAI’s Altman has invested in this US fusion power start-up as well
Researchers lament that it is difficult to find data, therefore the tech industry needs to be much more transparent about how much water and power it uses and how it is trying to reduce it. Businesses, particularly tech firms, are already under increasing pressure to reveal their environmental effects beyond carbon emissions. Water usage and other company impacts on the environment are being tracked via reporting systems being developed by the Taskforce on Nature-related Financial Disclosures.
These voluntary frameworks have drawn criticism for taking a while to show results. However, legislative initiatives are also growing. Though it might not succeed, a recent Democratic bill in the USA would establish guidelines for evaluating the precise environmental impact of artificial intelligence.
Data centre operators are now required by the EU’s Corporate Sustainability Reporting Directive to reveal their effect on the environment. It makes sense for tech companies to take the initiative rather than waiting to be compelled, as is the case with other regulatory sectors.
AI may help combat climate change, however, it must make sure that people view it as a component of the solution rather than a proponent of the issue.
(Tashia Bernardus)