‍What’s happening? Claims that AI can help tackle climate change have been criticised as greenwashing, with analysts arguing tech firms blur the distinction between traditional machine learning and energy-intensive generative AI. A review of 154 public statements found no evidence that popular tools such as chatbots or image generators have delivered material, verifiable emissions cuts. Much of the cited research lacked independent backing, with over a third of claims offering no evidence at all. Meanwhile, expanding data centres are driving rising electricity demand, particularly in the US. Critics warn that overstating AI’s climate benefits distracts from the sector’s growing energy use and associated emissions. (The Guardian)

‍

Why does this matter? There has been growing inspection into the taxonomies of AI, with the difference between traditional AI and generative AI not being emphasised in climate messaging. Traditional AI is machine learning that builds systems to learn from existing data sets, identifying patterns and making decisions with minimal human input to increase efficiency and accuracy of decision making. For example, through machine learning, complex climate and weather patterns can be tracked which are not obvious through human analysis alone, proving advantageous to understanding climate change.

‍

A different beast – On the other hand, generative AI does not just analyse data, it creates it. Generative AI platforms such as Gemini or Chat GPT are increasingly used in daily life to generate images, help create schedules and aid on work tasks to name a few, however energy consumption is much greater than traditional models. By 2030, generative AI is projected to consume 13x the energy of traditional AI. This creates a problem for climate change regarding scale and speed of adoption.

‍

The energy problem – The scale of energy needed is beyond current infrastructure capabilities. In Europe, electricity demand for data centres could rise 160% by 2030 – reaching 287 TWh – more than the total consumption of Spain in 2022. The expansion threatens to undermine the global emissions cuts, with newly added centres potentially emitting 44 million mt of CO2 into the atmosphere by 2030 in the US. Originally AI was presented as positive for the environment due to expansion through renewable energy sources, however, new green technology cannot keep up with sector expansion, with companies locking in contracts with fossil fuels to secure energy supplies.

‍

Water demand – Data centres are highly water intensive, as the hardware needs to be cooled. Currently, each 100-word AI prompt is predicted to use  one bottle of water (519 ml). By 2027, global AI demand is expected to account for 1.1-  1.7 trillion gallons of water withdrawal – more than four to six times Denmark’s total annual water consumption. While tech giants point to water recycling as a solution, the reality is complex.

‍

Water recycling – Minerals in water such as calcium and silica increase in concentration overtime as water evaporates. This means that water replacement is necessary to prevent equipment damage from high concentration minerals after a period of time. Furthermore, this water is also often drinking water from water companies. Although alternative sources are available such as water from oil and gas extraction and seawater, regulation and treatment costs prevent adoption. With at least 50% of the global population living in water-stressed conditions, industry consumption becomes a threat to local resource security.

‍

The social problem – A national review of 700 US data centres found that nearly half are located in areas with above-median environmental burdens. In Southaven, Mississippi, Elon Musk’s xAI is facing a lawsuit from the National Association for the Advancement of Coloured People (NAACP) for allegedly violating the Clean Air Act. The group claims xAI installed dozens of portable methane gas generators without permits in predominantly Black communities. These turbines emit nitrogen oxides and fine particulate matter linked to asthma and respiratory illness, effectively making these facilities the largest industrial pollution sources in the Southaven metropolitan area.

‍

Good governance – As use of generative AI continues to become easier and more accessible, causing increased expansion and money making, more regulations and governance controls must be put in place to ensure correct messaging. If companies are claiming green intentions, they must have the data and resolve to back it up.