In water-starved Middle East, large ChatGPT data centers are a cause of grave concern
A recent report has highlighted the undisclosed water footprint of large artificial intelligence (AI) models like ChatGPT on the environment. The Middle East and North Africa (MENA) region, which is the world's most water-scarce, faces similar risks due to the massive amount of water needed to cool down computational processes of advanced AI-powered language learning models. The report, "Making AI Less ‘Thirsty’: Uncovering and Addressing the Secret Water Footprint of AI Models", estimates that OpenAI’s ChatGPT needs to drink a 500-milliliter bottle of water for every 20 to 50 questions and answers exchanged. Similarly, GPT-3 and GPT-4 consume even more water. If used on a large scale and corporate level, this can cause severe damage.
MENA region is expected to see economic losses from climate-related water scarcity equivalent to 6-14% of gross domestic product (GDP) by 2050. The countries in the region have been placed in nine out of the 10 positions for the most water-stressed countries globally. Countries such as Qatar, Iran, Saudi Arabia, and others rely on large-scale air conditioning systems and desalination plants for freshwater, which require enormous amounts of energy and, therefore, water to operate, thus exacerbating an already bad situation.
High water consumption of ChatGPT and other models
Edurne Gil de San Vicente, program director at the Dubai-based non-governmental organization Water Alliance, said most companies in the MENA region do not disclose water usage data publicly as there is no legal obligation to do so. This makes it difficult to quantify the impact. ChatGPT needs a considerable amount of water to cool its computational processes. In addition, training GPT-3 in Microsoft's US data center can consume 700,000 liters (185,000 gallons) of water, which is the same amount of water needed to cool machines producing 370 BMW cars or 320 Tesla electric vehicles, as stated in the report.
Corporate level impact on the environment
The MENA region is the most water-scarce region globally, with hot climates and low water resources. 82% of wastewater is not recycled in the region, while Oman recycles 78% of its wastewater. The large-scale use of AI models like ChatGPT can exacerbate an already dire situation. Companies like Dubai Electric and Water Authority (DEWA), which plans to use ChatGPT, need to consider the high water consumption of these models.
The way forward
Government requirements for disclosing environmental, social, and corporate governance (ESG) data in MENA are slowly being adopted. The UAE and Saudi Arabia are leading in this respect, but there is a long way to go to incorporate universal and systemic sustainable practices in the region. The UAE has set the ambitious goal of reaching net-zero emissions by 2050; meanwhile, the country is also hosting the United Nations Conference of the Parties (COP28) in Abu Dhabi this year. Egypt hosted the climate action event COP27 the year before in Sharm El-Sheikh. Actions like these are helping to build awareness, encouraging research and development, and increasing investment in better corporate sustainable practices.
The Middle East and Africa green data center market accounts for around 10% of the global market. This takes place as major cloud service providers are expected to expand their presence in the region. For example, Microsoft announced its intent to invest in a new cloud data center in Saudi Arabia at the February LEAP technology conference in Riyadh. The kingdom is also exploring electricity-free cooling using salt and sunlight in a process that is currently under development at the King Abdullah University of Science and Technology.
