Is water the next shortage after electricity? Morgan Stanley: By 2030, data centers will consume 4.5 billion gallons of water per day!

The AI frenzy continues, as many Wall Street banks predict a surge in the number of generative artificial intelligence data centers in the United States by the end of this century. Various investment institutions are investing in the theme of "Powering up America," including grid companies, bulk commodities such as copper, gold, silver, and uranium, as well as artificial intelligence chip manufacturers. However, the issue of water consumption has been relatively overlooked. Morgan Stanley's Asia-Pacific stock research department recently published a research report titled "Deep Dive into Power, Cooling, Electric Grid and ESG implications," which elaborates on the impact of infrastructure on the development of artificial intelligence. The report specifically discusses the electricity consumption of artificial intelligence data centers. Additionally, Goldman Sachs previously found a significant increase in commercial electricity demand in Virginia amid the surge in "next-generation AI trading." Furthermore, even Blackstone Group CEO Steve Schwarzman and BlackRock Chairman and CEO Larry Fink have joined the investment theme of grid and AI, as there is a lot of upside potential in the coming years—unless AI demand collapses. It is worth noting that Morgan Stanley's report emphasizes at the end: "While data centers are easily focused on due to their high electricity consumption, the relatively high water consumption of data centers has been largely overlooked, which is also a fundamental characteristic of data center operations." Morgan Stanley cites data from Bluefield Research in the report, pointing out that the total water consumption of global data centers (including on-site cooling and off-site power generation) is growing by 6% annually from 2017 to 2022. The report states that by 2030, water consumption could soar to 450 million gallons per day. This is equivalent to approximately 681 Olympic-sized swimming pools of freshwater needed daily to cool global data centers. The report explains that the water usage of data centers is similar to carbon emissions and can be divided into Scope 1, Scope 2, and Scope 3. In this case, Scope 1 refers to water used for on-site server cooling, Scope 2 refers to off-site water used for power generation, and Scope 3 is considered water used in the server manufacturing supply chain. Specifically, Scope 2 water usage is indirectly related to data centers and power generation. Traditional power plants (such as coal-fired power plants, natural gas power plants, etc.) heat water to generate steam, which is used to rotate turbines for electricity generation. Water is also used for cooling in cooling towers. It is important to note that when using cogeneration, the water footprint of a data center's Scope 2 usage may be much larger than Scope 1 Research report indicates: "Considering the water intake required for operations in Zone 1 and Zone 2, by 2027, to meet the global AI demand, the water intake related to data centers may reach 4.2 billion to 6.6 billion cubic meters, exceeding half of the annual total water intake of the UK." The report further points out that in water-scarce areas, the significant water demand of data centers may intensify competition, potentially affecting the availability of water resources and even leading to the closure of data centers. "To dissipate the heat generated by servers, millions of gallons of water evaporate in the cooling systems every day, which accounts for a large portion of data center water usage."