The local vs. global challenges of AI resource use

- You've gotten a sense for how powerful AI tools can be in the fight against climate change and other sustainability challenges. Let's talk now about what it takes to build an AI tool. Like any tool, AI requires resources to build and use, resources like energy and water. In this video, I'll explain how water is used in the building and running of AI models, and I will put this in perspective of global water challenges and the sustainability transition. And in the following video, I will dive into AI and energy. So let's turn to water. Did you know that nearly half of the world's population could be living in areas facing water scarcity as soon as 2030? That's what the United Nations has warned. Today, many places in the world already face severe water limitations, but these challenges are projected to get worse. Global water demand is expected to increase by 55% by 2050. Manufacturing is expected to see a 400% increase. Thermal power generation might need 140% more water, and domestic use could jump by 130%. The expansion of AI operations is also increasing water demand, but AI's direct consumption is small relative to global demand, and fortunately, AI tools can also help to tackle the growing water crisis. The building and operating of AI tools can lead to water use in two ways. First, water is often used for cooling data centers, which can get very hot. Have you ever noticed your laptop feeling warm after being on for a while? Imagine a building full of computer processors. They can generate a tremendous amount of heat. Cooling is essential to prevent servers from overheating, which can lead to equipment failures. The second way building and running AI models can lead to water use is indirectly, through the water consumed in electric power generation, especially if the power plants burn fossil fuels. Renewable sources, like wind and solar, use virtually no water. This means indirect water use is essentially an energy issue, which I will cover in the next video. Let's take a closer look at the role of water in keeping data centers cool. Many data centers rely on water-based cooling systems. However, there are a variety of innovative cooling solutions in use today that don't require large amounts of water. These include advanced air cooling, liquid immersion cooling, and phase-change cooling. With continued advancements and broader adoption of these technologies, data centers will become less dependent on water. Currently at a global or even national scale, the combined direct and indirect water consumption by data centers is relatively small. For example, data centers in the US are responsible for only about 0.15% of national water consumption. Roughly three quarters of this is related to electricity generation, and as you will learn in the next video, AI currently drives only a small fraction of all data center energy use. However, locally, without proactive measures, data centers can add stress to some local water districts, particularly if they are already water-stressed. In addition to shifting away from water-based cooling systems, to address these challenges, many companies are increasingly investing in what is known as water replenishment. Water replenishment refers to activities that reduce use, recharge aquifers, or improve water quality in local watersheds and communities. AI can help implement and scale water replenishment. Consider this, across the US, households waste 3 billion cubic meters of water each year because of leaky pipes. This is about 30 times as much water as is estimated to be consumed by AI operations globally in a year. What if AI tools could be used to find and stop water leaks? Well, that is actually happening today. For example, FIDO Tech is a company that delivers AI-enabled leak detection and water management solutions, which are already reducing leakage in municipal water systems around the world. As water challenges around the world become more severe, what is becoming increasingly clear is that the AI toolbox can provide powerful tools for overcoming them if there are incentives and investments made to direct them to address these challenges.