AI Uses Water—but Where It Uses It May Matter More Than How Much

The Internet Isn’t Invisible: How Data Centers Are Changing the Environmental Cost of Our Digital Lives

Data centers give our digital lives a physical environmental footprint. Every AI request, streamed video, and cloud service ultimately depends on facilities that use real electricity, water, cooling, and land, even though that resource use is largely hidden from the people using them.  The Hidden Cost Behind Everyday Digital LifeStreaming a movie, saving family photos to the cloud, or asking an AI tool a question can feel almost effortless. A few taps on a screen, and the answer, image, or video appears.What is much harder to see is the physical infrastructure working behind that screen: buildings filled with servers, electrical equipment running around the clock, and cooling systems carrying away the heat those machines produce.That hidden infrastructure is growing quickly. Artificial intelligence is adding to demand that was already being driven by streaming, cloud computing, online shopping, banking, social media, and countless other digital services.The result is an environmental question that reaches far beyond the technology industry. The internet may feel invisible, but powering it requires real electricity, water, land, and infrastructure.For California residents, that raises a more useful question than whether data centers are simply good or bad for the environment. What matters is how they’re powered, how they’re cooled, and whether the places where they’re built have the resources to support them.The Cloud Has a Surprisingly Large Physical FootprintA large data center is typically a building, or sometimes a campus of buildings, containing the computing equipment that keeps digital services working.Rows of servers process and store information while networking equipment connects those machines to the rest of the world. All of that equipment needs electricity. It also produces heat, which means cooling becomes another major part of the operation. Add power distribution, backup systems, lighting, security, and other equipment, and the environmental footprint extends well beyond the servers themselves.Researchers at Lawrence Berkeley National Laboratory estimated that U.S. data centers consumed about 176 terawatt-hours of electricity in 2023, roughly 4.4% of total U.S. electricity use.Their newer projections show how quickly that could change. In an updated projection published in 2026, Berkeley Lab's reference case puts U.S. data-center electricity use at about 649 terawatt-hours by 2030, nearly four times the 2023 level, although considerable uncertainty remains around how fast demand will grow. Arman Shehabi, PhD, a staff scientist at Berkeley Lab who studies data-center energy and water use, has documented how AI is changing the industry's electricity needs.His research shows that AI is not the only force behind rising data-center demand, but increasingly power-intensive AI servers are helping accelerate that growth. That puts an entirely different backdrop behind something as ordinary as asking an AI assistant to plan dinner or summarize a document. The individual action may seem tiny. At scale, everyday digital activity depends on an expanding network of physical facilities.  AI Is Getting More Efficient. So Why Is Energy Use Still Rising?There’s an encouraging side to this story: computing technology keeps becoming more efficient.Newer hardware can accomplish more work with a given amount of energy, and the electricity required for some individual AI tasks is falling. Under ordinary circumstances, that might sound like a path toward lower overall energy use.The problem is scale.The International Energy Agency reported that global data-center electricity consumption increased 17% in 2025, even as the energy needed for individual AI tasks continued to improve.More people are using AI, companies are finding new applications for it, and some of those applications require increasingly intensive computing. It’s a little like making cars dramatically more fuel-efficient while simultaneously putting far more cars on the road and driving them much farther. Each trip may require less fuel, yet total consumption can still rise.This distinction matters when technology companies announce improvements in efficiency. Those gains are meaningful. But "more efficient" doesn’t necessarily mean "using less electricity overall."For an eco-conscious household, rising data-center demand doesn’t mean every streamed show or AI request needs to become a source of guilt.The more useful takeaway is that efficiency claims need context: a digital service can become more efficient per task while its total electricity use continues to grow as demand expands. In California, the Digital Boom Is Becoming an Infrastructure QuestionCalifornia is already planning for this growth.The California Energy Commission reported approximately 1,000 megawatts of existing data-center peak demand in the state as of December 2025. In its 2025 planning forecast, the commission projects statewide maximum data-center demand of about 5,800 megawatts by 2040, although future growth remains uncertain.In everyday terms, California is preparing for the possibility that data centers could become several times larger as a source of electricity demand.Utilities have received requests for even more potential capacity, but some projects are much further along than others, and not every proposed facility will be built.Even with that uncertainty, data centers have become significant enough for California to account for them specifically in its electricity forecasting.That matters because California's grid is being asked to handle several major changes at once. More vehicles are running on electricity.Buildings are gradually moving toward electric heating and appliances. Renewable power is expanding. Transmission and other grid infrastructure must keep pace with those changes. Large data centers add another substantial source of demand.Severin Borenstein, PhD, an energy economist at UC Berkeley's Energy Institute at Haas, has examined what happens when hyperscale facilities become unusually large utility customers.One concern is how infrastructure is planned and paid for when a single proposed customer can dramatically alter future electricity demand.That makes data centers more than private buildings. Once a project becomes large enough to require new substations, transmission capacity, additional power supply, or other grid improvements, questions about infrastructure and costs naturally follow. For Sacramento-area residents, that’s where an abstract technology story begins to feel much closer to home. Electricity doesn’t simply arrive at a data center. The grid has to be capable of delivering it. The Water Question Is More Complicated Than It SoundsElectricity is only half of one of the biggest environmental tradeoffs surrounding data centers.How that heat is removed can shift the environmental impact in different directions. Some cooling approaches rely more heavily on electricity, while others can place greater demands on water, creating a tradeoff that becomes especially important in places where either resource is under pressure.Berkeley Lab estimated that U.S. data centers directly consumed about 66 billion liters of water in 2023.But researchers also estimated a much larger indirect water footprint connected to producing the electricity those facilities consumed.That distinction is easy to miss.A data center may consume water directly through its cooling system, particularly when evaporation is used to remove heat.It can also be connected to water consumption elsewhere when the power plants supplying its electricity require water. Looking only at onsite water use can therefore leave out part of the picture. Cooling technology adds another wrinkle. Systems designed to minimize onsite water consumption may require more electricity. Cooling systems that rely on evaporation can sometimes improve energy efficiency,but also increase direct water consumption. Research involving Berkeley Lab scientists found that water consumption for computing workloads can vary enormously depending on hardware, server use, cooling technology, climate, and the electrical grid supplying the facility.That variation makes water use difficult to judge from a single number without knowing the conditions behind it. Peak demand adds another concern, particularly in California.A facility's annual water total doesn’t necessarily reveal what happens during a string of extremely hot summer days. Cooling needs can climb during the same hot periods when municipal water systems and electrical grids may already be facing heavier demand.Imagine two similar data centers. One operates in a place with plentiful water, moderate temperatures, and abundant low-carbon electricity.The other sits in a hot, water-stressed region where the electrical grid is already strained. The same technology could create very different environmental consequences.For California, that distinction is critical. The better question is not simply, "How much water does this data center use?" It is also, "Where is that water being used, when is demand highest, and what else is competing for it?" Three Questions Reveal More Than a “Green” LabelData-center sustainability can become complicated quickly, but three questions provide a useful way to cut through much of that complexity: energy, water, and location.Start with energy. How much electricity will a facility require, and where will it come from? Renewable and other low-carbon power can substantially improve the emissions picture.Some technology companies are also exploring nuclear and geothermal power, while investing in energy storage and other ways to secure reliable electricity. Just as important is when that electricity is needed. Facilities that can shift some computing work away from periods of heavy grid demand may be able to reduce pressure during the hours when electricity systems are most strained. But that flexibility has limits.Some computing can be delayed or shifted, while services that depend on immediate responses have far less room to move. How useful this approach becomes depends on the kind of computing a facility handles. Next comes water. The key question is whether a facility's water demands fit the supply, seasonal pressures, and competing needs of the community around it.A number that looks manageable over an entire year may carry different meaning when demand is concentrated during the hottest part of summer.Finally, there is location.Energy and water demands do not exist in isolation. Local climate, available resources, electrical-grid capacity, and surrounding development determine how easily a community can support a facility and what tradeoffs may come with it.These questions also help put environmental claims in perspective. A company saying a data center is powered by renewable electricity is providing useful information, but it’s not providing the whole picture. The same is true of claims about efficient servers or reduced water consumption.Water use is one example of why that broader context matters. Shaolei Ren, PhD, an associate professor of electrical and computer engineering at UC Riverside who studies the environmental effects of computing, has found that resource pressure can change with climate, cooling needs, and periods of extreme heat rather than being captured by one annual number.That flexibility has limits. Some computing work can be delayed or shifted, but services that depend on immediate responses may have far less room to move.The potential benefit therefore depends on what kind of computing a data center handles and how much of that work can realistically be scheduled around periods of heavy grid demand.That kind of flexibility is one more tool rather than a complete solution. Sustainability still depends on designing facilities around the resources actually available where they operate. Eco-Living Now Extends Beyond the Walls of HomeEco-friendly living has traditionally focused on choices close to home: household electricity, transportation, food, recycling, water conservation, and the products people buy.Those choices still matter. But digital infrastructure adds another layer, because many of its most important environmental decisions happen far beyond the household.A Sacramento resident can’t choose the cooling system inside a hyperscale data center or decide how a utility charges a company for connecting to the grid.But residents and communities can expect enough information to understand how a major project may affect resources they share.That makes transparency increasingly important. As data centers grow larger, communities need a clearer picture of their expected electricity and water demands, infrastructure needs, and potential effects on the surrounding area.Without that information, even well-intended sustainability claims can be difficult for the public to put into context.That doesn’t mean every data center should become a source of opposition. It means large digital facilities deserve thoughtful planning, much like other major developments that depend heavily on public infrastructure.The internet now supports work, communication, healthcare, commerce, entertainment, and countless ordinary routines. Its usefulness is not really in question. The challenge is making sure the physical systems behind that convenience grow in ways communities can reasonably support.For eco-conscious readers, that may be the larger shift in perspective. Sustainable living is no longer only about what happens inside the home.It also means understanding the infrastructure behind modern life and whether its demands can be supported responsibly by the communities and resources it depends on. Editorial TransparencyThis article was created to help Sacramento Living Well readers understand the less-visible environmental infrastructure behind everyday digital life.It approaches data centers as an Eco-Living issue involving energy, water, development, and responsible resource use rather than treating technology itself as either inherently harmful or sustainable.The goal is to give readers useful context for understanding a rapidly changing issue without placing responsibility for large infrastructure decisions on individual consumers.How This Article Was ResearchedThis feature was informed primarily by research and projections from Lawrence Berkeley National Laboratory, the International Energy Agency, the California Energy Commission, and California's electricity-grid planning system.Additional context came from research and expert work by Arman Shehabi of Berkeley Lab, Shaolei Ren of UC Riverside, and Severin Borenstein of UC Berkeley's Energy Institute at Haas.Research was compared across electricity demand, water consumption, cooling technology, grid planning, and California-specific infrastructure considerations so that environmental tradeoffs could be presented with appropriate context rather than as isolated statistics.If protecting the planet is part of how you live well, explore Eco Living — and discover more stories about mindful living on Sacramento Living Well.---Created by the Sacramento Living Well Editorial Team — part of DSA Digital Media, highlighting responsible choices and everyday sustainability.