An interactive data story that breaks down how AI actually uses water, separating headlines from measurable reality through clear definitions
AI is often accused of quietly draining water resources. The claim is alarming, but the underlying numbers are rarely unpacked. This story examines what AI water use actually means, relying on publicly available data and clearly defined terms.
A key distinction is that AI-related water consumption occurs in two fundamentally different ways: onsite and offsite.
Onsite water use refers to the water consumed directly inside data centers. This is primarily used for cooling servers and maintaining safe operating temperatures for hardware. In reporting terms, this is typically labeled as data center water consumption, the actual water used at the facility itself.
Offsite water use refers to water consumed outside the data center, most commonly at nearby power plants that generate the electricity required to run the infrastructure. Electricity production can require substantial water for cooling. Offsite water use therefore includes the broader energy overhead of operating AI systems: data center overhead, CPU and RAM maintenance, idle machines, full system dynamic power draw, and active TPU and GPU consumption. While this water is not used inside the data center walls, it is indirectly attributable to its electricity demand.
This analysis follows definitions and estimates published by Google and independent researchers, separating these two categories to avoid conflating fundamentally different measurements.
Sources include:
By moving step by step, from definitions to measurement, and then from measurement to comparison, the picture becomes clearer. A single AI prompt uses very little water on its own. The impact becomes meaningful only at scale, and even then, context and comparison are essential.
The goal is not to dismiss environmental impact. AI infrastructure consumes both energy and water, and those impacts deserve scrutiny. The goal instead is to replace vague claims with measurable quantities, separating onsite facility consumption from offsite electricity-related water use so the discussion rests on transparent assumptions rather than aggregated headlines.
Questions like this cannot be resolved with a single static chart.
Vizzu allows the analysis to unfold visually. Definitions turn into numbers, numbers into comparisons, and assumptions remain visible at every stage. Viewers can follow how conclusions are constructed rather than being presented with a final figure stripped of context.
Because the visuals evolve smoothly, it becomes easier to see where data originates, how categories are defined, and why results look the way they do. That clarity is critical when discussing complex infrastructure systems that are easy to oversimplify.
Vizzu helps transform careful analysis into a transparent narrative—one that invites reasoning instead of reaction.
