Data centres are on track to double their electricity and water consumption by 2030 as AI workloads scale up, according to UN researchers. This buildout is putting utilities in the United States on a collision course with their clean energy commitments — and Nevada is already showing the cracks.

The scale of the problem is worth grasping in concrete numbers. Data centres used roughly 4.4% of all U.S. electricity in 2023, per the Department of Energy. The International Energy Agency projects that data centre electricity demand will grow at about 15% per year from 2024 to 2030. For comparison, overall U.S. electricity demand is expected to grow at only 2.1% annually, according to the U.S. Energy Information Administration. So data centres are expanding seven times faster than the grid as a whole.

Renewables are getting built. The U.S. is expected to increase renewable electricity from roughly 24% of the mix in 2025 to 27% by 2027, with nuclear staying steady. But here's the catch: that new clean power is being absorbed before it can actually reduce how much coal and natural gas we burn. The demand curve is steeper than the supply curve. In other words, new capacity is being consumed almost as soon as it's online, unable to displace fossil fuels or strengthen the grid.

The first place this breaks is at the state level. Nevada's largest utility has signalled it may fail to meet its 2030 clean energy mandate, and it points directly to data centre growth as the reason. Nevada is not alone — it is the canary. The state offers cheap land and a welcoming regulatory environment for data centre companies (known as "hyperscalers"), and its power grid was built for the old demand profile. What happens in Nevada will repeat in Arizona, Texas, and other Sun Belt and Mountain West states where the same math applies.

What the Numbers Actually Show

The 15% annual growth figure for data centres deserves a closer look. If you compound that rate over six years, you get the UN's headline: data centre electricity consumption roughly doubles. But the baseline matters for understanding what this means. If data centres already consume 4.4% of U.S. electricity and double that share by 2030, they would be pulling roughly 8–9% of a national grid that is itself expanding. This is not a temporary spike. It is a structural load — a permanent new category of demand — that changes how utilities have to plan for the next 10 to 20 years.

Here is how the timing problem works. When a hyperscaler wants to build a data centre, it signs a power purchase agreement with a utility. The utility then faces a choice: either build new generation to serve it (ideally renewable, to meet state clean energy targets), or pull the power from existing power plants that often run on fossil fuels. The problem is speed. A data centre can be built and running in 18 months to two years. A wind farm or solar facility takes three to five years to permit, build, and connect to the grid. That gap — the time between when the data centre turns on and when the clean power arrives — is when utilities resort to fossil fuels. Over thousands of megawatts and dozens of projects, that gap accumulates into missed climate targets.

Water is the second, quieter constraint. Most large data centres use evaporative cooling — essentially water-based air conditioning that trades electrical efficiency for water consumption. Double the power draw, and you roughly double the water draw. This lands hardest in Nevada, Arizona, and parts of Texas, all of which are simultaneously building out data centre capacity and struggling with water scarcity. No one is coordinating these two trajectories at a federal level.

The deeper issue is not temporary or cyclical. The U.S. energy transition was designed around an assumption: electricity demand would grow slowly or stay flat. But AI infrastructure is injecting a new, geographically concentrated, always-on load that tests both the speed of renewable energy build-out and the credibility of state clean energy laws. Utilities that made renewable energy promises when demand was lower now face a real choice between hosting new data centre load and keeping their climate commitments.

What happens next depends on several moving pieces. Federal permitting reform could speed up transmission and renewable construction. Technology companies could follow through on their public commitments to power data centres with 24/7 carbon-free energy at the portfolio level — or those commitments could soften quietly as the gap between demand and supply widening widens. The Nevada signal matters not because it is unusual, but because it is arriving early. A dozen other states are now in the same position. How this resolves in Nevada will shape what happens in all of them.