Why AI Data Centers Are Straining the Power Grid

Data centres powering artificial intelligence are on track to double how much electricity and water they use by 2030, according to UN researchers. This growth is colliding head-on with promises made by utilities and states to run on clean energy.

The scale of the problem is stark. In 2023, data centres used about 4.4% of all electricity consumed in the United States, according to the Department of Energy. The International Energy Agency projects that data centre electricity demand will grow at roughly 15% per year through 2030 — more than four times faster than overall U.S. electricity demand, which the government expects to grow at about 2.1% annually.

Here is the crux: new renewable energy coming online is being swallowed up by data centres before it can reduce carbon emissions. Renewables are expected to grow from roughly 24% of U.S. electricity generation in 2025 to 27% by 2027, according to energy forecasts. But the demand curve is steeper. Every megawatt of new solar and wind is being claimed to power AI systems, leaving the grid no cleaner overall.

We are already seeing the real-world consequences. Nevada's largest utility recently said it may not meet its state mandate to run on clean energy by 2030 — directly because of data centre growth, according to AP News. Nevada is not an outlier. It is a preview of what other states will face.

Why This Gap Between Power and Promises Matters

The 15% annual growth number deserves a closer look. If data centre electricity use compounds at that rate for six years, it roughly doubles — which aligns with the UN projection. But consider the starting point: if data centres already account for 4.4% of U.S. electricity in 2023 and double by 2030, they will consume roughly 8-9% of the national grid by then. That is a permanent structural change to electricity demand, not a temporary blip.

Utilities plan their power supplies a decade or two into the future. When a tech company signs a deal to buy power, the utility must build new generation to serve it — ideally wind farms or solar installations to meet clean energy goals. But there is a timing problem. A solar farm or wind turbine takes three to five years to permit, build, and connect to the grid. A data centre can be running in half that time. The data centre arrives before the clean power does. In that gap, utilities turn to existing coal or natural gas plants to fill the load, and state clean energy targets slip.

Water adds a second pressure. Large data centres need massive amounts of water for cooling — think of it like how a power plant needs water for steam. When data centre power use doubles, water consumption roughly doubles too. Nevada, Arizona, and parts of Texas are simultaneously building out data centres while already managing severe water shortages. These two trends are not being coordinated at the federal level, which amplifies the local stress.

The broader tension is structural, not temporary. The U.S. energy transition was designed assuming electricity demand would stay roughly flat or grow slowly. AI infrastructure is now injecting a new, always-on, geographically concentrated power demand that tests both how fast utilities can build renewable energy and whether state clean energy laws can actually be met. Utilities that promised to go clean during an era of flat demand now face a genuine choice: grow with data centres or stick to their clean energy pledges. They cannot easily do both.

What happens next rests on three things: whether the federal government speeds up permitting for new power lines and energy projects, whether tech companies actually build the 24/7 carbon-free energy systems they have promised, or whether those promises quietly weaken as the electricity gap widens. The Nevada signal is worth watching — not as an isolated problem, but as a warning sign for the dozen states now building out the most data centre capacity.