Google's AI Buildout Drove a 37 Percent Jump in Its Electricity Use Last Year

Google consumed 37 percent more electricity in 2025 than it did in 2024. That is not a number that appears buried in a footnote. It is a headline figure from the company's own environmental reporting, and it tells you something important about the real cost of the AI infrastructure boom that every major tech company is currently racing to build.

To put that in context, Google's total electricity consumption in 2025 reached approximately 34 terawatt-hours, up from around 24.7 terawatt-hours the year before. For reference, 34 terawatt-hours is roughly equivalent to the annual electricity consumption of Denmark, a country of nearly six million people. That is one company, in one year.

The primary driver is data centre expansion to support AI workloads. Training large language models and running inference at scale, meaning responding to millions of user queries through products like Gemini, requires vast amounts of compute. That compute runs on GPUs and custom AI accelerators, which draw significantly more power than the general-purpose server hardware they are replacing. Google's own Tensor Processing Units are more efficient than off-the-shelf alternatives, but efficiency gains are being outpaced by the sheer growth in demand.

The Renewable Energy Complication

Google has been committed for years to matching its electricity consumption with renewable energy purchases, and the company maintains that it continues to do so. But there is an important distinction between buying renewable energy certificates and actually running your data centres on clean power at every hour of the day.

The electricity grid does not work like a subscription service. When Google buys a renewable energy certificate, it is paying a wind farm or solar project to put clean electricity onto the grid somewhere. That does not mean the specific electrons powering a Google data centre in Virginia at midnight are from a solar panel. At night, or on days when wind generation is low, those facilities are drawing from whatever mix of generation the local grid offers, which often includes gas and coal.

Google has been pushing toward what it calls 24/7 carbon-free energy matching, where it tries to ensure clean power is available on the grid in the same location and at the same time as its consumption. But achieving that at scale, especially with consumption growing this fast, is an enormous challenge. The company's own 2025 environmental report acknowledged that it fell short of its 24/7 matching targets in several regions.

This matters for a few reasons. First, the communities living near these data centres bear the environmental cost of the additional grid demand regardless of what certificates Google purchases. Second, the surge in data centre construction is already influencing energy infrastructure planning across the United States and Europe, prompting utilities to keep older fossil fuel plants running longer than they otherwise would have.

What This Means for Tech's Climate Promises

Google is not alone in this position. Microsoft, Amazon, and Meta have all reported significant increases in energy consumption tied to AI infrastructure, and all face the same fundamental tension between ambitious climate commitments and the practical realities of exponential demand growth.

Microsoft's 2025 sustainability report showed its carbon emissions had actually increased since 2020, the year it pledged to become carbon negative by 2030. Amazon has similarly struggled to hit its renewable matching targets as AWS expands. The pattern is consistent: climate commitments made before the generative AI boom are straining under the weight of the infrastructure those products require.

None of this means AI development should stop. But it does mean the tech industry needs to be more honest about the trade-offs involved, and regulators and investors should be asking harder questions. A 37 percent annual increase in electricity use from a single company is not a rounding error. It is a material impact that deserves scrutiny proportional to its scale.

The irony is not lost that Google is simultaneously using AI to help optimise energy systems and predict renewable generation. The technology may eventually help solve the problem it is currently making worse. But right now, the numbers are moving in the wrong direction.