Google’s annual Environmental Report, published June 30, 2026, reveals that the company’s electricity consumption rose 37 percent in 2025—its largest single-year increase on record. The surge, driven almost entirely by the expansion of AI data centres, pushed Google’s total electricity use up more than 250 percent since 2019 and drew roughly 42 million megawatt-hours over the year, a figure comparable to New Zealand’s entire annual power consumption.
What Happened
The 37 percent jump comes despite Google purchasing renewable energy certificates to match 100 percent of its electricity consumption for a ninth consecutive year. Operational carbon emissions actually fell 2 percent year-on-year, largely because of those renewable purchases and long-term clean energy contracts. However, supply-chain emissions—the carbon produced by manufacturing servers, constructing facilities, and sourcing materials—grew 25 percent, adding approximately 2.3 million tonnes of CO₂-equivalent to Google’s total footprint. Data centre construction was the single largest driver of that increase.
Google’s own report contains a striking admission: its “AI infrastructure buildout is currently accelerating faster than the grid is decarbonising.” The statement acknowledges a growing tension between the company’s long-standing climate commitments and the demands of its core business strategy. Renewable energy certificates can offset operational electricity use on paper, but they cannot reduce the carbon physically embedded in the steel, concrete, and silicon that make each new data centre possible.
Why It Matters
Google’s disclosure is the clearest signal yet that the AI boom is arriving with a substantial environmental price tag. As one of the world’s largest technology companies, Google’s energy profile shapes grid planning, water use policy, and energy investment decisions across every region where it operates. A 37 percent electricity increase in a single year from just one company is not a rounding error—it is a structural shift in global power demand.
The cooling challenge runs alongside the electricity challenge. AI inference and training generate intense heat requiring large volumes of water or energy-intensive cooling systems to manage. Canadian startup Wafr Technologies recently raised $100 million to address precisely this problem, with its technology claiming to cut AI data centre water consumption by as much as 95 percent. The scale of Google’s electricity figures makes clear why investors are pouring money into data centre infrastructure solutions at an accelerating pace.
Background and Context
Google is far from alone in this predicament. The entire tech industry is racing to expand AI infrastructure at a pace that grid operators and environmental regulators were not prepared for. Microsoft has committed to powering its data centres with nuclear energy before the end of the decade. Amazon Web Services is signing gigawatt-scale renewable contracts across three continents. The collective electricity demand from AI data centres is now large enough to influence national energy policy in the United States, Ireland, Singapore, and beyond.
Other hyperscalers are doubling down on expansion. Meta recently announced plans to rent out its spare AI compute to third parties, directly challenging AWS and Microsoft Azure—an expansion that requires exactly the kind of energy-intensive infrastructure that Google’s report puts numbers on. Meanwhile, Microsoft’s $2.5 billion Frontier Company initiative, designed to embed AI engineers inside large enterprises, will further drive demand for compute resources and the electricity needed to power them.
“Our AI infrastructure buildout is currently accelerating faster than the grid is decarbonising.”
Google Environmental Report, 2026
What Comes Next
Regulators and investors are beginning to scrutinise AI energy consumption more closely. The International Energy Agency reported earlier this year that data centre electricity use surged globally in 2025, and there are growing calls in both the European Union and the United States for mandatory energy reporting from hyperscalers. Google’s proactive disclosure in its sustainability report may be a strategic move to shape that regulatory conversation rather than react to it.
For consumers and enterprises, the environmental footprint of AI products is likely to become a more prominent factor in purchasing decisions. As AI becomes embedded in everything from search to drug discovery to enterprise software, the energy cost of each query and inference run compounds at scale. Google’s 37 percent figure is a baseline; without significant efficiency improvements or a step change in clean energy deployment, a similar or larger increase seems probable when the 2026 figures are published next year.
