According to www.techtimes.com, Google’s AI infrastructure drove a 37% year-over-year increase in total electricity consumption in 2025 — the largest single-year jump in the company’s history — while its supply chain carbon emissions rose 25%.
Record Energy Demand and Physical Grid Reality
Google’s 2026 Environmental Report, published on June 30, 2026, confirms that its data centers consumed approximately 42 million megawatt-hours of electricity across 2025 — a volume comparable to New Zealand’s entire annual electricity use. This surge contributed to cumulative electricity growth of more than 250% since 2019. The report coincided with Amazon’s parallel sustainability disclosure, which recorded a 16% emissions increase for 2025 — marking the first side-by-side public accounting of AI’s planetary cost.
Although Google matched 100% of its electricity use with renewable energy certificates (RECs) for the ninth consecutive year — and reported a 2% decline in operational Scope 1 and Scope 2 emissions — the physical electrons powering its servers remain indistinguishable from fossil-fueled grid supply. As peer-reviewed research in Nature Climate Change concluded, this accounting method is “unlikely to result in actual emission reductions,” and under current trends, 42% of corporate emissions reduction commitments will not translate into real-world mitigation.
Scope 3: Where the Carbon Debt Accumulates
Under the Greenhouse Gas Protocol, Scope 3 emissions — covering upstream value chain activities including chip manufacturing, server production, and logistics — now constitute roughly 80% of Google’s total carbon footprint. These emissions grew 25% year-over-year in 2025, with data center construction alone contributing 2.3 million metric tons of carbon dioxide equivalent. Google’s total Scope 3 emissions have roughly doubled since its 2019 baseline.
The company explicitly attributes this growth to hardware manufacturing on fossil-heavy grids, naming semiconductor suppliers in Taiwan, Japan, Vietnam, and India. A REC purchased in Iowa or Texas cannot decarbonize the electricity used at a foundry thousands of miles away — a geographic and accounting disconnect that defines the structural gap in AI climate accountability.
Efficiency Gains vs. Aggregate Demand: The Jevons Paradox
Google has emphasized per-query efficiency: a technical paper from August 2025 estimated that the median Gemini text prompt uses just 0.24 watt-hours — a 44-fold improvement over the prior 12 months. Yet this efficiency has not reduced total load; it has accelerated adoption. As Google’s own data shows, electricity consumption rose 37% in 2025 even as per-prompt energy use plummeted — illustrating the Jevons paradox, where falling computational cost drives higher aggregate demand.
Water use follows the same pattern: operational water consumption climbed 34% to 10.9 billion gallons in 2025. In The Dalles, Oregon, Google’s water use quintupled between 2012 and 2025, reaching approximately 550 million gallons — about 40% of the city’s annual municipal supply.
Clean Energy Procurement: Scale and Structural Limits
Google signed agreements for more than 12 gigawatts of net-new clean energy in 2025 alone — its largest annual procurement ever and exceeding its combined total from the two preceding years. Projects included restarting Iowa’s 600-megawatt Duane Arnold nuclear plant with NextEra Energy, a 3-gigawatt hydropower framework with Brookfield, and a fusion energy power purchase agreement with Commonwealth Fusion Systems.
Despite a 37% electricity demand increase, Scope 1 and Scope 2 emissions fell 2% — a genuine engineering achievement. Yet the report candidly states: “our climate moonshot is getting harder.” That phrase reflects not projection but observed trajectory: the gap between AI infrastructure growth and grid decarbonization continues to widen. Sasha Luccioni, co-founder and chief scientific officer of the Sustainable AI Group, underscored the urgency:
“We’re essentially in a climate crisis and we should not be having emissions growth at all, arguably, and yet the data centers are going in the opposite direction.” — Sasha Luccioni, co-founder and chief scientific officer of Sustainable AI Group
Source: techtimes.com
Compiled from international media by the SCI.AI editorial team.
