AI

Google electricity use jumped 37% in 2025 on AI data center surge

The tech giant's power consumption now rivals entire nations, though carbon emissions held flat through clean energy purchases.

Omega Editorial· July 2, 2026· 3 min read

Record power consumption driven by AI infrastructure

Google's electricity consumption surged 37 percent in 2025, marking the largest single-year increase in the company's history, according to details first reported by Ars Technica from Google's latest sustainability report. The jump reflects the accelerating pace of AI data center construction across the tech industry.

Google's data centers alone consumed more than 42 million megawatt-hours of electricity in 2025, up from 30.6 million megawatt-hours the previous year. That places the company's data center energy footprint on par with the total electricity consumption of countries including New Zealand, Denmark, and Nigeria.

The 2025 spike continues a sharp upward trajectory. Google's total electricity use has climbed more than 250 percent since 2019, driven by growth in Google Cloud, YouTube video streaming, and infrastructure supporting AI products and services. The company saw a 27 percent increase in 2024 alone.

Why it matters

The scale of Google's energy demand illustrates a fundamental tension facing the AI industry: computational requirements are growing faster than the electrical grid can decarbonize. As enterprises evaluate AI deployments, the infrastructure's energy footprint represents both a cost consideration and an environmental challenge that may influence regulatory frameworks and competitive positioning in the coming years.

Emissions held flat through clean energy strategy

Despite the steep rise in power consumption, Google reported a 2 percent reduction in operational emissions over the same period. The company has matched 100 percent of its electricity use with renewable energy purchases for nine consecutive years. In 2025, Google's purchase agreements for 12 gigawatts of clean energy represented its largest annual commitment to date.

Google acknowledged in its sustainability report that simply matching total consumption with renewable purchases doesn't guarantee clean power at every location and hour. The company has shifted focus toward what it calls "24/7 carbon-free energy," which requires more granular accounting of hourly and local clean energy matches.

However, supply chain emissions from contracted manufacturers and suppliers grew 25 percent, which Google attributed to an Asia-Pacific supply chain operating on grids with limited carbon-free energy access.

Natural gas investments complicate clean energy narrative

Google has invested more than $3.8 billion between 2010 and 2025 in projects expected to bring 7.5 gigawatts of clean energy online. The company's portfolio includes advanced nuclear, fusion energy, enhanced geothermal, long-duration energy storage, and natural gas with carbon capture.

Yet Michael Thomas, CEO of the Cleanview data platform tracking renewable energy projects, noted in an April 2026 analysis that Google's $40 billion Texas data center investment includes a campus potentially powered by a 933-megawatt natural gas plant without carbon capture technology. That facility's turbines could produce 4.5 million tons of carbon dioxide annually, according to Thomas's assessment. A Google spokesperson told Thomas the company has not finalized power agreements for that data center.

The sustainability report stated that achieving climate goals "will not be linear" given that AI infrastructure buildout is "currently accelerating faster than the grid is decarbonizing."

Details were first reported by Jeremy Hsu at Ars Technica.

#google#data centers#ai infrastructure#energy consumption#carbon emissions#renewable energy

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in AI

AI· 3 min read

Startup Deploys 100+ AI Cameras to Make San Francisco Streets Searchable

Orchestra wants to turn urban video feeds into structured data for police, insurers, and autonomous vehicle companies—without selling raw footage.

Via AI Watch · Jul 2, 2026
AI· 3 min read

Google's AI Tutor Delivered Year-Long Learning Gains in 8 Weeks

DeepMind rebuilt Gemini from the ground up to guide rather than answer, then tested it in Sierra Leone classrooms with measurable results.

Via AI Watch · Jul 2, 2026
AI· 3 min read

Publishers Build Parallel Websites for AI Agents, Not Humans

Time, The Economist and others are creating markdown versions and new protocols to stay visible as AI systems reshape web traffic.

Via AI Watch · Jul 1, 2026