Google wants to organize the world’s information. As it turns out, that goal comes with a carbon footprint roughly the size of a small country.
According to the company’s 2025 Environmental Report, total greenhouse gas emissions in 2024 reached approximately 11.5 million tons of CO2 equivalent. This is a 51% increase from Google’s 2019 baseline, and is overwhelmingly due to the infrastructure required to run AI models like Gemini.
Data center power consumption alone increased 27% year over year. The company attributes this primarily to growth in AI workloads and overall supply chain operations. For context, the International Energy Agency predicts that global data center electricity usage could double to approximately 945 TWh by 2030.
efficiency paradox
In fact, the company achieved a 12% reduction in data center energy emissions in 2024 thanks to clean energy investments and more efficient hardware. But total emissions are still ballooning because the demand for AI computing is growing faster than the efficiency gains can offset.
Google’s numbers for individual AI tasks are really impressive when looked at individually. The median energy consumption of one Gemini text prompt is just 0.24 watt-hours, producing 0.03 grams of CO2 equivalent and using 0.26 milliliters of water. According to the company, this means up to a 44x improvement in carbon footprint per inference task over the past year.
Google’s data centers now deliver more than six times more computing power per unit of electricity than they did five years ago. On the hardware side, Google notes that the 7th generation Ironwood TPU is nearly 30 times more power efficient than the first Cloud TPU in 2018.
Clean energy balancing act
Google reports an average compliance rate of 66% for carbon-free energy, backed by more than 8 gigawatts of clean power capacity contracts.
The IEA predicts that U.S. data centers are on track to reach record peak electricity demand between 2026 and 2027.
What this means for investors
Rising energy procurement costs are worth monitoring closely, especially for Alphabet investors. Securing 8 GW of clean power contracts won’t be cheap. And competition for renewable energy capacity will only intensify as hyperscalers compete to build AI infrastructure simultaneously. These costs will ultimately be reflected in capital expenditures and could weigh on profits if AI monetization does not scale proportionately.
European regulators have already shown a willingness to impose environmental standards on technology companies, and U.S. policymakers are increasingly focused on the energy needs of data centers. Companies that can demonstrate real progress on sustainability, not just in task-specific efficiency metrics, but also actual reductions in total emissions, will be better positioned once the regulatory environment materializes.
