Google has reported a significant rise in its carbon emissions as it intensifies its focus on artificial intelligence (AI). According to the company’s 2025 sustainability report, its so-called “ambition-based emissions” surged by 11 percent over the past year, reaching 11.5 million metric tons of carbon dioxide, reflecting a 51 percent increase since 2019.
This increase moves Google further away from its target of reducing greenhouse gas emissions by 50 percent by 2030 based on 2019 levels. The “ambition-based” emissions metric notably excludes certain supply chain emissions that Google claims are beyond its control, including specific purchased goods and services and food programs.
A detailed breakdown found in the report’s appendix reveals that the total emissions actually amounted to 15,185,200 metric tons of carbon dioxide, an amount comparable to the annual emissions of nearly 40 gas-fired power plants. While the report does indicate a 12 percent decline in emissions from data centers, the overall energy consumption is still on the rise due to the demands of AI technologies. Even with this reduction, emissions from data centers remain significantly elevated compared to 2019 levels.
The report attributes several external factors beyond Google’s direct influence as obstacles to its climate ambitions. These include the rapid advancements in AI, which complicate future predictions regarding energy needs and emissions. Other cited challenges encompass shifts in climate and energy policies, delays in the rollout of carbon-free technologies, and a deficiency of renewable energy sources in certain regions. Google acknowledges that these external conditions could impact the cost, feasibility, and timeline of its environmental initiatives, emphasizing the need for flexibility.
The energy consumption associated with AI is not unique to Google; other tech giants such as Microsoft and Meta are also experiencing increased energy demands due to similar developments. Current projections indicate that AI workloads may surpass the energy consumption of Bitcoin by the close of this year. Despite ongoing efforts from some companies like DeepSeek to develop more energy-efficient AI models, the trend of constructing energy-intensive data centers persists, as exemplified by Meta’s plans for a large gas-powered facility in Louisiana. Additionally, significant political moves, such as an executive order signed by former President Donald Trump promoting coal use for powering AI data centers, further complicate the landscape.
