Environmental Report by Google 2024: Successes and Ambiguities in AI Energy Consumption

In 2024, Google published an environmental report spanning over 80 pages. The document details the company’s efforts to address environmental issues and reduce its own impact on the environment. However, the report avoids a specific discussion on the energy consumption related to AI, possibly because the figures are much higher than the company would like to reveal.

The report contains a wealth of interesting data, showcasing the diversity of projects undertaken by a company as large as Google. For instance, Google implements a water replenishment program aimed at compensating for the water used at its facilities and achieving a positive water balance. This project involves identifying and funding watershed restoration and irrigation management. Thanks to these efforts, Google has managed to offset 18% of its water usage, with this figure increasing each year.

The company also emphasizes the potential benefits of AI in combating climate change, such as optimizing irrigation systems, creating more efficient routes for transportation, and forecasting floods. These initiatives can bring real benefits in various fields, and it’s important to acknowledge that Google is making significant strides here, even though many large companies do not.

However, the section "Responsible Management of AI Resource Consumption" clearly shows that Google does not provide specific information on the energy consumption of its AI systems. The company claims that data centers consume only 1.3% of the world’s energy, and Google's share is no more than 10% of that amount, which means 0.1% of global energy consumption. Nevertheless, this does not diminish the significance of the AI energy consumption issue.

In 2021, Google set a goal to achieve net-zero emissions by 2030, although the company acknowledges many uncertainties in how this will be accomplished. Since 2020, Google’s emissions have continued to rise. In 2023, the total greenhouse gas emissions were 14.3 million tons of CO2, 13% more than the previous year and 48% more than the baseline year of 2019. This is due to increased energy consumption by data centers and supply chain emissions. With further integration of AI into Google’s products, reducing emissions becomes increasingly challenging due to the growing energy demands of AI.

The company explains that predicting the environmental impact of AI is a complex and evolving process, and historical trends do not fully reflect the future trajectory of AI. Differentiating AI energy consumption from other workloads becomes less relevant, so Google prefers to provide overall data center metrics.

This raises doubts about whether Google truly does not know the exact contribution of AI to overall electricity costs. The company claims its custom AI server blocks and other technologies reduce energy consumption, yet full data on AI energy consumption remains undisclosed. It is important to continue requesting more specific information from Google and to monitor how the situation evolves in the future.