Google’s Sustainability Report for 2024

In our earliest years, we set out on an ambitious path to contribute to a more sustainable future. An essential aspect of this mission is to share the knowledge we have gained and to be open about our progress and the obstacles we face. This is especially important now, as technological progress is intersecting with the imperative for an energy transition. Our annual Environmental Report delves deeply into our endeavors to leverage technology, specifically AI, to instigate positive environmental transformations and conduct our operations sustainably.

Our strategy for harnessing AI for sustainability involves scaling up AI and using it to hasten climate action, which is just as critical as dealing with the environmental consequences tied to it. To reduce our environmental impact, we have established cutting-edge and efficient infrastructure for the AI era. This includes Trillium, our latest Tensor Processing Unit (TPU) generation, which boasts over 67% higher energy efficiency than TPU v5e. Through tried and tested methodologies, we have found ways to significantly decrease the energy needed to train an AI model and cut related emissions. These practices are presently employed across Google.

We are committed to fabricating the most energy-efficient computing infrastructure globally, bolstered by responsible water management practices and a dedication to waste reduction. On average, a Google-owned and managed data center is around 1.8 times more energy-efficient than a standard enterprise data center. In 2023, our data centers achieved an average annual power usage effectiveness of 1.10, far surpassing the industry average of 1.58, which means our data centers consumed roughly 5.8 times less auxiliary energy per unit of IT equipment energy.

Recently, we introduced a water risk framework to better pinpoint climate-conscious cooling solutions that take into account the availability of carbon-free energy (CFE), watershed health, and future water requirements. We view our expanding infrastructure as an avenue to steer the innovations and investments necessary to drive a low-carbon economy.

AI carries immense potential for steering climate action. AI has the capability to help mitigate 5–10% of global greenhouse gas (GHG) emissions by 2030. We are propelling climate action through AI in three principal areas:

• Information Organization: AI is utilized for fuel-efficient routing, analyzing traffic, topography, and a vehicle’s engine to propose the most effective route. This initiative has facilitated over 2.9 million metric tons of GHG emissions reduction since its inception, equivalent to removing approximately 650,000 fuel-based cars from the roads for a year.
• Prediction Enhancement: We developed an advanced global hydrological AI model and merged it with publicly accessible data sources to forecast floods up to seven days in advance in over 80 countries, covering regions where more than 460 million people reside. This aids these communities in preparing for and responding to riverine floods.
• Optimization Improvement: Green Light is an AI-driven tool that assists city traffic engineers in optimizing traffic light timings to cut down on stop-and-go traffic and fuel consumption. This technology has the potential to lower stops by up to 30% and diminish emissions by up to 10% at intersections.

Through our products, we aim to aid individuals, cities, and other partners in collectively slashing 1 gigaton of carbon equivalent emissions annually by 2030. Additionally, we will persist in developing technologies that help communities adapt to the impacts of climate change.

Source: Google Blog