Microsoft and Google, two tech giants driving the expansion of artificial intelligence (AI), are facing challenges in balancing their green goals with the growing emissions resulting from their datacenter infrastructure. Both companies acknowledge the environmental impact of AI growth and the need to address it. Microsoft's scope 3 emissions have increased by over 30% since 2020, while Google has reported a 48% increase in greenhouse gas emissions over the past five years [7821f5ed].
Google has reported an increase in greenhouse gas emissions in their latest Environmental Report. This is mainly due to their data centres that have expanded as a way to support AI developments, that use about 7-10% of the electricity used by all data centres around the world. Last year alone, emissions grew by 13% from the previous year, totaling 14.3 million metric tons of carbon dioxide. Servers in the data centres require cooling systems and as they expand, more servers are being powered. As a result, Google’s water consumption went up from being at 6% in 2022, to 18% being last year. Google is on a mission to power all its operations with completely clean energy by 2030. They’ve already managed to keep their energy usage 100% renewable for 7 years straight. With the planned advancements in AI, Google plans to reduce the energy needed to train AI models, which in turn makes these processes up to a hundred times more energy-efficient [73f8c4f8] [7821f5ed].
Microsoft plans to invest £2.5 billion over the next three years to expand its AI datacenter infrastructure in the UK and has announced new datacenter projects worldwide. However, the rapid growth of AI models and their associated hardware is driving up electricity consumption in datacenters. The International Energy Agency predicts that datacenters' total electricity consumption could double from 2022 to 2026, reaching 1,000 TWh. By 2030, AI is expected to account for 4.5% of global energy generation [bdfcbd31].
Google's Chief Sustainability Officer, Kate Brandt, acknowledges the challenges in balancing AI growth with emission reduction. However, Google remains committed to its goal of reaching net zero emissions by 2030. The company aims to achieve this by investing in renewable energy and enhancing the efficiency of its datacenters [d5ae60dd].
Google's 2024 Environmental Report highlights the company's use of AI to drive environmental change and sustainability. They have achieved 90% carbon-free energy in 10 grid regions and contracted 4 GW of clean energy generation capacity. Google's AI initiatives include fuel-efficient routing, flood prediction, and an AI-based Green Light tool. They aim for net-zero emissions by 2030 and have replenished 1 billion gallons of water. Despite progress, GHG emissions rose by 13% in 2023. Google is investing in clean energy projects and emphasizes the need for systems-level change and collaboration [75d3505d].
The energy consumption of AI is not limited to datacenters. Large language models, such as ChatGPT and Gemini, consume significant amounts of energy during training and inference. Analysts estimate that 90% of the energy cost of AI is in the inference phase. Redirecting and retraining human labor in other fields could be a potential solution to mitigate the environmental impact of AI. Despite the energy-intensive nature of AI, it can also save energy in other areas, such as writing and illustrating, compared to human labor [bdfcbd31].
To address the environmental impact of AI, both Microsoft and Google are investing in renewable energy and working towards carbon neutrality. Microsoft aims to become carbon negative by 2030, while Google aims to achieve net zero emissions by the same year. Both companies recognize the need to balance AI expansion with emission reduction and are taking steps to mitigate the environmental impact of their operations [bdfcbd31] [d5ae60dd] [7821f5ed].
Google's climate emissions have risen by 48% since 2019, drawing attention to the environmental impact of artificial intelligence (AI) in the tech industry. AI applications consume significant amounts of electricity, with data centers consuming 40% of electricity on computing and cooling. The global energy demand of data centers, cryptocurrencies, and AI combined reached 460 TWh in 2022, and projections suggest it could double by 2026. AI alone could consume between 85.4-134.0 TWh annually. Tech companies are exploring renewable energy sources to reduce their carbon footprints. The intersection of AI and climate change remains a critical issue for the future [274d7584] [39b93e1b].