Posted on March 25, 2025 by Neil Bradley
#Corporate Responsibility #Corporate Sustainability Reporting #ESG Issues #Sustainability Reporting #Water
By Neil Bradley, Sustainability Analyst – G&A Institute
In recent decades, artificial intelligence (“AI”) has become an increasingly hot topic as one of the most transformative technologies available today that can be widely applied to solve complex, contemporary issues such as climate change and national security.
In the early 2000s, the U.S. Department of Defense kick-started practical applications of AI when its Defense Advanced Research Projects Agency (DARPA) began producing some of the first virtual assistant technology (Source: Boddu, 2023).
Since then, the absolute boom in technology advancement has driven demand for more efficient, complex, and customizable AI systems across the corporate world. The health care industry uses AI to analyze personal and public health data, allowing for early diagnoses and prevention as well as efficient tracking of public health metrics.
Across numerous global supply chains, in logistics for fulfillment and optimizing shipping routes, use of robotic AI systems has increased in popularity. In banking, AI powers automatic recognition systems to help prevent and mitigate fraud. Social media sites and entertainment providers — such as Netflix and Spotify — use complex AI-powered algorithms to tailor individualized content to consumers.
That’s not all — AI has become increasingly common for personal use, as well. For example, Apple released the “Siri” virtual assistant in 2011, using the same technology DARPA had developed for military use. In 2024, a combined 248.6 million U.S. consumers used virtual assistants like Google Assistant, Siri, or Alexa at least once a month.
Like virtual assistants, “chatbots” have also gained popularity in recent years, largely in part to industry leaders such as WeChat and ChatGPT. Chatbots allow users to interact with a virtual “bot”, which can generate human-like responses and assist with a huge variety of tasks. As of February 2025, ChatGPT alone has over 400 million active users per week!
Business applications of AI, as well as the increased accessibility of AI systems for personal use, have helped grow the worldwide AI market value to a projected US $243 billion USD in 2025, and this number is expected to continue growing exponentially over the next few years. By 2030, expert forecasts estimate the worldwide AI market value will reach nearly $827 billion USD.
Recently another emerging application of AI has been making waves, this time in the corporate sustainability space: the use of generative AI in creating content for companies’ ESG disclosures. Unlike traditional AI technology that uses coded script, generative AI analyzes data to generate tangible content ranging from images to text to music, and even predictive statistical models (Source: University of Illinois, 2024).
ESG reports typically take months to produce and require synthesizing global, cross-functional data, in addition to contributions from stakeholders, into a comprehensive and consistent public-facing picture of corporate sustainability performance. As explained by Luke Elder, Google’s Sustainability Reporting Lead, companies can use generative AI to summarize information, factcheck, and draft report content from employee inputs, which can help avoid months of back and forth between stakeholders.
However, neither traditional nor generative AI queries exist in a vacuum. Both make significant use of natural resources, and this creates social burdens. Training any AI technology requires massive amounts of electricity for computing input datasets, as does the operational phase in which end users submit queries. This is because powerful computers, servers, and other infrastructure stored in heavily resource-intensive data centers are necessary to analyze the sheer volume of information associated with potential queries.
On a global scale, 460 terawatt-hours of electricity was consumed by data centers in 2022, only three terawatt-hours less than the entire nation of France. By 2030, data centers could consume up to 10% of all U.S. electricity (Source: Davenport et. Al, 2024). Electricity used to power data centers is often generated using fossil fuel sources, which yield greenhouse gas (GHG) emissions. The resulting air pollution from data centers alone has cost the U.S. an estimated $5.4 billion in healthcare expenses since 2020, with Google accounting for nearly half of these expenses (Source: Darley, 2025).
In addition, AI processing is heavily dependent on water to absorb heat from computing equipment (Source: Zewe, 2025). Data centers’ heavy use of water and energy not only impacts the environment, but also the communities in and around the facilities that fuel these powerful technologies.
In 2022, an Iowa group of data centers powering OpenAI’s most advanced large language model, GPT-4, was the subject of a lawsuit by residents of the local district, who estimated that the company used 6% of the district’s water (Source: Crawford, 2024).
The proportion may be even higher in regions experiencing water stress, such as in California, where Google is based, or the rest of the American Southwest. AI’s environmental impacts affect each locality differently, a fact which does not always appear reflected in the company’s energy decisions: for example, Google uses 97% carbon-free energy at its data center in Finland, but only 4% to 18% of the energy consumed at its data centers in Asia — where water stress is more common — is carbon-free (Source: Ren & Wierman, 2024).
Given the social and environmental concerns linked to generative AI programs and data centers, should they be avoided by a company seeking to improve its impacts on the world? Not necessarily. Although generative AI has a heavy environmental footprint, some of the concern could be addressed by improving data center sustainability, particularly by reducing their negative externalities.
Because of the disproportionate impacts that surrounding communities bear, facilities need to look beyond typical green transition strategies such as purchasing carbon credits or transitioning to renewable sources over time. For example, entering into power purchase agreements (PPAs) or virtual power purchase agreements (VPPAs) may help companies reduce their global emissions totals, but will not necessarily lead to a meaningful reduction in GHG emissions specifically from data centers¹. Instead, experts suggest the following potential solutions:
Implementing innovative strategies like these could greatly improve data center sustainability, in turn reducing the negative impacts associated with use of generative AI.
Bringing this full circle: Suppose a company seeks to improve the social and environmental impacts of its operations, and then to report on them for stakeholder awareness. Should it reduce the human cost of preparing ESG disclosures with the help of generative AI? What if doing so creates more emissions and water consumption that then needs to be reflected, in turn, in the environmental disclosures the company is preparing? The research points to a careful look at tradeoffs, with two paths forward depending on the type of company.
First, companies that have operational control over the AI model being used in report drafting – like Google itself when preparing its own ESG disclosures – have an even greater responsibility to prioritize innovative data center strategies to reduce environmental impacts, such as those highlighted above.
Second, firms that do not have operational control over models but want to use AI as a tool in their ESG reporting process should prioritize due diligence in selecting their AI vendors. This may include calculating both the financial return on investment and environmental footprint of potential platforms, considering lower-footprint alternatives before landing on generative AI, and analyzing potential impacts across the entire lifecycle of the technology.
¹ This is because of a concept called additionality, which is crucial in the green data center transition. Additionality means that decarbonization projects are only “additional”, and not inherently beneficial, if they are implemented due to monetary incentives associated with carbon credits. Otherwise, leakage can occur, in which clean energy sources are diverted from existing users instead of truly increasing capacity (Source: https://offsetguide.org/high-quality-offsets/additionality/).
ABOUT THE AUTHOR
Neil Bradley – Sustainability Analyst, G&A Institute
Neil Bradley is a Sustainability Analyst at G&A Institute. His role is primarily focused on sustainability report development, ESG news research and analysis, and supporting a variety of additional client projects ranging from nature-related risk assessment to ESG datapoint gap analyses.
Published by Governance & Accountability Institute, Inc.
Design & Contents Copyright © 2005 – 2025 By G&A Institute unless otherwise noted. All rights reserved. Sustainability Update™ is a service mark of G&A Institute, Inc.
