The Environmental Impact of AI: Positive or Negative?

2025-03-24

Morgane Benoist | Sustainability Consultant

Share this article

In 2023, 30.5 billion digital devices were used by 5.35 billion people — an average of around six devices per person. Digital technologies are now an inseparable part of our daily lives, shaping how we work, communicate, and interact with the world. Among these technologies, artificial intelligence (AI) is rapidly gaining ground, transforming industries and redefining possibilities. But as AI’s influence grows, so does its environmental footprint — raising an essential question: is AI’s overall impact on the planet positive or negative?

The Rising Use of AI and Digital Technologies

AI is rapidly expanding across industries, from manufacturing and logistics to healthcare and entertainment. Many companies are using AI to plan and optimize operations, leading to savings in energy, money and resources. Here are some publicized examples where AI led to energy savings:

AI may contribute to reduce commercial buildings energy consumption and carbon emissions by 8% to 19% by itself, while even higher reductions could be observed combined with low-carbon power generation and energy policies.
Rakuten Mobile has developed an AI model that can reduce network power consumption by 25%
Mitsubishi UFJ has implemented Enneteye in 200 facilities nationwide. Enneteye is an AI-powered service that visualizes energy consumption and provides tailored recommendations for energy-saving actions. During a one-year trial, the bank achieved a 9% reduction in electricity consumption during winter and a 7% reduction in summer.

From an environmental perspective, AI models are already helping monitor pollution, track biodiversity and predict climate phenomena by finding patterns and providing data to researchers, supporting positive environmental outcomes.

During the current ‘AI spring,’ models such as ChatGPT, DALL-E, Stable Diffusion, and Midjourney — all first released in 2022 — have helped a wider audience adopt and integrate AI technologies into their daily lives and work. According to a study by Boston Consulting Group, 43% of employees globally used AI models at least once a week in 2024, with this proportion rising to 82% among executive-level employees. In this study Japanese respondents were the least convinced and the most anxious about its results. The proportion of employees using AI was also lower than the world’s average, which is explained in the study by a lack of employee training.

AI: Bane or Boon for the Environment?

The environmental costs of AI are complex. While AI currently accounts for only a small fraction of global energy consumption (estimated at 0.03%), this figure is set to grow exponentially as AI becomes embedded in everyday software (with one of the latest examples being Microsoft Copilot being automatically activated on Windows). Microsoft president Brad Smith highlighted this challenge in a 2024 Bloomberg interview:

“In 2020, we unveiled what we called our carbon moonshot. That was before the explosion in artificial intelligence. So in many ways the moon is five times as far away as it was in 2020, if you just think of our own forecast for the expansion of AI and its electrical needs.”[1]

Even with efforts to optimize AI models, the so-called “rebound effect” remains a challenge. As technology improves and energy efficiency rises, demand often increases — wiping out the environmental benefits. This pattern emerged with the transition between 3G and 4G networks, and a similar race for faster, more powerful AI models is now underway. According to estimates, the AI industry is projected to consume ten times more electricity in 2026 compared to 2023.

(Source: “Electricity 2024 Analysis and forecast to 2026” – IEA)

The environmental impact isn’t limited to electricity consumption. The manufacturing of digital devices or data centers requires rare metals like indium, gallium, and tantalum — resources that are becoming scarcer. Mining these materials often involves harmful environmental and social practices. Moreover, AI infrastructure demands significant water resources. Training large models like GPT-3 consumed over 700,000 liters of water — used both for cooling servers and for generating renewable energy to power them.

Policies for sustainable AI still sparse

Policymakers have begun regulating AI, but environmental concerns remain secondary.

The Hiroshima AI Process, launched in May 2023 under Japan’s G7 presidency, aims to promote the “safe, secure, and trustworthy development and use of AI” and establishes 12 principles to guide stakeholders throughout the AI lifecycle. Principle 9 encourages the development of AI for “global benefit,” addressing challenges such as climate change. However, it does not mention mitigating the environmental impact of developing the AI tools intended for these purposes.

The EU AI Act, passed in 2024, primarily addresses privacy, transparency and safety. It goes a little further than the Hiroshima AI Process by encouraging companies to adopt voluntary eco-friendly AI practices but lacks enforceable environmental mandates. It also focuses mainly on the impact of energy consumption but other points, such as water usage, are not considered.

In Japan, the AI Operator guidelines (AI事業者ガイドライン) , published by MIC and METI in 2024, similarly mention environmental considerations in passing. A similar document, the Guidebook on Measures to Address Risks in the Use of Generative AI (テキスト生成AI利活用におけるリスクへの対策ガイドブック ) published the same year by the Digital Agency likewise prioritizes privacy and fair use over sustainability.

Worldwide directives on sustainable AI seem off the table for now: during the AI Summit in Paris in February 2025, both the US and UK refused to sign a non-binding document stating that “making AI sustainable for the people and the planet” was one of the priorities for inclusive and sustainable artificial intelligence, despite 60 countries including France, China, India, and Japan signing it.

In parallel to this summit, a Coalition for Sustainable AI has been launched around more than 100 companies, with 14 countries and 7 international organizations supporting it , with objectives such as advocating for the development of methods to assess AI’s environmental impact, or ensuring that AI infrastructure and software will be built and maintained in line with global environmental commitments, showing there is a budding movement to advocate for a greener AI.

The next few years will likely see regulations and more binding sustainability guidelines emerging in several countries, but without stronger environmental directives at the moment, companies face a choice: wait for regulation to catch up or proactively integrate sustainability into their digital and AI strategy. The latter option not only helps future-proof companies, but also position them as sustainability leaders.

So, what should companies do to encourage a sustainable use of AI?

In an era driven by digital transformation (DX), reassessing digital habits may feel counterproductive and even contrary to the prevailing trend. Still, there are 3 steps that companies can take to reduce their environmental footprint linked to AI and digital devices use:

Calculate digital emissions

The first step to consider actions is to first assess the situation. Currently, when calculating a company’s Scope 3 GHG emissions, Internet and AI usage fall under the “Purchased Goods and Services” and is relying on secondary data, making it hard to isolate AI’s specific impact. Companies should first seek localized data for greater accuracy, and then include all devices -computers, smartphones, network usage- in their Scope 3 calculations.

Recognize the importance of the digital footprint and promoting digital responsibility towards all employees

Over half of digital technologies impact on the environment stems from user equipment. This highlights the importance of reducing unnecessary device use and extending hardware lifespans. Companies without a digital sustainability plan should create one, encompassing all digital technologies — not just AI — and encouraging mindful technology use.

Reconsider AI’s necessity

While leaving even the simplest of tasks to AI leads to time gains (and avoided headaches), not all tasks require AI. In 2023, Alphabet (Google) reported an AI-generated response consumed 10 times more energy than a traditional Google search. This represents a significant additional energy load to a company’s Scope 3 emissions. Sometimes, a simpler approach (even using paper!) is more energy-efficient.

Conclusion

AI holds potential to drive progress and tackle environmental challenges —but it’s not a silver bullet. The rebound effect remains a persistent obstacle: as AI models become more efficient and accessible, usage skyrockets, offsetting the energy savings. This cycle, already seen with past technological advances, threatens to repeat unless businesses and policymakers take deliberate action.

Technology alone won’t solve the environmental crisis. Companies must rethink their approach to digital consumption, question whether AI is necessary for every task, and prioritize sustainability alongside innovation. Only through conscious, informed choices can we ensure that AI’s benefits outweigh its environmental costs.

For those interested in a deeper understanding of digital technologies’ environmental impact, joining a Digital Collage workshop is a great next step. This workshop — built on insights from leading research — breaks down the complexities of digital sustainability into an accessible, card-based format designed to foster awareness and actionable change.