Navigating the Complex Intersection of AI and Sustainability: A Data-Driven Examination of Energy Consumption and Environmental Implications

Julia Daviy
5 min readJul 7, 2023

The transformative power of generative artificial intelligence (AI) has created unprecedented opportunities for addressing global challenges, including climate change and sustainability. However, the rapid development and widespread adoption of generative AI have also exposed potential threats to environmental sustainability. In this article, we provide a data-driven examination of the complex relationship between AI and sustainability, delving into the potential energy consumption, e-waste generation, and hidden environmental costs of AI-based solutions.

The Dual Impact of AI on Sustainability

On the one hand, generative AI has the potential to accelerate progress toward the United Nations’ Sustainable Development Goals (SDGs) by optimizing energy consumption, predicting renewable energy generation, improving energy grid management, and reducing waste. AI-driven urban planning can lead to smarter, more efficient cities, reducing their environmental impact and improving residents’ quality of life.

On the other hand, the substantial energy consumption and carbon emissions associated with AI models raise concerns about their environmental impact. Escalating energy requirements contribute to the strain on the global power grid and the depletion of non-renewable resources. Additionally, the rapid development of AI technology and the constant need for updated hardware may lead to an increase in electronic waste, exacerbating soil and water pollution.

AquaFusion 2100: Harnessing Modular Fusion Energy for Desalination & Natural Ecosystem Restoration — Sustainable Innovations Challenge by Julia Daviy x AI

Hidden environmental costs

AI-driven solutions in agriculture, transportation, and manufacturing may inadvertently contribute to environmental degradation. Here are five examples illustrating such hidden environmental costs:

  1. Increased Energy Demand

AI solutions in manufacturing and transportation often require substantial computational resources, leading to increased energy consumption. Even AI systems aimed at optimizing energy use, such as smart grids, need substantial power to function, adding pressure on the already strained energy sector.

Large-scale AI models, such as OpenAI’s GPT-3, can consume vast amounts of energy during their training and inference processes. Data centers supporting AI applications consume around 200 TWh of energy per year, accounting for about 1% of global electricity consumption.

2. Resource Extraction & Inequality

AI technologies, particularly those requiring advanced hardware, often rely on rare earth elements and other non-renewable resources. The extraction of these materials can lead to habitat destruction, water pollution, and other forms of environmental degradation.

The high costs associated with implementing and maintaining advanced AI systems may exacerbate resource inequality between developed and developing nations. By 2025, the global AI market is expected to reach USD 190.61 billion, potentially leaving resource-poor nations struggling to access the necessary infrastructure and resources to adopt sustainable AI technologies.

3. Carbon-intensive Supply Chains

AI-enabled optimization of supply chains can sometimes favor efficiency or cost-reduction at the expense of environmental sustainability. For instance, AI algorithms might prioritize faster shipping routes over more fuel-efficient ones, leading to increased carbon emissions. The energy-intensive training process of models like Google’s BERT contributes to carbon emissions and climate change. BERT generates 1,438 pounds of CO2 emissions per GPU during training.

4. Monoculture Farming

AI-driven precision agriculture aims to increase crop yields and efficiency. However, it may unintentionally promote monoculture farming. This practice, while profitable in the short term, threatens biodiversity, degrades soil health, and makes crops more susceptible to diseases and pests.

5. E-waste accumulation

As industries turn towards AI-powered automation, the replacement of older, non-AI equipment can lead to an increase in electronic waste. In 2019, e-waste generation reached a record of 53.6 million metric tons, with only 17.4% being recycled. The improper disposal of obsolete equipment and components can have detrimental environmental consequences, such as soil and water pollution.

Energy Consumption Projections for ChatGPT and Other AI Applications

AI applications’ energy consumption is expected to rise significantly as their popularity and usage grow. Let’s consider ChatGPT as an example:

Current energy consumption: according to existing estimations, at this stage ChatGPT consumes between 1.1 million to 23 million kWh of electricity per month for 13 million users. This range is equivalent to the annual electricity consumption of approximately 100 to 2,150 American households.

If we base our projections on linear growth with user numbers, we can estimate the energy consumption for 244 million users by scaling up the existing numbers.

Therefore, with 244 million users in June 2023, the projected monthly energy consumption of ChatGPT could be equivalent to the energy consumption of about 225,000 American households.

For a larger perspective, the high-end estimation is roughly equivalent to the annual electricity consumption of small countries like Grenada or Saint Kitts and Nevis.

Projected energy consumption for 1 billion users: If we assume that generative AI applications like ChatGPT reach 1 billion users by the end of 2023, the energy consumption could range from 84.6 billion to 1.77 trillion kWh per year, given a linear growth in energy consumption with user numbers.

This projected consumption is roughly equivalent to the annual electricity production of Spain or Italy.

Addressing the Complex Intersection of AI and Sustainability

To harness the potential of AI while mitigating its environmental challenges, a comprehensive approach that fosters collaboration across sectors and disciplines is essential:

  • Make AI greener: Create smarter, energy-saving AI systems and equipment. This can help reduce energy use and lower carbon emissions.
  • Use AI to help the planet: AI can be used to better manage power grids and increase the use of renewable energy. This can help us use energy more efficiently across the globe.
  • Team up for the environment: Bring together AI experts, environmental scientists, and policymakers. By working together, they can tackle the environmental challenges linked to AI and prevent unwanted effects.
  • Set clear rules: We need easy-to-understand rules and standards for developing AI. These should focus on sustainability and ensure that electronic waste is disposed of responsibly.

The intricate relationship between AI and sustainability underscores the importance of adopting a balanced approach to AI development and implementation. As AI continues to reshape our world, it is crucial to recognize and address the potential environmental challenges, such as the projected increase in energy consumption and e-waste generation, while harnessing its transformative potential for sustainable development.



Julia Daviy

Write about Impact & Innovations | Co-Fndr Imageneria, Ib Centre. About: | Berkeley Haas | Co-Chair Berkeley Female Founders & Funders