Sustainable AI: Balancing innovation and eco-responsibility

While prompting an AI model or asking AI a question might save you some energy, energy flows from somewhere. But how much, and what does it mean for our planet and communities? Read on to raise your level of awareness on the environmental impacts of AI and simple ways you can practice more intentional AI that’s better for you, your community, and our planet.

What makes generative AI different?

Artificial Intelligence (AI) refers broadly to technologies that simulate human intelligence in machines, a concept that has been around for decades. From chatbots and speech recognition to self-driving cars, AI encompasses a wide range of capabilities.

Today, when most people mention “AI,” they’re often referring to generative AI – a specific subset powered by large language models (LLMs). What sets generative AI apart is its accessibility, its ease of use, and its ability to interpret and respond to plain, unstructured language. The outputs are easy to understand, making the technology broadly approachable. As a result, anyone with internet access can interact with it.

AI at Rocket

At Rocket, generative AI is accelerating the way we work. Long gone are the days of finding a designated note taker in meetings; Slack AI will summarize huddles, bullet points, and action items. A data scientist forgetting the specific syntax for a rarely used function no longer needs a Google search and deep dive through Stack Exchange – just ask Copilot.

Plus, Rocket’s in-house generative AI tool, Navigator, gives every team memberquick and easy access to cutting edge models leveraging internal documents and processes for their outputs. Navigator also supports custom app development, allowing users to add context and fine-tune generative AI to specific use cases.

There’s no question that AI is an incredibly valuable and powerful tool, especially for boosting impact and creating efficiencies, but it’s equally important to recognize the high environmental price.

Environmental impact of generative AI

Climate change is no longer a distant threat – it’s something we’re feeling right now. From rising sea levels and more extreme storms to hotter temperatures and devastating wildfires, the planet is under pressure. In this climate crisis, every pound of CO2 matters, and every drop of freshwater counts.

The causes of climate change are complex and far-reaching, but one major contributor is carbon emissions. We’re already seeing serious effects:

• Across the U.S., flooding and droughts have become more intense since the early 1900s.
• Here in Rocket’s hometown of Detroit, average temperatures have gone up by 1.4 degrees.
• Southeast Michigan has seen 11% more annual rainfall compared to past decades.
• The great lakes – especially Lake Erie – are now more vulnerable to harmful algal blooms, which threaten both people and wildlife.

Generative AI, while powerful and full of possibilities, also uses a lot of resources.  That impact often gets overlooked, but it’s worth paying attention to. As users of AI, we assume the important responsibility to ensure AI is sustainable, safe, and inclusive – especially when it comes to our work.

• The data centers used to power these chatbots already suck up about 2% of the world’s energy consumption [16]
• Training for GPT-3 caused 552 tons of CO2 emitted, equivalent to the average power used by 120 U.S. homes for 1 year [1]
• 700,000 liters of water were required to train GPT-3 [3]
• A single query uses 5 x more electricity than a simple web search [5]
• GPT-4 uses 17 oz of water to write one 100-word email [3]
• In January 2023, GPT-3's usage equaled the energy consumption of 175,000 humans [2]
• Projections estimate that generative AI will require 4 - 21% of the world’s electricity supply by 2030 [10] [1]
• GPT-4 had 570x the parameters of GPT-3 [2]

These trends are troubling, but not irreversible. The good news? There’s a better way forward: Sustainable AI.

What is sustainable AI?

With smarter choices, cleaner infrastructure, and more-efficient technology, we can harness the power of AI without accelerating the climate crisis. “Sustainable AI is a movement to foster change in the entire lifecycle of AI products (i.e. idea generation, training, re-tuning, implementation, governance) toward greater ecological integrity and social justice” [4].

Sustainability includes more than just environmental impacts, but for the purposes of this article, we are focused on environmental sustainability. However, the solutions for environmental sustainability are frequently the same solutions as social sustainability.

Sustainability takes many forms. In the context of generative AI, this can include things like monitoring consumption and developing new technology to reduce the cost of training and computing. This can be done by using sustainable materials to create hardware to reduce electronic waste, creating thoughtful data center designs that can help lower the volume of water loss, and investing in sustainable energy sources like solar and wind, which also reduces the carbon footprint of generative AI.

What is being done?

At the scale generative AI operates, energy and water consumption aren’t just environmental concerns – they’re expensive. And when costs climb, even the most profit-driven companies take notice.

That’s why tech giants like Google, Amazon, and Meta are investing heavily in more energy and water-efficient technologies at their data centers [7] [13]. Microsoft has gone a step further by pledging to be carbon negative by 2030 by doing things like investing in carbon reduction and removal technology [14].

These high-level efforts signal a growing shift toward more sustainable AI. While the motives may vary, the result is the same: greener infrastructure for more resource-intensive technology.

It’s not just big tech. Individuals can also make a big difference by using AI more mindfully and efficiently.

Navigating AI, intentionally

If you’re a generative AI user that uses the biggest model for even the smallest tasks, you might be harming yourself, and the planet in the process. While individual consumers may not have any say in the environmental cost of training generative AI models, we do have control over how and why we use them. Even small changes can make a big difference. Most importantly, stay informed – and be sure to tune in for more updates from the sustainable AI group at Rocket.

Like all things, it’s important to remember balance when using AI, and not falling into the (admittedly easy) habit of “just ask AI.” Instead, try to incorporate some simple, sustainable AI practices in your AI encounters.

Examples of sustainable AI

Sustainable AI at Rocket

As we continue to explore the potential of generative AI, we’re also aware of its environmental impact. We know this is a shared responsibility – and we’re committed to doing our part.

As team members, we can lean on the pillars of our ISMs to have a positive impact with sustainable AI.

• We can raise our level of awareness by keeping up with current research and staying informed, sharing research among peers, and joining Rocket’s sustainable AI community. This enables us to find a better way by discovering new and efficient ways of interacting with, building, and deploying models.  By sharing and collaborating on ideas, we can push ourselves as individuals and a company to mitigate negative impacts.
• Launching and learning our findings and ideas empowers us to move quickly and keep up with the rapidly changing pace of generative AI.

Rocket’s sustainable AI community has already been working together to brainstorm ideas, including monitoring, data analysis, and sharing knowledge on using generative AI in more sustainable ways. We don’t have all the answers yet, but we’re asking the right questions and making thoughtful choices with sustainability in mind. Every step forward, no matter how small, is part of the bigger picture. Together we can shape a future where innovation and environmental responsibility go hand in hand.

Contributors to this article include Tina Han, Data Scientist, Shivali Yedulapuram, Associate Machine Learning Engineer at Rocket Mortgage and Jennifer Shapiro.

References

1. Foy, Kylie. (2023, September, 22) "AI Models Are Devouring Energy: Tools to Reduce Consumption Are Here if Data Centers Will Adopt." MIT Lincoln Laboratory, www.ll.mit.edu/news/ai-models-are-devouring-energy-tools-reduce-consumption-are-here-if-data-centers-will-adopt.
2. D. M. Dvořák, Pikulik, M. and Kroupa, T. (2024) "Resource Efficiency in Data Centers with Machine Learning.” ScienceDirect.  https://www.sciencedirect.com/science/article/pii/S0925231224008671#fn3.
3. Johnson, R. (2024) "Generative AI Data Center Water Use." TechRepublic,  https://www.techrepublic.com/article/generative-ai-data-center-water-use/.
4. A. R. Bhatia, M. R. Van Dijk, and Singh, R.K (2021) "Sustainable Artificial Intelligence." SpringerLink. https://link.springer.com/article/10.1007/s43681-021-00043-6.
5. Zewe A. (2025, January, 17) "Explained: Generative AI's Environmental Impact." MIT News. news.mit.edu/2025/explained-generative-ai-environmental-impact-0117.
6. Unspecified. (2025, June, 13) "Climate Change and Water." NOAA. www.noaa.gov/education/resource-collections/climate/climate-change-impacts.
7. Ren, S. and Wierman, A. (2024, July, 14) "The Uneven Distribution of AI's Environmental Impacts." Harvard Business Review. hbr.org/2024/07/the-uneven-distribution-of-ais-environmental-impacts.
8. Lindwall, C. (2022, October, 24) "What Are the Effects of Climate Change." Natural Resources Defense Council.  www.nrdc.org/stories/what-are-effects-climate-change#humans. 
9. Climate Change Impacts. University of Michigan, https://glisa.umich.edu/wp-content/uploads/2021/02/DCAC_Climate_Impacts.pdf.
10. Yao, Y. (2024, October, 10) "Can We Mitigate AI’s Environmental Impacts?" Yale Environment 360, Yale School of the Environment. https://environment.yale.edu/news/article/can-we-mitigate-ais-environmental-impacts.
11. Unspecified. (2025, June, 1) "How AI Use Impacts the Environment." World Economic Forum, www.weforum.org/stories/2025/06/how-ai-use-impacts-the-environment/. 
12. Moss, S. (2024, December, 9) "Microsoft's Zero Water Evaporation Data Centers." Data Center Dynamics. www.datacenterdynamics.com/en/news/microsofts-upcoming-data-centers-to-use-closed-loop-zero-water-evaporation-design/.
13. Jones, N. (2018, September 12) "Artificial Intelligence and Climate Change." Nature.  http://www.nature.com/articles/d41586-018-06610-y.
14. Smith, B. (2020, January, 16) "Microsoft Will Be Carbon Negative by 2030." Official Microsoft Blog. blogs.microsoft.com/blog/2020/01/16/microsoft-will-be-carbon-negative-by-2030/.
15. Unspecified. (2024, May, 14) “AI is poised to drive 160% increase in data center power demand” Goldman Sachs. https://www.goldmansachs.com/insights/articles/AI-poised-to-drive-160-increase-in-power-demand.
16. Altman, S. (2025, April, 19) “Sam Altman Admits That Saying "Please" and "Thank You" to ChatGPT Is Wasting Millions of Dollars in Computing Power.” Futurism. https://futurism.com/altman-please-thanks-chatgpt.
17. Tech Seven Partners. (2024). AI etiquette: Why you should add a please and thank you when using AI. https://techsevenpartners.com/ai-etiquette-why-you-should-add-a-please-and-thank-you-when-using-ai/.