Skip to main content


applied intelligence Live

16 Oct 2023

How can we sustain AI growth and protect the planet?

How can we sustain AI growth and protect the planet?
In the last five years, the field of AI has made significant progress, with breakthrough applications in autonomous driving, language translation, security systems, and generative transformations.

However, as AI use grows, so too does its carbon footprint. Omdia and AI Business’s report AI Sustainability: Two Sides of a Coin explores the relationship between AI and sustainability, looking at the extent of the environmental issues created by AI computation, how AI expansion can be managed to reduce such costs, and whether AI’s use in fighting the climate crisis compensates for its own impact.  

There is an environmental cost to training and deploying AI models

While AI has become more commonplace in workforces, these new AI processes and applications do consume a huge amount of energy: more electricity is required to train a single average-sized machine learning (ML) model than 100 US homes use in an entire year. Data centers often use thousands of chips inside thousands of server racks and require significant water and electrical input to run. The combination of this high energy use and companies’ desire to explore ever more use cases in GenAI could cause a new environmental crisis.

AI is also being used in the fight against climate change

However, AI is also being used for sustainability to achieve sustainable development goals in water, air, and energy conservation. It is combatting environmental issues in at least six major areas:

  • Reducing energy consumption: Renewable energy accounts for less than 15% of electrical generation worldwide. Therefore, an organization reducing its energy consumption will directly improve its sustainability.
  • Improving air quality: AI can assess the impact of air quality on the environment and guide policy development in this area.
  • Optimizing logistics: Organizations’ supply chains often account for more than 90% of their carbon emissions—making this an area ripe for AI optimization.
  • Eliminating defective production waste streams: Product waste, including from common practices like free customer returns, has a hidden but important carbon cost. Many returns could be avoided with better quality control, which is where AI comes into play.
  • Predicting, preparing for, and recovering from natural disasters: AI-based early warning systems have the potential to save lives and help in the recovery from a natural disaster. For example, ML models can be used to classify earthquakes and immediately assess the Tsunami risks.
  • Analyzing greenhouse gas generation: It is difficult for businesses to analyze how business travel, commuting, waste, and third-party deliveries affect their greenhouse gas generation. AI can help companies by scanning a large amount of online text that may otherwise be inaccessible to them.

Undoubtedly, there will be new uses and applications of AI to help in the fight against climate change as we unlock AI’s full potential.

Measuring and monitoring the environmental impact of AI processes is key

Additionally, effective management can help mitigate some of AI’s negative impacts. Few businesses track the energy cost of an IT project, but they will likely need to capture and monitor the energy consumption data for any AI processes moving forward. Companies should consider the carbon footprint of training the model, running inference on the model, and of the supporting computing hardware. It is also important to examine how and where data is stored: organizations should consider the environmental advantages of certain sites and opt for more environmentally friendly locations, using infrastructure specifically tuned to the needs of AI training where possible. Finally, there should be increased transparency around each model’s methodology. AI does not provide audit trails to explain its decision processes and, partly for this reason, both its capabilities and dangers have been overstated. All too often, a company builds its models with the best of intentions, but due to a lack of instrumentation and measurement, doesn’t detect when the model is being unfair, or even incorrect. AI transparency can avoid downstream waste caused by incorrect AI results.

Whether AI’s uses in countering climate change outweigh its own environmental impacts remains to be seen. In order to sustain the growth of AI and manage its acceleration while minimizing its environmental damage, organizations need to closely manage model training and inferencing, as well as the operational impact on the environment. With the increasing number of international sustainability standards, companies can legally require vendors to disclose AI energy consumption and improve ML toolchain efficiency. Businesses making use of AI must pay attention to both sides of the debate, accurately representing AI’s cost and value to enable sustainable AI development.

Click here to access the full report and find out more. You can also listen to the experts at The AI Summit New York this December 6-7, at the Javits Center, who will be discussing the latest insights into sustainable AI practices and practical applications. Secure your place now.


Explore Our Full Blog Library