Getty Images

AI to boost sustainability if carbon costs are kept in check

AI has the potential to drive ESG goals and improve sustainability outcomes, but using the tech also creates considerable environmental cost with the amount of energy needed.

AI, especially generative AI today, is an important set of technologies needed to advance sustainability goals -- but there are potential downsides to adoption.

AI and sustainability have become closely interlinked, and every IT vendor that has some stake in AI should think about how AI can help meet sustainability goals, which include environmental, social and corporate governance (ESG), according to Bjoern Stengel, an analyst at IDC.

This is a story about risks and opportunities. Potential downsides include the footprint of AI, the energy consumption, responsible AI – ethics, bias -- but there's also the potential that AI has to accelerate sustainable transformation.
Bjoern Stengel Global sustainability research and practice lead, IDC

"This is a story about risks and opportunities," Stengel said during the recent IDC Directions conference in Boston. "Potential downsides include the footprint of AI, the energy consumption, responsible AI -- ethics, bias -- but there's also the potential that AI has to accelerate sustainable transformation."

Willing spirit, but weak follow-through

IDC research suggests that interest in using AI to meet sustainability goals is significant. In a recent survey of IT decision-makers, 76% of 1,390 respondents reported AI as being either "critical" or "very important" to their organizations' initiatives to meet sustainability targets.

But there are questions about how far along enterprises are in implementing AI for sustainability use cases. Just 25% of survey takers said they are integrating AI for sustainability, and about 40% said that they are just getting started, Stengel said.

Deploying AI to meet environmental and energy usage goals is the most important sustainability concern for organizations, according to the survey. The top issues are managing energy demands and carbon emissions (30%); circularity, including e-waste and end-of-life management (29%); and water usage (14%). Deploying AI to help meet ethical or social practices was less prominent and include data security and customer privacy (13%); addressing unfair and biased decision-making (8%); and unethical business conduct (6%).

Sustainability concerns are also driving the procurement of AI technologies, and 68% of decision-makers said that sustainability considerations play either a "critical" or "very important" role in planning and procuring AI applications, according to the survey.

Sustainability use cases for AI

There are broad use cases for AI, according to Stengel. Traditional AI is being used to help manage energy usage and operations, including tracking the footprint of AI, gaining insights into energy consumption or optimizing operations.

Generative AI tools are being used more on the front end to help figure out regulatory requirements and pull together information from different data sources, Stengel said during his presentation. One reason genAI is being implemented for this is that few people in organizations are experienced with ESG reporting and there's an enormous amount of data involved, he added.

TechMarketView's "SustainabilityViews" report for 2023 found that the most common use cases for AI include analyzing environmental data for strategy and planning, ESG reporting, agritech, biodiversity monitoring, supply chain routing and smart building optimization.

Most of the AI applications for ESG are in project phases rather than product phases right now, according to Craig Wentworth, principal analyst at TechMarketView. AI tends to be used in combination with other technologies such as IoT and geospatial to build models that help organizations better understand climate-related issues such as extreme weather events or activities that have sustainability effects such as sewage spills.

"These tend to be instances of general problem sets, like pattern spotting or prediction but applied in sustainability scenarios -- all of which represent a great fit for AI," he said.

For example, AI is being used to extract relevant emissions data from reams of unstructured sources or modelling behavior to plug gaps where no data exists, and genAI is being used to generate final reports, Wentworth said.

The impact of AI will be broad across industries, as it's poised to revolutionize sustainability technology and services in terms of expertise, pace, innovation and cost, according to Ambika Kini, senior analyst for sustainability technology at Everest Group.

For expertise, AI enables personalized learning experiences and rapid creation of educational content that can facilitate upskilling and knowledge dissemination.

For pace, the automation and streamlining of processes and the querying of large data sets enables sustainability initiatives to be executed faster and accelerates iterations, analytics and simulations for use cases.

For innovation, genAI can facilitate the design and prototyping of more sustainable products across a variety of industries, such as pharmaceuticals and sustainable agriculture.

For cost, AI can be used to reduce manpower requirements and decrease the time and cost of deploying sustainability services. AI-driven applications can automate tasks and optimize resource allocation, making sustainability applications more cost effective and accessible to a broad range of organizations and industries.

Some enterprise use cases for AI.
A look at how industries such as manufacturing and retail are using AI for ESG.

Insatiable power needs dampen AI gains

Using AI for ESG initiatives comes at a high sustainability cost, however.

AI will increase the demand for power generation significantly, Stengel said. The percentage of AI-related energy consumption within data centers will increase dramatically, with an inevitable increase in carbon emissions.

"Most of the emissions are linked to the usage of the IT infrastructure, but you have to consider there's embedded carbon in building the infrastructure," he said. "The more physical assets you build, the more there is to consider around the circularity and end-of-life management of these products."

The impacts might be mitigated to a certain degree as the majority of AI will likely happen in the cloud in colocation data centers that are increasingly powered by renewable energy, Stengel said.

"It's a pretty complex equation between increase of renewable data sources, but also an increase in energy consumption and power capacity," he said.

Putting ESG data in the cloud might help meet environmental goals, but it can have negative effects on the social and governance pieces by introducing challenges for data privacy and sovereignty, Stengel said.

The increasing use of AI in sustainability comes at an environmental price, Wentworth agreed, and climate credentials of cloud providers must be questioned.

"This means not just averages across their networks worldwide, but specific questions about how low-emission the energy is that's used in data centers in each region, as that can inform when and where to run workloads," he said. "Plus, the 'green efficiency' of the code built around the AI to deliver overall applications, which is another area to query suppliers on when looking for the most sustainable AI applications."

Power consumption remains the primary concern as the adoption of AI increases, according to Kini. But many IT service providers are investing in R&D efforts to make their models greener to curb the energy consumption and reduce the carbon footprint.

"Cloud service providers are using green data center technologies and renewable energy sources to host their cloud and AI models, [and are] supplemented by mature carbon calculators that help measure the carbon emissions," she said.

There are several other challenges for AI applications in sustainability, Kini said, including algorithmic bias, privacy concerns around sensitive sustainability data, the transparency of the models and job displacement.

"Many AI models, particularly complex deep learning models, are often black boxes, making it difficult to understand how they arrive at their decisions and can hinder trust and accountability in sustainability efforts," she said. "As AI automates tasks, some jobs may be lost, so it's important to consider the potential social impact of AI implementation."

Jim O'Donnell is a senior news writer for TechTarget Editorial who covers ERP and other enterprise applications.

Dig Deeper on Sustainable IT

Data Center
Mobile Computing