Training a single AI model can emit as much carbon as five cars do in their entire lifetimes.

As AI becomes more powerful, its environmental toll is garnering a lot of scrutiny–from the immense energy it uses to the e-waste it leaves behind. Yet, in a twist of irony, AI might also be our best hope for fighting climate change. 

So, will AI save the environment, or will it help destroy it faster? 

Environmental Impact

So, what is the impact of AI on the environment? How detrimental is it? The UN Environment Programme (UNEP) has identified AI’s direct environmental impacts. This includes the extraction of raw materials and minerals for hardware, data center construction and operation, and energy consumption. In addition to electricity for operations, water for cooling and it also produces waste and e-waste.

MIT research revealed that the rapid development of these generative AI models requires a staggering amount of electricity. This leads to increased carbon emissions and pressures on the electric grid. 

At the center of the issue are data centers, which are temperature-controlled buildings that house computing infrastructure, such as servers, data storage drives, and network equipment. Not only do they require massive amounts of electricity to operate but it also needs water to cool down electrical components. The UNEP reports that globally AI-related infrastructure may soon consume six times more water than Denmark, with a population of 6 million. This is largely concerning due to the water scarcity crisis many areas of the world are experiencing. 

While data centers have been around since the 1940s, the sheer power density required by AI training runs seven or eight times more energy than a normal load. Moreover, by 2026, the expectation is that the electricity consumption of data centers will approach 1,050 terawatts.

What does this mean for carbon emissions? Logically, AI contributes to higher carbon emissions; because of the huge amounts of electricity it needs to operate. 

However, it is hard to determine the amount of AI-based computation at a data center because of the lack of datasets. Moreover, the International Energy Agency reports that AI data centers currently account for 0.5% of carbon emissions. However, Carbon Direct reports that AI likely consumes 0.04% of global electricity based on AI computer chips sales. This results in 0.01% of global greenhouse gas emissions (GHG).

In a study conducted by the University of Massachusetts, it was found that the process of training large AI models can emit more than 626,000 pounds of carbon dioxide. This is equivalent to nearly five times the lifetime emissions of the average American car. 

The French-American AI Startup, Hugging Face, released a report estimating the overall emissions of its own LLM, BLOOM. Through comparison, it estimated Open AI’s GPT-3 to emit 500 metric tons of carbon dioxide, while Meta’s OPT emits more than 75 metric tons. The global tech sector accounts for 1.8-3.9% of global GHG emissions. While AI contributes only a fraction of that, its footprint is still considerable for a single field.  

The momentum of AI carbon emissions is not clear, it depends on how AI applications continue to roll out. Its net impact can only be controlled depending on how the applications will operate and the regulatory frameworks that will respond.

Environmental Benefits

So, how can AI save the environment? The UNEP argues that AI can play a role in tackling environmental challenges. This can range from designing more energy-efficient buildings, to monitoring deforestation, to optimizing renewable energy deployment. 

Despite its impacts on the environment, generative AI has immense potential–from streamlining workflows to accelerating scientific research. For instance, the UNEP also indicates that AI can be a tool to mitigate climate crises. Recently, it announced the Coalition for Environmentally Sustainable Artificial Intelligence at the Artificial Intelligence (AI) Action Summit in Paris. The coalition’s aim of ramping up global momentum to place AI on a more environmentally sustainable path.

A prominent advantage of AI is environmental monitoring, as it would be a step forward in the usually unclear landscape of carbon monitoring. The resulting data could help corporations determine how to lower their emissions footprint. Even more so, it can aid policymakers in their pursuit to hold polluters accountable.

In 2022, the UNEP launched the World Environment Situation Room, which is a digital platform leveraging AI’s capabilities to analyze complex, multifaceted datasets. Within this digital ecosystem is the International Methane Emissions Observatory (IMEO), which leverages AI to revolutionize the approach to monitoring and mitigating methane emissions.

There are simply monitoring tasks that are too hefty for humans, such as melting icebergs. AI has been trained to measure changes in icebergs 10,000 times faster than humans. In addition, it can also generate detailed maps of ocean litter, enabling more efficient waste removal efforts. 

By improving weather prediction models, AI can support climate resilience. For instance, in the UAE, the climate tech startup FortyGuard uses AI technology to cool cities by identifying, controlling, and cooling overheated outdoors. Furthermore, it monitors outdoor temperatures using cloud-based AI tools. Thus, allowing urban planners, developers, industries, and governments to make environmentally conscious decisions.

Environmental Cost

On the flipside, there are ethical concerns with using AI for climate mitigation. If these models are trained on skewed or incomplete data, they may be biased. For example an AI model that is trained to value economic growth over sustainability, might suggest policies that put short-term financial gain ahead of long-term environmental sustainability.

AI has to become part of the solution, not the problem, if it will actually mitigate the effects of climate change. This means reducing its own environmental footprint. With cleaner energy and sustainable practices, it could be well on its way to solve climate crises. 

The efficiency AI provides will always come at a cost, however, long-term, its benefits could be game-changing.  AI can run simulations instead of expensive real word testing, it can optimize electricity distribution or help design better batteries and solar panels. It can also contribute to the construction of energy-efficient buildings, generation of  low carbon materials, and monitor deforestation.

AI is already part of our collective reality. Its environmental impacts are undeniable, but so are its potentials. If used wisely, it could be a game-changer. If mismanaged, it risks becoming just another high-emitter helping accelerate climate change.