In recent years, artificial intelligence (AI) has become an integral part of our daily lives, powering everything from social media algorithms to medical diagnostics. However, as AI technologies rapidly evolve, the environmental impact of these systems has been largely overlooked, with a growing concern over their carbon footprint. With the rise of AI, carbon accounting—tracking and reducing the environmental impact of various industries—has gained prominence. However, this focus has largely obscured the specific environmental costs of AI itself, which have remained hidden in plain sight.

AI’s energy consumption has skyrocketed as machine learning models become more complex and require immense computational power. A recent study by the Shift Project, a French think tank, revealed that the global data storage and transmission networks used to support AI systems account for about 1% of global energy consumption, a figure expected to double by 2030. Moreover, a single AI training session, particularly for large language models (LLMs), can produce up to 284 tons of CO2 emissions—equivalent to the lifetime emissions of five average American cars.

The key challenge in carbon accounting for AI lies in its indirect emissions. Unlike traditional industries such as manufacturing, where emissions can be tracked through direct production activities, the environmental cost of AI is distributed across vast networks of data centers, cloud services, and energy-hungry algorithms. For example, training a single AI model often requires thousands of processors running for days or even weeks. This intensive use of electricity—especially when sourced from fossil fuels—contributes to the carbon emissions of AI technologies.

Furthermore, the complexity of modern AI models, such as GPT-4 or Google’s Bard, requires not only vast amounts of energy to train but also to fine-tune and deploy. The cloud infrastructure, which hosts these models, is one of the largest contributors to AI’s carbon footprint. In 2020, data centers around the world were responsible for about 2% of global emissions, a number expected to rise as AI adoption continues.

Despite this, AI’s carbon footprint is often minimized or ignored in corporate sustainability reports. Many tech companies that deploy AI systems focus on carbon offset initiatives or sustainable energy sources for their data centers, but these measures only partially mitigate the environmental cost. Without a thorough understanding and transparent accounting of AI’s energy consumption, it becomes difficult for businesses and consumers to make informed decisions about how to reduce their environmental impact.

One step towards addressing this issue is the development of green AI—a movement that advocates for AI research and applications to be designed with sustainability in mind. This includes efforts to reduce the energy needed for training models, the development of more efficient algorithms, and using renewable energy sources for powering AI data centers. For instance, Google has reported that it uses renewable energy for all its data centers, and Microsoft has pledged to be carbon negative by 2030. However, these initiatives alone may not be enough to curb the growth of AI’s environmental impact.

Another challenge is the difficulty of accurately measuring and reporting the carbon footprint of AI technologies. As AI continues to advance, it’s essential to establish robust standards for carbon accounting in the tech industry. The AI community needs to embrace transparent reporting practices, and regulations may need to evolve to ensure that the environmental costs of AI are fully accounted for and addressed.

The carbon footprint of AI is not a theoretical problem—it is a growing reality that has immediate implications for the environment. As AI becomes increasingly embedded in every industry, its hidden environmental costs must be recognized and mitigated. By better accounting for the energy consumption and emissions of AI technologies, and pushing for more sustainable AI development practices, we can begin to balance the benefits of this transformative technology with the need to protect the planet for future generations.