Estimating AI’s Carbon Footprint: A Necessary Trade-Off for a Better Future?

Introduction

Artificial intelligence (AI) is transforming industries, promising innovations that improve productivity, health, and safety. However, with its rapid expansion comes rising concerns about its environmental impact. A recent study by Meng Zhang and colleagues quantified AI’s carbon footprint, providing a clearer picture of its energy demands and emissions. But how significant is AI’s impact, and how does it fit into the broader context of global challenges? Here ShrinkThatFootprint argues that the impact is actually quite small compared to the potential positive impact of AIs, but nevertheless there is an opportunity to reduce its carbon footprint.


Contextualizing AI’s Carbon Footprint

ShrinkThatFootprint has characterized the carbon footprint of training GPT-3. Since the release of GPT-3, AIs and more accurately, large language models (LLMs) has achieved broad usage. Correspondingly one would imagine that with hundreds of millions, possibly billions of users, this software may be incurring a heavy carbon cost. Zhang’s study found that from 2020 to 2024, 79 prominent AI systems collectively emitted 102.6 million metric tons (Mt) of CO₂ equivalent annually, a figure comparable to the emissions of smaller countries like Iceland. For perspective, the global human carbon footprint is approximately 40 gigatons (Gt) of CO₂ per year. This makes AI responsible for less than 0.3% of global emissions.

Comparing AI’s emissions to major contributors highlights its relative scale:

  • Concrete and cement production: ~2.9 Gt CO₂ annually, or 28 times AI’s footprint.
  • Agriculture and food systems: ~10 Gt CO₂ annually, nearly 100 times AI’s emissions.
  • Transportation: ~7.3 Gt CO₂ annually, over 70 times AI’s contribution.

These comparisons show that while AI’s emissions are significant, they are dwarfed by other sectors. Yet, its growth trajectory raises valid concerns about sustainability.


The Trade-Off: Costs vs. Benefits

Despite its environmental impact, AI bears enormous promise for addressing global challenges and improving civilization in meaningful ways.

1. Enhancing Productivity

AI optimizes supply chains, automates repetitive tasks, and enhances resource allocation, reducing inefficiencies across industries. For example, logistics systems powered by AI can cut transportation emissions while saving costs.

2. Advancing Health Outcomes

AI-driven tools enable breakthroughs in personalized medicine, disease prediction, and drug discovery. These innovations not only save lives but also reduce resource-intensive trial-and-error approaches in healthcare.

3. Ensuring Safety

AI systems are instrumental in disaster management and infrastructure maintenance, predicting failures and optimizing emergency responses, thus minimizing risks to human life and property.

4. Tackling Climate Change

AI accelerates progress in renewable energy technologies, climate modeling, and precision agriculture. These contributions indirectly reduce emissions in high-impact sectors like energy and agriculture.


Charting a Sustainable Path for AI

While AI offers immense benefits, its energy-intensive growth cannot continue unchecked. Zhang’s study proposes actionable steps to balance AI’s development with sustainability:

Adopt Renewable Energy

One promising development in this area is Microsoft’s recent commitment to a 20-year power purchase agreement (PPA) with a revived Three Mile Island nuclear power plant in Pennsylvania. Scheduled to reopen in 2028, the 837 MW facility will power Microsoft’s AI data centers across Pennsylvania, Chicago, Virginia, and Ohio. This deal highlights how tech companies are beginning to align their growth with cleaner energy sources, reducing AI’s dependence on fossil fuels.

Improve Energy Efficiency

Advances in hardware and software can make AI systems more efficient. Smaller, task-specific models could replace the current trend of ever-larger, energy-intensive systems.

Enforce Transparency and Accountability

Mandatory reporting of energy use and emissions would push companies to adopt greener practices. Standardized metrics would help policymakers and researchers accurately assess AI’s impact.

Leverage Economic Incentives

Carbon taxes, projected to cost AI companies ~$10 billion annually, could incentivize innovation in energy-efficient AI and broader adoption of sustainable technologies.


AI: A Double-Edged Sword

AI’s dual nature as both a contributor to and a solution for global challenges cannot be ignored. Its carbon footprint is a necessary trade-off for advancements in productivity, health, and safety, as well as for tackling climate change itself. However, the industry must commit to sustainable practices to ensure that this trade-off remains justifiable.

Microsoft’s nuclear PPA underscores the potential for transformative partnerships between AI developers and clean energy providers, setting a precedent for how technology can grow sustainably.

By focusing on renewables, efficiency, and accountability, AI can lead the way toward a carbon-neutral future—one where its transformative power not only drives progress but also aligns with global sustainability goals.


Conclusion

AI’s environmental impact, though notable, is a fraction of the global total. Its immense societal benefits make it a necessary investment in a better future. However, achieving a sustainable balance requires a coordinated effort from industries, policymakers, and researchers.

If guided responsibly, AI could become more than a tool for innovation—it could be a key player in building a sustainable, thriving civilization.

Staff Writer
+ posts

Leave a Comment