Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

---

The rapid growth of Artificial Intelligence (AI), especially Large Language Models (LLMs), fuels digital transformation but raises computational demand, making data centers major energy users and emission sources. The Information and Communication Technology (ICT) sector contributes 2–4% of global emissions. Assessing AI’s carbon footprint is vital for sustainability and policy planning. This narrative review systematically searched Scopus, Web of Science, PubMed, ScienceDirect, and Google Scholar from January 2019 to October 2025. Keywords related to AI, data centers, carbon, greenhouse gas emissions, and green AI were combined using Boolean operators. Included studies covered original research, reviews, and technical reports on measuring or mitigating AI’s carbon footprint. Studies focused only on AI’s environmental applications or hardware design were excluded. Data were qualitatively categorized and analyzed. AI’s carbon footprint arises from the full model lifecycle—including embodied carbon, training, inference, and end-of-life—along with growing computational demand, hardware efficiency, and geographic carbon intensity variations. Currently, 369 generative models emit 10–18 million tons of CO₂ annually, projected to reach 245 million tons by 2035. Efficient architectures like Mixture-of-Experts (MoE) can reduce energy use tenfold; Tensor Processing Units (TPUs) are about 50% more efficient than GPUs; and data centers with a Power Usage Effectiveness (PUE) of 1.1–1.4 outperform those above 1.6. Geographic location can cause 5- to 10-fold differences in carbon intensity. Green AI techniques—such as knowledge distillation, quantization, data optimization, renewable energy, and tools like Code Carbon—can cut emissions by up to three orders of magnitude. AI’s growing carbon footprint challenges the shift to a low-carbon economy. Mitigation requires Green AI, transparency, standardized metrics, and efficient data centers. Sustainable AI depends on collaboration among researchers, industry, and policymakers, with sustainability as a key principle.

Keywords: Artificial intelligence, Carbon footprint, Data centers, Greenhouse gas emissions, Green AI.

Full-Text [PDF 1198 kb]   (9 Downloads)    

Add your comments about this article : Your username or Email:

---

---