AI’s Environmental Footprint in Navigating Challenges and Digital Diplomacy | Insights from Vincent Aldy Hermawan
THINKING GREEN IN THE AGE OF THINKING MACHINESENVIRONMENTAL POLICY IN FOCUSINTERVIEWS AND INSIGHTS
Lakeisha Drusillia Wijaya
1/23/20263 min read
AI’s Environmental Footprint in Navigating Challenges and Digital Diplomacy
Artificial intelligence (AI) now permeates nearly every aspect of life. However, AI's rapid growth also carries a significant environmental footprint in the form of energy, water, and carbon emissions. As Fengqi You of Cornell University noted, "AI is transforming every sector of society, but its rapid growth has a significant footprint in energy, water, and carbon consumption."
By 2030 alone, the rapid growth of AI infrastructure is projected to produce 24-44 million tons of CO₂ per year, equivalent to adding 5-10 million cars to the road, and consume 731-1,125 million cubic meters of water per year; enough to meet the needs of 6-10 million in American households.The data shows that the AI revolution is not only a matter of efficiency and innovation, but also a serious environmental issue.
Drastically Increased Energy Consumption
Reflecting this scenario, we can understand that AI creates problems such as energy-intensive training models. MIT News notes that training models with billions of parameters requires a significant amount of electricity, increasing carbon emissions and straining the power grid.
This issue resonates with the interview conducted with Mr. Vincent Aldy Hermawan, an undergraduate from Binus ASO University, explaining that AI model training is the most energy-intensive part of the process compared to running the model (inference) or simply maintaining the data center. Training requires thousands or millions of computationally intensive iterations to achieve high accuracy, resulting in a much larger energy footprint.
Water Consumption and Challenges in Tropical Regions
Besides energy, water consumption for server cooling is also a major concern. AI servers require cooling systems that consume millions of liters of water per day to maintain safe operating temperatures. MIT notes that cooling generative AI hardware has the potential to deplete municipal water supplies and damage local ecosystems.
A Cornell study estimates that water consumption for AI in the US alone could reach over one billion cubic meters per year. In tropical regions like ASEAN, this pressure is even greater because high ambient temperatures increase cooling requirements.
The ASEAN Climate Change (ACCEPT) report estimates that data center electricity demand in Southeast Asia will increase from 9 TWh in 2024 to 68 TWh in 2030. If electricity supplies remain reliant on fossil fuels, emissions could increase up to 7.6 times the initial estimate, adding to an already significant climate burden.
Challenges in Public Oversight
One of the biggest issues that is often overlooked is the lack of transparency from technology companies regarding the energy consumption, carbon emissions, and water usage generated by their AI systems. Polytechnique Insights noted that Google only reports total data center electricity consumption (24 TWh in 2023) without separating out specific AI usage.
This was mentioned in an interview by Mr. Vincent as he felt companies were not being transparent enough, and even if they did disclose the data, the public might still be dissatisfied because AI's energy consumption is indeed substantial. Consequently, it makes it difficult for researchers, policymakers, and the public to understand the true impact of AI.
Towards a More Sustainable AI?
Sustainable AI development cannot begin without first understanding its environmental footprint. Both scientific findings and interviews indicate that measuring energy, water, and emissions impacts is the most fundamental step before undertaking any optimization.
Large-scale AI technologies require significant energy and resources, so without clear monitoring, their growth has the potential to exceed environmental limits. Due to the significant computational and infrastructure burden, optimization processes will be ineffective if organizations are unaware of their actual consumption.
Google and the ASEAN AI Governance Guidelines emphasize the importance of comprehensive power usage monitoring. Organizations can reduce AI's footprint by using smaller, more efficient models, optimizing model lifecycles, employing energy-efficient architectures, selecting renewable energy-powered data centers, and utilizing carbon tracking tools like CodeCarbon. Google even reports that hardware and software innovations can reduce the energy footprint per AI request by tens of times. However, researchers like Strubell caution that "green AI" does not necessarily mean zero emissions, as renewable energy remains limited and training large models can sacrifice energy allocated to other sectors. Interviewees also emphasized that purchasing renewable energy does not automatically mean zero emissions if it is not accompanied by an increase in clean energy capacity.
The need to measure and control the impact of AI applies not only to companies but also to governments and society. The IEA encourages countries to strengthen electricity grids, increase clean energy generation, and improve data center efficiency. Interviewees echo this view, stating that regulation is necessary to maintain environmental sustainability. ASEAN has taken steps through the ASEAN AI Governance Guidelines (2024) and the construction of the YTL Green Data Center in Malaysia, which runs entirely on green energy. These efforts demonstrate how digital diplomacy, energy policy, and technological innovation must go hand in hand.
The journey towards more sustainable AI requires strict monitoring, efficient technology, and environmentally sound public policies. Scientific data, interviews, and regional initiatives demonstrate one important point: AI must not develop at the expense of ecological boundaries. By accurately measuring its impact, optimizing technology, and establishing appropriate regulations, AI can be part of a future that is not only intelligent but also responsible for the planet.
news.cornell.edu/stories/2025/11/roadmap-shows-environmental-impact-ai-data-center-boom
https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117
https://www.polytechnique-insights.com/en/columns/energy/generative-ai-energy-consumption-soars/
https://en.reset.org/looking-at-the-entire-life-cycle-tips-for-sustainable-ai-development-and-use/