Environmental Impacts of Artificial Intelligence and Data Centers: A Sustainable Development Perspective

Introduction

Artificial Intelligence (AI) applications, commonly recognized through mobile apps and chatbots, rely heavily on data centers—large-scale facilities housing thousands of computers that process AI requests. These data centers vary in size and are operated by major technology companies such as Google, Microsoft, and Amazon. The emergence of AI-specific data centers has introduced significant environmental challenges, particularly concerning energy consumption, water use, and air pollution. This report emphasizes the relevance of these issues within the framework of the United Nations Sustainable Development Goals (SDGs), particularly SDG 6 (Clean Water and Sanitation), SDG 7 (Affordable and Clean Energy), SDG 11 (Sustainable Cities and Communities), SDG 13 (Climate Action), and SDG 3 (Good Health and Well-being).

Energy Consumption of AI Data Centers

AI Workloads and GPU Usage

AI workloads predominantly utilize Graphics Processing Units (GPUs), which offer high processing power but consume substantially more energy than traditional Central Processing Units (CPUs). The energy demand escalates with the complexity of AI outputs, with text generation requiring less energy compared to image or video generation.

Quantifying Energy Use

1. A single AI text query can consume approximately 114 joules, equivalent to running a microwave for 0.1 seconds.
2. Larger AI models may use up to 6,706 joules per response, comparable to running a microwave for 8 seconds or powering an e-bike for 400 feet.
3. Generating a standard-quality image requires about 2,282 joules, while a high-quality five-second video can consume over 3.4 million joules, equivalent to running a microwave for over an hour or riding an e-bike for 38 miles.

Scale of Energy Usage

• ChatGPT reportedly processes over one billion text requests and tens of millions of image requests daily.
• The total electricity consumption for these AI requests is estimated to be equivalent to the annual power usage of over 3,000 U.S. homes.
• Data centers accounted for 1.9% of U.S. electricity consumption in 2018, increasing to 4.4% in 2023, with projections reaching 6.7% to 12% by 2028.

SDG Relevance

These energy demands directly impact SDG 7 (Affordable and Clean Energy) and SDG 13 (Climate Action), highlighting the need for sustainable energy solutions and improved energy efficiency in AI infrastructure.

Water Usage in AI Data Centers

Cooling Requirements and Water Consumption

Data centers generate substantial heat during operation, necessitating cooling systems. Water-based cooling methods, favored for cost efficiency, involve evaporative cooling processes that consume large volumes of water.

Water Use Statistics

• In 2023, U.S. data centers consumed approximately 66 billion liters of water directly for cooling purposes.
• This volume is equivalent to the annual residential water use of over 500,000 people or irrigating approximately 11,101 acres of almond orchards.
• Water consumption by data centers has tripled over the past decade, rising from 21.2 billion liters in 2014.
• Indirect water use, such as that associated with electricity generation for data centers, is estimated at 800 billion liters in 2023.

Geographical and Environmental Concerns

Approximately half of data centers are located in regions experiencing water scarcity, including California, Arizona, and Texas. This situation exacerbates competition for limited water resources between data centers, local communities, and agriculture, posing challenges aligned with SDG 6 (Clean Water and Sanitation) and SDG 11 (Sustainable Cities and Communities).

Air Pollution and Public Health Impacts

Sources of Air Pollution

• Diesel backup generators used in data centers emit significant air pollutants during operation and maintenance.
• Electricity powering data centers often originates from fossil fuel power plants, releasing fine particulate matter (PM2.5), nitrogen dioxide (NO2), and sulfur dioxide (SO2).
• Manufacturing of AI hardware and construction materials contributes additional air pollution.

Health and Economic Consequences

A recent study estimates that the public health burden from U.S. data centers in 2030 could exceed $20 billion annually, primarily due to increased asthma and cardio-pulmonary diseases. These impacts disproportionately affect low-income communities, raising environmental justice concerns consistent with SDG 3 (Good Health and Well-being) and SDG 10 (Reduced Inequalities).

Policy Recommendations and Community Engagement

Essential Actions

1. Implement policies mandating community engagement during data center planning and development.
2. Enforce environmental protections to mitigate energy, water, and air pollution impacts.
3. Promote transparency and accessibility of environmental data from technology developers.
4. Ensure equitable distribution of benefits derived from AI technologies.

Educational Initiatives

Collaborative efforts with academic institutions have produced educational resources to raise awareness about AI’s environmental footprint. An example is the interactive website AI Energy and Water, which provides detailed information on AI’s implications for energy and water sustainability.

Conclusion

The environmental footprint of AI and its supporting data centers presents multifaceted challenges that intersect with several Sustainable Development Goals. Addressing these requires coordinated efforts involving technological innovation, policy intervention, and community participation to promote sustainable AI development that aligns with global sustainability objectives.

1. Sustainable Development Goals (SDGs) Addressed or Connected

1. SDG 6: Clean Water and Sanitation
   • Water consumption and water scarcity issues related to data centers’ cooling processes and their impact on local water resources.
2. SDG 7: Affordable and Clean Energy
   • High electricity consumption by AI data centers and the need for sustainable energy sources.
3. SDG 11: Sustainable Cities and Communities
   • Impact of data centers on local communities, especially low-income areas affected by pollution and increased energy costs.
4. SDG 12: Responsible Consumption and Production
   • Environmental costs of AI hardware manufacturing and data center operations, including resource use and pollution.
5. SDG 13: Climate Action
   • Carbon emissions and air pollution from energy use and manufacturing processes related to AI data centers.
6. SDG 3: Good Health and Well-being
   • Health impacts from air pollution caused by data centers and associated activities.
7. SDG 10: Reduced Inequalities
   • Environmental justice concerns where low-income communities bear disproportionate health and environmental burdens.

2. Specific Targets Under Those SDGs Identified

1. SDG 6: Clean Water and Sanitation
   • Target 6.4: By 2030, substantially increase water-use efficiency across all sectors and ensure sustainable withdrawals and supply of freshwater.
2. SDG 7: Affordable and Clean Energy
   • Target 7.2: Increase substantially the share of renewable energy in the global energy mix.
   • Target 7.3: Double the global rate of improvement in energy efficiency.
3. SDG 11: Sustainable Cities and Communities
   • Target 11.6: Reduce the adverse per capita environmental impact of cities, including air quality and waste management.
4. SDG 12: Responsible Consumption and Production
   • Target 12.2: Achieve sustainable management and efficient use of natural resources.
   • Target 12.4: Achieve environmentally sound management of chemicals and wastes.
5. SDG 13: Climate Action
   • Target 13.2: Integrate climate change measures into national policies, strategies, and planning.
6. SDG 3: Good Health and Well-being
   • Target 3.9: Reduce the number of deaths and illnesses from hazardous chemicals and air, water, and soil pollution and contamination.
7. SDG 10: Reduced Inequalities
   • Target 10.2: Empower and promote the social, economic and political inclusion of all, irrespective of income or other status.

3. Indicators Mentioned or Implied to Measure Progress

1. Energy Consumption Indicators
   • Electricity consumption by data centers as a percentage of total national electricity use (e.g., 4.4% in 2023, projected 6.7%-12% by 2028).
   • Energy use per AI request (joules per query for text, image, video generation).
2. Water Use Indicators
   • Direct water consumption by data centers (e.g., 66 billion liters in 2023).
   • Indirect water use associated with electricity generation for data centers (e.g., 800 billion liters in 2023).
   • Comparison of water use with local agricultural water needs and residential water consumption.
3. Air Pollution and Health Impact Indicators
   • Emissions of criteria air pollutants: PM2.5, NO2, SO2 from data centers and associated activities.
   • Public health costs and incidence of asthma and cardio-pulmonary diseases linked to air pollution from data centers.
   • Economic valuation of health impacts (e.g., $20 billion per year estimated public health burden in 2030).
4. Environmental Justice Indicators
   • Distribution of pollution and health impacts by income and community demographics.
   • Access to benefits and decision-making in AI technology deployment.

4. Table: SDGs, Targets and Indicators

Source: blog.ucs.org