Session #25 - Voluntary Carbon Credit Markets & AI Energy Use Study Notes

Course Administration and Final Exam Details

Final Exam Schedule: The exam will take place on Monday, May 4 at $1\text{ pm}$ .
Format: The format will be identical to the midterm exam.
Content Focus: While the exam is technically cumulative, the primary focus will be on the material covered during the second half of the semester.
Review Session: A review session is scheduled for May 3 at $5\text{ pm}$ via WebEx.
Policy Brief Extension: The deadline for the policy brief has been extended to May 6 at midnight.
Assignment Status: Assignment #2 has been graded and returned.

Fundamentals of Voluntary Carbon Credit Markets

Definitions and Key Concepts

Carbon Credit: Defined as 1 credit equaling $\text{1 ton CO}_2\text{ eq}$ ( $\text{carbon dioxide equivalent}$ ) that is either avoided relative to a baseline or removed directly from the atmosphere and stored.
The "Promise": A carbon credit represents a financial commitment toward an action that reduces greenhouse gas (GHG) concentrations in the atmosphere.
Market Mechanism: Credits are traded and sold on established exchanges, providing an efficient framework for emission reduction. * Major Exchanges: Verra, Gold Standard, and the Clean Development Mechanism (CDM), which is a program run by the United Nations.

Market Participants

Buyers: Primarily emitters of greenhouse gases. Their motivations include: * Compliance: Meeting government-mandated regulations. * Sustainability Pledges: Achieving corporate-driven goals that exceed regulatory requirements.
Sellers: These are firms, non-profits, and various organizations capable of "promising" verifiable carbon reductions.

Types of Credits

Avoidance Credits: Focus on preventing emissions that would have otherwise occurred (reducing the flow of carbon to the atmosphere).
Removal Credits: Focus on extracting existing carbon from the atmosphere (reducing the stock of carbon).

Carbon Offsetting Techniques

Various strategies are used to create carbon credits by either sequestering carbon or reducing emissions at the source:

Increasing Carbon Sequestration: * Reforestation: Planting trees in areas that were previously forested. * Afforestation: Establishing forests in areas that have not recently been covered by trees. * Soil Carbon Projects: Utilizing regenerative agriculture practices to trap carbon in the soil. * Wetland Restoration: Often referred to as "blue carbon" offsets, focusing on coastal and aquatic ecosystems.
Increasing Energy Efficiency: * Example: Replacing traditional charcoal and wood-burning methods with bioethanol cookstoves.
Renewable Energy Offsets: Replacing fossil-fuel-based energy production with renewable sources.
Waste and Landfill Management: Capturing methane or reducing emissions from waste streams.

UN Sustainable Development Goals (SDGs) and Co-benefits

Carbon projects, particularly removal projects, often produce significant co-benefits that align with the United Nations' Sustainable Development Goals (SDGs).

The SDG Framework

Adoption: Adopted by all UN Member States in 2015.
Core Principle: Recognizes that ending poverty must occur alongside improvements in health and education, reduction of inequality, and the promotion of economic growth, all while tackling climate change and preserving ecosystems.
Purpose: Provides a shared blueprint for peace and prosperity for people and the planet, serving as an urgent call for action by all countries in global partnership.

The 17 Sustainable Development Goals:

No Poverty
Zero Hunger
Good Health and Well-being
Quality Education
Gender Equality
Clean Water and Sanitation
Affordable and Clean Energy
Decent Work and Economic Growth
Industry, Innovation, and Infrastructure
Reduced Inequalities
Sustainable Cities and Communities
Responsible Consumption and Production
Climate Action
Life Below Water
Life on Land
Peace, Justice, and Strong Institutions
Partnerships for the Goals

Challenges to Voluntary Carbon Credit Markets

Structural and Operational Issues

Potential for Double Counting: A situation where a country claims credit for emission reductions to meet its national pledges (NDCs) while simultaneously selling those same reductions on the private marketplace.
Time-span Issue: Many "high quality" projects only require sequestration or reduction for a period of $40\text{ years}$ . Critics argue this duration should be significantly longer to be effective for climate stability.
Improper Use: Credits should ideally be used for "hard-to-reach" carbon after a company or country has maximized their own direct reductions. Instead, they are often used to delay aggressive reduction measures, potentially undermining broader climate action.

Integrity and Measurement Concerns

Quality Control: Evidence suggests many projects over-estimate the actual carbon offset. This includes the "protection" of forests that were not actually at risk of being cut down.
Measurement Scandals: Difficulties in measuring sequestration led to a recent market collapse. The market value shrank from $\$1.9\text{ billion}$ in 2022 to $\$723\text{ million}$ in 2023.
Corporate Monopoly: The highest-quality projects are often monopolized by large corporations, leaving smaller firms with fewer reliable offset options.

Prospects for Market Recovery and Optimization

Despite recent setbacks, several factors suggest the voluntary market may recover and strengthen:

Corporate Commitments: The world’s largest companies (including those in aviation, heavy industry, and AI) have made large emission reduction promises that will require offsets to reach $100\%$ reduction goals.
Increasing Regulation: The World Bank has identified $75\text{ carbon-pricing schemes}$ globally, now covering $24\% \text{ of global emissions}$ , up from only $7\%$ a decade ago.
Regulatory Synergies: * Article 6 of the Paris Agreement: Allows countries to partially meet emission reduction pledges by purchasing credits. * California Cap and Trade: This government-mandated scheme allows participants to use a portion of credits sourced from the voluntary market.
Economic Impact: The market provides an efficient mechanism for wealth transfer from developed nations to the developing world, funding sustainable progress and SGDs.

Artificial Intelligence (AI) and Energy Consumption

Per-Prompt Energy Comparison

Simple Text Query: One query in a simple model uses approximately the same amount of energy as running a microwave for $8\text{ s}$ .
Video Generation: Creating a $5\text{ s}$ , low-resolution video uses as much energy as running a microwave for over $1\text{ hr}$ .

Cumulative Statistical Impact

National Consumption: AI accounts for $4-5\% \text{ of US electricity consumption}$ . The US dominates the global AI infrastructure with a $44\%$ share. Combined, the US and China account for $70\% \text{ of global capacity}$ .
Global Consumption: AI accounts for approximately $1.5\% \text{ of global electricity consumption}$ . Many countries are net importers of AI outputs, while the electrical load remains hosted in the US.
ChatGPT Usage: With $2.5 \times 10^9\text{ queries/day}$ , the emissions are equivalent to driving a regular gasoline car for $280,000,000\text{ miles}$ ( $2.8 \times 10^8\text{ miles}$ ) in a single year.
Growth Projections: By 2030, global energy demand specifically for AI is expected to be $2-4\text{ times}$ higher than predicted levels for 2025.

Factors Determining Energy and Climate Impact

Language Model Type: Large models require significantly more power than small ones.
Query Type: Complexity increases from Text to Image to Video.
Energy Source: The climate impact varies based on whether the data center is powered by coal/natural gas versus geothermal/nuclear (wind and solar are often insufficient due to the need for "firm," steady-state power).
Cooling Method: Significant energy or water is required to dissipate heat generated by processors.
Data Center Location: Proximity to clean energy and water sources is critical.

Data Center Cooling and Water Use

Cooling is essential for data center operations, using one of two primary methods:

Wet-Cooled Data Centers

Resource Use: These use water as the primary coolant.
Energy Efficiency: They are generally more energy-efficient, using $10-40\% \text{ less energy}$ than dry-cooled systems.
Water Intensity: They require $1-2\text{ liters}$ of water per $\text{kWh}$ of energy load.
Corporate Examples: * Google: Consumes $6-8 \times 10^9\text{ gallons}$ of water annually, which is comparable to the total water used for the entire California agricultural sector in a year. * Microsoft: Water usage increased by $34\%$ between 2021 and 2022.
Water Source: Approximately $70-80\% \text{ of the water used}$ is potable freshwater.

Dry-Cooled Data Centers

Resource Use: These use energy to power cooling fans and systems.
Energy Intensity: Cooling accounts for $10-30\% \text{ of the total energy load}$ .
Water Impact: These systems use almost zero water.

Solutions for AI Sustainability

Energy Demand Reduction

Strategic Siting: Locating data centers in regions with "firm" clean energy sources like geothermal or nuclear power.
Hardware Efficiency: Developing and using more efficient computer chips.
Model Optimization: Employing smaller, more specialized language models.

Water Use Reduction

Siting for Water Stress: Building data centers in areas with low water stress.
Thermal Circularity: Diverting warm water from cooling systems to heat homes or swimming pools.
Renewable Recycling: Using renewable energy to cool and then reuse existing water supplies.

Integrated Supply Mapping

According to research (Xiao et al. 2025 Nature Sustainability), location planning should consider water and energy supplies together. The goal is to identify locations that fall into the top percentiles for both low water footprint and low carbon emissions per unit of energy.