Episode Summary
Executive Summary: The conversation examines the explosive growth of AI data centers and their rising costs in electricity, water, and public backlash. Casey Crownhart explains that while individual AI prompts use relatively little energy, the cumulative demand from billions of queries and massive infrastructure buildouts is driving major increases in power and water consumption, straining local grids, pushing up utility concerns, and complicating climate goals.
Main Topics: Scale of the AI data center boom (Priority: 5/5): The discussion frames the current wave of data center construction as far larger than earlier resource-intensive booms like crypto mining, with spending and electricity demand reaching eye-popping levels globally. Electricity use from AI queries and infrastructure (Priority: 5/5): Crownhart distinguishes between the small energy cost of one query and the much larger systemwide impact of billions of queries plus the electricity needed to build and run data centers. Water consumption and cooling methods (Priority: 5/5): The segment explains how data centers and the power plants that support them consume large amounts of water, especially in dry states, and reviews cooling alternatives that may reduce water use but create other trade-offs. Geographic concentration in water-stressed regions (Priority: 4/5): Companies are building in places like Arizona, Nevada, and Texas because of cheap power and available land, even though these regions are already under water stress. Political and community backlash (Priority: 4/5): Rising electricity rates and local concerns are prompting election messaging, project delays, and cancellations as communities resist new data center developments. Climate pledges versus AI demand growth (Priority: 4/5): The piece questions whether tech companies can keep net-zero promises while AI growth rapidly increases energy demand, making clean-energy procurement harder to keep pace with. Infrastructure timing mismatch (Priority: 5/5): The conversation highlights that energy systems, renewables, and nuclear projects take years to build, while data centers can be deployed much faster, creating a structural mismatch.
Key Arguments: AI’s resource footprint is not just about one chatbot query; it is the cumulative effect of massive, continuous usage across billions of interactions. Data centers already account for a meaningful share of global electricity use, and that share is projected to rise quickly by 2030. A large portion of AI-related water use is indirect, coming from the power plants that supply data centers, not just the facilities themselves. Dry states are attractive to builders because companies prioritize cheap power and land, but this intensifies local water stress. Cooling alternatives like direct liquid cooling and immersion cooling can save water, but they may cost more or use more electricity. Public pressure is becoming politically significant, affecting elections, utility-rate debates, and the cancellation or delay of major projects. Tech companies’ climate commitments are being tested by unprecedented electricity demand that clean-energy procurement alone may not offset fast enough. The real issue is systemic infrastructure, not merely personal choice by consumers using AI tools.
Data Points: Global electricity share from data centers (2024): 1.5% - International Energy Agency estimate cited by Crownhart. Projected global electricity share from data centers by 2030: Double from 2024 levels - IEA projection for data center electricity consumption. Global investment in AI and data centers in 2025: $580 billion - Compared with spending on global oil supply. Global oil supply investment in 2025: $540 billion - Used as comparison to AI/data center investment. Google Gemini average query electricity use: 0.24 watt-hours - Google’s estimate of energy per query. OpenAI ChatGPT query electricity use: 0.34 watt-hours - OpenAI’s estimate of energy per query. New data centers in development since 2022 located in water-stressed areas: Two-thirds - Arizona, Nevada, and Texas were cited as key examples. Water use from AI linked to power plants: Over 60% - Estimate that much of AI’s water footprint is indirect power-generation water use. Data centers’ share of total U.S. water use: 0.3% - One report’s estimate of national water consumption. Projected rise in data centers’ water use by 2030: Double from 2023 levels - Forecast of continued growth in water demand. Energy penalty for some alternative cooling methods: Up to 10% more energy - Certain water-saving cooling systems may require more electricity than evaporative cooling. Water use by a single data center versus homes: More than an entire county’s homes in some cases - Illustrates local-scale impact despite modest national share. Projects delayed or canceled due to community pushback: $93 billion - Data Center Watch estimate from March to June.
Pivotal Quotes: "This is overall a systems conversation that we need to be having." — Casey Crownhart: Her closing argument that AI’s impacts are infrastructure-wide, not just about individual user choices. "Data centers accounted for about 1.5% of the world's electricity consumption in 2024. And that's set to double by 2030." — Casey Crownhart: She explains the scale and expected growth of energy demand. "We spent more on data centers than the oil supply." — Casey Crownhart: She underscores the magnitude of 2025 AI/data center investment compared with global oil development spending.
Implications: AI’s growth is becoming a public utility and climate issue, not just a tech trend. Expect tighter scrutiny of data center siting, rates, water access, and energy sourcing as communities, regulators, and companies confront the infrastructure trade-offs.