AI’s Hidden Thirst: A Nuclear Solution for the Environmental Cost of Computing
Presentation Type
Poster Presentation
In Person or Zoom Presentation
In-Person
Campus
Daytona Beach
Status
Student
Student Year and Major
First Year, Aerospace Engineering
Presentation Description/Abstract
The expansion of artificial intelligence (AI) has intensified global energy consumption and revealed substantial water requirements associated with data center operations. The deployment of large-scale AI models requires extensive electrical power and cooling infrastructure, resulting in significant carbon emissions and freshwater consumption when supported by fossil fuel–based energy grids. Addressing these challenges requires sustainable, low-carbon, and water-efficient energy sources that can support continuous high-density computing.
The proposed research examines nuclear power as a mitigation strategy for the environmental impacts of AI by analyzing its potential to supply stable, carbon-free electricity with reduced water use. Meta’s $750 million data center in Newton County, Georgia, reportedly uses about 500,000 gallons of water daily, accounting for roughly 10% of the county’s supply, contributing to well depletion and reduced water pressure. In contrast, Microsoft has entered a 20-year power-purchase agreement to supply its data-center operations using electricity from the planned restart of the Three Mile Island Unit 1 nuclear generating station. This is expected to deliver 835MW of carbon-free baseload power by 2028. AI’s environmental footprint may be reduced by providing stable, carbon-free electricity with reduced lifecycle water intensity.
Lifecycle analyses demonstrate that nuclear generation exhibits lower operational water withdrawal and consumption rates than most thermoelectric alternatives. However, few studies have quantified or compared the lifecycle water intensity and carbon performance of AI data centers powered by nuclear versus conventional sources. Future research should focus on site-specific water accounting, techno-economic modeling of SMR–data-center co-location, and regulatory frameworks to guide sustainable AI energy strategies.
Keywords
Artificial Intelligence (AI), Data Centers, Energy Consumption, Water Footprint, Nuclear Power, Sustainability, Carbon Emissions, Small Modular Reactors (SMRs), Lifecycle Analysis, Sustainable Computing Infrastructure
AI’s Hidden Thirst: A Nuclear Solution for the Environmental Cost of Computing
The expansion of artificial intelligence (AI) has intensified global energy consumption and revealed substantial water requirements associated with data center operations. The deployment of large-scale AI models requires extensive electrical power and cooling infrastructure, resulting in significant carbon emissions and freshwater consumption when supported by fossil fuel–based energy grids. Addressing these challenges requires sustainable, low-carbon, and water-efficient energy sources that can support continuous high-density computing.
The proposed research examines nuclear power as a mitigation strategy for the environmental impacts of AI by analyzing its potential to supply stable, carbon-free electricity with reduced water use. Meta’s $750 million data center in Newton County, Georgia, reportedly uses about 500,000 gallons of water daily, accounting for roughly 10% of the county’s supply, contributing to well depletion and reduced water pressure. In contrast, Microsoft has entered a 20-year power-purchase agreement to supply its data-center operations using electricity from the planned restart of the Three Mile Island Unit 1 nuclear generating station. This is expected to deliver 835MW of carbon-free baseload power by 2028. AI’s environmental footprint may be reduced by providing stable, carbon-free electricity with reduced lifecycle water intensity.
Lifecycle analyses demonstrate that nuclear generation exhibits lower operational water withdrawal and consumption rates than most thermoelectric alternatives. However, few studies have quantified or compared the lifecycle water intensity and carbon performance of AI data centers powered by nuclear versus conventional sources. Future research should focus on site-specific water accounting, techno-economic modeling of SMR–data-center co-location, and regulatory frameworks to guide sustainable AI energy strategies.