The Hidden Cost of AI Operations
As artificial intelligence becomes embedded in legal practice—from document review to contract analysis—few professionals consider the environmental footprint behind each query. New research from Vanderbilt University challenges industry claims that energy disclosure requirements would slow AI development, arguing instead that transparency would benefit consumers, competition, and the environment simultaneously.
At Lawtechnology.ai, we track developments that shape how legal professionals evaluate and adopt AI tools. This research raises important questions about the sustainability metrics that should inform enterprise AI procurement decisions.
The Scale of AI’s Energy Demand
The numbers are striking. In 2023, U.S. data centers consumed 176 terawatt hours of electricity—enough to power 16.3 million homes—resulting in 61 million metric tons of carbon dioxide emissions. AI currently accounts for approximately 15% of total data center demand, but researchers predict that share could triple by 2028.
Water consumption is equally significant. A small one-megawatt data center annually draws about 7 million gallons of water for direct cooling, plus approximately 10 million gallons indirectly to cool the power plants that supply its electricity.
As AI capabilities expand through multi-step reasoning models, chatbots, and image/video generation, energy consumption rises accordingly.
The Transparency Gap
Despite consumers routinely considering energy efficiency data for appliances and vehicles, accurate information on AI model electricity consumption remains elusive. The research, authored by Vanderbilt’s Michael P. Vandenbergh, Ethan Thorpe, and Jonathan Gilligan, identifies why this gap persists.
Energy consumption for individual AI tasks depends on a complex web of proprietary variables: the program code, executing chips, data center infrastructure, and model size and complexity. Rapid technological advances can render energy research obsolete before publication. Proprietary routing algorithms and specialized chips further complicate assessment.
The researchers tested four publicly available AI footprint calculators—tools designed to estimate energy consumption for specific tasks. Despite controlling input variables and standardizing output, the highest energy estimate was 58 times greater than the lowest for the same AI model. This variance renders existing consumer tools essentially useless for informed decision-making.
Challenging Industry Arguments
Some industry advocates argue that AI’s potential productivity gains and environmental problem-solving capabilities mean development should be prioritized over disclosure requirements. Others claim AI’s workplace efficiency creates net environmental benefits.
The Vanderbilt researchers find merit in the efficiency argument for certain tasks—AI may indeed use less energy than a human coder working on a desktop in an air-conditioned office for some operations. But they identify the fundamental problem: without actual energy consumption data, such comparisons remain speculative. The very information needed to validate industry efficiency claims requires the transparency industry resists.
The Case for Disclosure
The researchers contend that transparency “need not become a barrier to rapid development.” Instead, energy disclosures would:
Inform consumers: Legal departments and law firms making AI procurement decisions could factor energy efficiency into vendor selection, just as they consider other operational costs.
Drive competition: Transparent energy metrics would create market incentives for efficiency improvements, rewarding providers who achieve better performance per watt.
Improve efficiency: Public benchmarking would accelerate efficiency-focused innovation across the industry.
Enable accurate comparison: Organizations could make legitimate assessments of AI versus human task completion environmental costs.
Regulatory Pathways
Despite current deregulatory trends, the researchers argue that voluntary disclosures could establish transparency norms that benefit the AI industry. The precedent exists: Energy Star ratings transformed appliance markets without stifling innovation, creating competitive advantages for efficient manufacturers.
Implications for Legal Technology Adoption
For legal professionals evaluating AI tools, this research suggests adding sustainability metrics to procurement criteria—while acknowledging current data limitations. As law firms increasingly face ESG reporting requirements and client sustainability mandates, the environmental footprint of legal technology infrastructure will become harder to ignore.
The broader regulatory trajectory may eventually require such disclosures regardless of voluntary action. Organizations that develop energy efficiency evaluation frameworks now will be better positioned when—not if—transparency requirements emerge.
Access the complete analysis from The Regulatory Review for additional detail on the research findings and policy implications.