Introduction
The expansion of generative artificial intelligence is reshaping the global technology landscape, but at what environmental cost? A new study published in Nature reveals that AI server deployment in the United States could generate an annual water footprint ranging from 731 to 1,125 million cubic meters and additional carbon emissions from 24 to 44 million tonnes of CO₂-equivalent between 2024 and 2030. These figures depend on the scale of expansion and raise urgent questions about the sustainability of a rapidly growing sector.
Research Context
The study analyzes the combined energy, water, and climate impact of operational AI servers in the United States between 2024 and 2030, a critical period characterized by future uncertainties but of great strategic importance. The United States was selected as the study region due to its dominant position in the global AI market. The research is based on temporal projection models and regional frameworks, assuming that server deployment mirrors current large-scale AI data center distribution patterns.
The data center sector, according to the International Energy Agency, already contributes 0.6% of global carbon emissions due to electricity consumption. With AI growth, this energy consumption could double by 2026, threatening decarbonization targets under the Paris Agreement, which include a 53% reduction in data center emissions by 2030 and net-zero goals for the AI sector.
Expansion Scenarios and Environmental Impact
The research presents five scenarios for cumulative AI server installations in the United States from 2024 to 2030: low demand, low power, mid-case, high application, and high demand. The mid-case serves as the base scenario, while low and high demand set projection bounds. The low power scenario assumes server efficiency gains, while high application considers increased adoption driven by efficiency.
Even in the most conservative scenarios, a considerable increase in energy, water, and carbon footprints of AI servers is outlined. The highest scenario yields the highest environmental impact, largely surpassing previous forecasts for the entire US data center market, underscoring the environmental risks of unchecked AI server expansion.
Geographic Distribution of Impacts
State-level AI server allocation shows significant variations in Power Usage Effectiveness (PUE) and Water Usage Effectiveness (WUE). Southern states such as Florida exhibit higher PUE and WUE than northern states such as Washington, reflecting climate impacts. Moreover, grid factors demonstrate notable sensitivity to location, highlighting the importance of the local grid for AI servers' environmental impacts.
Centralized installation in water-stressed regions could perturb local water balance and threaten supply for millions. The indirect water footprint contributes 71% of the total, with direct use at 29%.
Carbon Emissions and Energy Consumption
The climate impact of AI servers will stem primarily from their operations (Scopes 1 and 2) and supply-chain activities (Scope 3), including manufacturing and end-of-life treatment. Scope 2 emissions from indirect energy purchases are expected to constitute a substantial portion and indicate high dependency on increased AI server energy consumption.
AI server energy predominates over infrastructure energy. This consumption pattern raises concerns not only about energy provisioning challenges but also water scarcity and climate change issues stemming from the energy-water-climate nexus of AI data centers.
Prospects for Net-Zero Pathways by 2030
The research evaluates best- and worst-case scenarios of key drivers, including industry efficiency improvements, server location distribution, and grid decarbonization. Results show that the AI server industry is unlikely to meet its net-zero aspirations by 2030 without substantial reliance on highly uncertain carbon offset and water restoration mechanisms.
Although best practices may reduce emissions and water footprints by up to 73% and 86% respectively, their effectiveness is constrained by current energy infrastructure limitations. These findings underscore the urgency of accelerating the energy transition and point to the need for AI companies to harness the clean energy potential of Midwestern states.
Efficiency Improvements: PUE and WUE
Over the past decade, efficiency gains in the data center industry have stabilized environmental costs despite a doubling of computing instances. The research examines existing potential for further improvements through system optimization and technology adoption.
The best-practice scenario suggests notable reduction potential: over 7% PUE reduction and over 85% WUE reduction, despite the high efficiency of AI data centers compared with industry averages (PUE 1.58, WUE 1.8). PUE reduction yields over 7% reductions in total energy consumption and carbon emissions. WUE reduction efforts result in over 29% reduction in total water footprint.
Advanced Technologies: Liquid Cooling and Optimization
The study explores the potential impact of adopting advanced technologies within AI data centers, focusing on advanced liquid cooling (ALC) and server utilization optimization (SUO). Best ALC adoption can reduce about 1.7% of energy consumption, 2.4% of water footprint, and 1.6% of carbon emissions of AI servers by 2030.
For SUO, the best-case scenario, representing total adoption by 2030, results in a 5.5% reduction in all footprint values, while the worst-case scenario, representing frozen adoption, leads to a 7.3% increase by 2030. Maximum reductions of energy, water, and carbon drawn from existing potential are approximately 12%, 32%, and 11% respectively.
Policy and Industry Implications
Coordinating efforts of private actors and regulatory interventions would ensure the competitive and sustainable development of the AI sector. The research emphasizes that AI companies should harness the clean energy potential of Midwestern states, where favorable climate conditions and renewable energy resources could significantly reduce environmental impact.
Factors such as industry efficiency initiatives, grid decarbonization rates, and spatial distribution of server locations within the United States drive deep uncertainties in estimated water and carbon footprints. Without rapid and coordinated action, the sector risks compromising global climate goals even as it seeks to drive technological innovation.
Conclusion
The research published in Nature offers a comprehensive view of the environmental impact of AI servers in the United States, highlighting urgent challenges and mitigation opportunities. With projections indicating water consumption up to 1,125 million cubic meters annually and carbon emissions up to 44 million tonnes of CO₂-equivalent by 2030, the sector faces decisive choices.
Existing best practices can significantly reduce impact but require substantial investments in clean energy infrastructure and advanced cooling technologies. The path to net-zero targets by 2030 remains uncertain without strong commitment from both industry and regulators. The energy transition cannot be delayed if we want to ensure that AI, a powerful tool for human progress, does not become an accelerator of climate change.
FAQ
What is the projected water footprint of AI servers in the United States by 2030?
According to the Nature research, the annual water footprint of AI servers in the USA could range from 731 to 1,125 million cubic meters between 2024 and 2030, depending on the scale of expansion.
How much additional carbon emissions will American AI servers generate?
Projected additional annual carbon emissions range from 24 to 44 million tonnes of CO₂-equivalent between 2024 and 2030.
Can AI servers achieve net-zero targets by 2030?
The study indicates that the AI server industry is unlikely to meet net-zero aspirations by 2030 without substantial reliance on highly uncertain carbon offset and water restoration mechanisms.
What is the difference between direct and indirect water use in AI data centers?
Direct use refers to server cooling (29% of total), while indirect use relates to water needed to produce the electricity consumed by data centers (71% of total).
How much can industry best practices reduce environmental impact?
Best practices may reduce emissions by up to 73% and water footprints by up to 86%, though their effectiveness is constrained by current energy infrastructure limitations.
What are PUE and WUE in AI data centers?
PUE (Power Usage Effectiveness) measures data center energy efficiency, while WUE (Water Usage Effectiveness) measures water use efficiency. Lower values indicate greater efficiency.
Which American states have the greatest environmental impact for AI servers?
Southern states such as Florida show higher PUE and WUE than northern states such as Washington, due to climate differences affecting cooling systems.
What is the data center sector's contribution to global carbon emissions?
According to the International Energy Agency, data centers and data transmission networks currently contribute 0.6% of total global carbon emissions due to their electricity consumption.