This journal article creates a model for predicting the climate impact of data centers using certain variables and uses sensitivity analysis to show which variables are the most important factors in increasing impact. The largest factors include the energy efficiency of the infrastructure in the data center as well as the emissions from electricity sources. Additionally, they provide potential solutions for what climate neutrality can look like with data centers, which include on-site renewable energy production.

The article argues that global data center capacity is in the midst of an infrastructure investment supercycle driven by AI and cloud demand with roughly 100 GW of new capacity expected come online by 2030 and deployment patterns increasingly shaped by power availability and regional market conditions. The report uses industry projection of a 14% compound annual growth rate through 2030, construction cost dat averaging $10.7 million per MW, and regional breakdowns which shows the Americas commanding about 50% of global capacity while APAC and EMEA grow at slower rates. This resource is important because it provides current, data rich evidence of how unevenly data center expansion is distributed across the globe. For our project, this report reinforces the thesis that data center growth concentrates in already privileged region while the shift from centralized AI training to distributed inference workloads may eventually reshape geographic strategies but has not yet meaningfully closed the gap for underserved areas.

The paper discusses the environmental colonialism that occurs and creates an imbalance in terms of data centers present in various countries. An example of this is shown in rural Ireland, where a person may have a hard time finding sources of power to go about their daily activities. However, big tech companies could have access to nearby sources such as windmills, thus cutting off access to the local population, and bringing data politics into play. This relates to our project because it has to do with our research question regarding who bears the environmental and social burdens of data center growth.

This article covers how OpenAI’s chief executive Sam Altman admitted publicly that AI’s effects will consume a lot of power, and that its effects on energy systems will be large. However, Altman has also begun investing in nuclear fusion, which is expected to become the next big energy source. In the next few years, AI will begin needing as much power as entire nations. This article relates to our project because it shows the negative effects of energy consumption, and further highlights how renewable energy is needed. 

Rather than treating AI as a more virtual process, this source examines the environmental effects of generative AI by taking into account the electricity and freshwater consumption required to run data centers. It argues that even one use of ChatGPT and similar systems leaves a trace of about 10 milliliters of water and 4.3 grams of CO2, which is less than driving or taking a shower but more than low-energy digital activities like basic web browsing. This argument is supported by a direct comparison of AI’s water consumption and emissions per use with everyday activities and other online practices (such as social media, video calls, and streaming), placing AI in that range.The source continues to scale, showing how, when turned on on massive platforms, AI-guided demand could transform into billions of gallons of water and tens of terawatt-hours of electricity. It is relevant to our project because it highlights how difficult it is to quantify these effects given the lack of transparency provided by businesses and gives us tangible numbers on a per-query and at-scale basis that we can see right away. Although many of the headline figures are still based on estimates and secondary reporting, it does present the issue as one that can be resolved by referring to mitigation paths, such as efficiency improvements, renewable energy, and improved disclosure.

The International Energy Agency (IEA) is a global nonprofit organization aimed at creating reports about energy consumption and production across the globe and supporting the clean energy transition. They also provide analysis, data, and policy recommendations. This report specifically provides predictions about energy demand from data centers beyond 2034, split by country and electricity source.

This article argues that integrating data centers with renewable energy source plants can stabilize the electricity grid, reduce the carbon footprint of data centers, and promote investment in renewable energy source plants. The article came to this conclusion after testing using real-world data in two scenarios where data centers are used: AWS computing for training Machine Learning algorithms and Bitcoin mining. This conclusion is important because it brings to light a benefit of data centers that is not mentioned in the general narrative of data centers being harmful to the environment. In our thesis, this article will help us conclude that data centers can be beneficial to the local population and to the environment, if the right steps are taken.

This article argues that cloud infrastructures, despite sounding ethereal, should be understood as data centers that profoundly impact the environment around them, causing a burden on local water resources and infrastructure. The article is based on a case study of Google’s South Carolina data center and uses 3 years of local newspaper articles as evidence of the data center’s impact on the community. This case study is important because it gives a concrete example of the impact data centers can have on a local community. In our thesis, this article will help us tell a narrative about the negative impact of data centers on the local people and environment.

This article references a report published by the University of Michigan School of Public Policy, where it discusses the impacts of data centers on resources like water and electricity and policy actions local and larger governments can take to reduce the impacts. It points out that data centers cannot run on renewable energy alone, since the uptime requirement is extremely rigid. The report provides examples of harm caused by data centers in Michigan, Nebraska, Virginia, Utah, Georgia, Indiana, and Washington state. The report recommends that the US follow the German Energy Efficiency Act model.

In addition to examining the digital images that end users view, this article also looks into the ecological effects of generative AI by examining the carbon footprint and the amount of freshwater used to run data centers. He contextualizes the idea of generative AI with the broader growth of the world’s data infrastructure, showing how even small per-query footprints can have significant effects on millions of users. Even though a single AI-human interaction results in very little water use and a small amount of CO2 emissions, the total amount of data has a big influence on Earth’s overall energy and water consumption. To put the impact of AI in perspective, the article uses industry reports and earlier research to contrast its footprint with other everyday activities like driving, streaming, and household energy use. It also emphasizes how efficiency advancements, the use of renewable energy, and cost disclosures can all help to reduce these environmental expenses. This article presents generative AI as a technology that has a noticeable impact on the environment but that can be resolved, making it a problem that can be resolved by carefully planning and powering data centers.

This article argues that data centers should be understood as energy prosumers requiring global coordination of renewable energy integration and waste heat utilization. To demonstrate this claim, the article reviews cooling technologies, renewable integration methods, real world projects, and economic analyses showing 7-46 payback periods. This article is important because it provides analysis of data centers’ role in the energy system with real world implementation examples. In our thesis, it will show how location decisions intersect with regional energy availability and climate conditions. The Nordic country examples show how geographic and political factors privilege certain locations over others.

This collection of journal articles discusses the cloud computing infrastructure monopoly of technology industry empires like Google, Amazon, and Microsoft, and it argues that governments have been unable to redistribute such infrastructure to account for more intensive computing. As evidence, the authors explain the history and exact geographic spread of data centers in the world by cloud computing providers like Amazon Web Services, Microsoft Azure, and Google Cloud, as well as their different levels of supported GPUs. This resource is important because it asserts how the United States has dominated public cloud computing over other nations, which also answers one of our research questions about which countries have more concentrations of data centers. As a result, this resource supports our thesis because it touches upon geopolitical power competition due to cloud computing, as well as the stark divide between countries that have access to computing resources and those that do not.

This journal article about data center infrastructure in Sweden discusses energy gentrification, which manifests from electricity grids that reach their maximum capacities and forces disproportionate allocation of societal resources. To be specific, energy gentrification from such a capacity deficit causes competition between data centers in Sweden and housing residents for land resources, thus displacing socioeconomically weaker groups in favor of an environment with higher economic status. As evidence, this resource studies data centers in Sweden built by companies like Microsoft, their energy consumption, and current policies targeting energy gentrification. This research is important because it analyzes in-depth how electricity grid deficits in Sweden exploit local communities and how it calls for stronger legislation. As for how it benefits our thesis, this resource could offer a socioeconomic perspective into our research about data centers, especially how centers actively displace communities.

The article argues that the expansion of cloud edge data centers while reducing latency of users already near existing infrastructure widens the cloud digital divide by disproportionately benefiting wealthier countries and leaving the underserved population further behind. The authors use latency measurements from over 4,800 globally distributed RIPE Atlas probes, US census income data, and remotely sensed nighttime light imagery as a proxy for wealth to quantify inequality and unfairness in data center placement across six continents. This resource is important because it moves beyond the traditional digital divide framework of internet access or speed and instead introduces cloud latency as a new dimension of digital inequality. For our project, this article directly supports the thesis that where data centers are build reflects and reinforces existing socioeconomic hierarchies which provides concrete metrics like the concentration index and inequality ratio that show how edge deployments disproportionately benefit wheatley urban areas rather than underserved regions.

This article examines the environmental implications of artificial intelligence, arguing that its development has led to disparities in environmental harm across different regions, such as through water footprint. The authors assert that regional environmental costs should be redistributed to achieve a more proportionate resource allocation between countries. As evidence, this resource quantifies the environmental costs and proposes how specific countries are more susceptible to harm; for instance, the usage of carbon-free energy differs dramatically between Google’s data centers. This resource is important because it provides a well-rounded explanation and potential solution to the unequal distribution of AI usage across data centers. As a result, this resource is beneficial for our thesis because it demonstrates how data centers harm local communities by depleting them of carbon and water resources in favor of complex AI models.

This article determines the environmental footprint that data centers leave on their local populations. Specifically, they calculated the carbon and water footprints of data centers within the United States. Their findings show the vast energy consumption that data centers do, and suggest that new data centers are found in more environmentally favorable areas in order to reduce the impact. This relates to our research because it discusses the environmental impact of data centers, and offers renewable solutions to this problem. 

Rather than treating AI as a more virtual process, this source examines the environmental effects of generative AI by taking into account the electricity and freshwater consumption required to run data centers. It argues that even one use of ChatGPT and similar systems leaves a trace of about 10 milliliters of water and 4.3 grams of CO2, which is less than driving or taking a shower but more than low-energy digital activities like basic web browsing. This argument is supported by a direct comparison of AI’s water consumption and emissions per use with everyday activities and other online practices (such as social media, video calls, and streaming), placing AI in that range. The source continues to scale, showing how, when turned on on massive platforms, AI-guided demand could transform into billions of gallons of water and tens of terawatt-hours of electricity. It is relevant to our project because it highlights how difficult it is to quantify these effects given the lack of transparency provided by businesses and gives us tangible numbers on a per-query and at-scale basis that we can see right away. Although many of the headline figures are still based on estimates and secondary reporting, it does present the issue as one that can be resolved by referring to mitigation paths, such as efficiency improvements, renewable energy, and improved disclosure.

This article argues that LLMs pose critical sustainability challenges through their energy consumption, carbon emission, and resource utilization that require sustainable practices. The article uses training cost data, environmental impact studies, and the Beta Lab solar powered case study as evidence for this claim. The article is important because it provides comprehensive cost breakdowns and environmental mechanisms that quantify AI infrastructure impact. In our thesis, it will provide data on energy consumption, carbon emission, water usage for AI data centers. This supports our environmental analysis and economic examination of data center cost.

The article argues that surging demand for AI ready data centers is outpacing supply and that the speed at which companies and investors expand capacity across the value chain will determine how quickly AI is deployed at scale. The authors draw on McKinsey’s proprietary modeling of AI adoption trend, chip power consumption and workload projection to estimate that global data center demand could more than triple from 60 GW to up to 298 GW by 2030 alongside industry data on colocation pricing, vacancy rates, and capital spending. This resource is important because it shows how AI is accelerating the physical expansion of data centers which raises urgent quotations about energy access, geographic concentration, and who benefits from this infrastructure. For our project, this article strengthen the argument that data center growth is not simply a neutral technological trend but a process shaped by power availability and economic incentives which pushes new facilities toward regions with abundant energy while potentially deepening digital inequality for areas that leak the infrastructure to attract investment.

This article argues that data centers consume massive amounts of electricity and harm the environment through carbon emissions, extremely high water consumption, and electronic waste. The article uses statistics from various sources as evidence for these claims, such as data centers contributing to 2.5% to 3.7% of global greenhouse emissions. This article is important because it quantifies the impacts of data centers through data, allowing for a more in-depth analysis of how data centers harm the environment. In our thesis, this article will help us demonstrate the global impact of data centers and provide data to back up our claims.