Sustentabilidade ambiental na implantação de data centers de IA: análise dos impactos energéticos e hídricos na região do Triângulo Mineiro

Abstract

The advancement of global digitalization has driven the expansion of AI data centers, which are essential infrastructures for data storage, processing, and transmission. However, these facilities require substantial amounts of electricity and water resources while also contributing to greenhouse gas emissions, raising concerns about their environmental sustainability. In this context, this dissertation aims to analyze the water and energy feasibility associated with the implementation of an AI data center in the municipality of Uberlândia, Minas Gerais, Brazil, considering the regional supply and demand conditions of these resources in the Triângulo Mineiro region. The research is characterized as a descriptive study with a qualitative approach supported by secondary data, employing an integrative literature review, documentary analysis, and a case study as methodological strategies. Initially, international cases of AI data centers are examined, focusing on their environmental impacts and mitigation strategies. Subsequently, an assessment of the energy and water conditions of the study area is conducted, followed by an estimation of the facility's resource consumption based on internationally recognized metrics, namely Power Usage Effectiveness (PUE) and Water Usage Effectiveness (WUE). The results indicate that, in aggregate terms, the Triângulo Mineiro region has a water supply compatible with the project's initial operational capacity, although localized and seasonal water scarcity remains, particularly during the dry season of the Cerrado biome. Regarding energy availability, the findings reveal greater concern due to increasing competition for electricity resources, suggesting that the feasibility of implementing the AI data center in the region is highly conditional rather than unrestricted. As the Technical and Technological Product (TTP) of this research, the METAS-DC – Method for Assessing Data Center Sustainability was developed, a replicable instrument structured around environmental, energy, water, social, and economic dimensions, designed to support the sustainability assessment of data centers and to inform decision-making by public and private managers. It is expected that the findings will contribute to the ongoing debate on the sustainability of digital infrastructure expansion by providing evidence to support the implementation of AI data centers in a manner that is more balanced and aligned with the regional conditions in which they are established.