The rapid global adoption of advanced artificial intelligence (AI) and machine learning (ML) systems is catalyzing a new wave of digital infrastructure expansion. While much attention has focused on the capabilities and societal impacts of AI, comparatively less scrutiny has been given to the material infrastructure—specifically, high-performance data centers—that enables its growth. This article investigates whether increased AI/ML deployment is driving significant increases in energy and water consumption worldwide. Findings indicate that AI-specific workloads now constitute a rapidly growing share of total data center operations, with training and inference of large models contributing disproportionately to electricity demand and cooling requirements. In tandem, water usage for data center cooling—often overlooked—has grown substantially, especially in regions already facing water stress. The paper concludes with implications for theory, engineering practice, and policy, advocating for mandatory resource transparency, efficiency regulation, and lifecycle accountability in AI development. It also identifies research gaps in regional modeling, AI workload differentiation, and cross-sector environmental assessments.

The continuous increase in computational power of GPUs, essential for advancements in areas like artificial intelligence and data processing, is driving the adoption of liquid cooling in data centers. Skived copper cold plates featuring parallel straight channels are a mature technology, but they lack design freedom due to manufacturing limitations. As chips become increasingly complex in their design with the transition towards heterogeneous integration, these parallel straight channels are not able to address critical areas of concentrated high heat flux (hotspots) on a chip. A single hotspot exceeding the upper temperature limit can cause the full chip to throttle and hence limit performance. In addition, this would require a reduction in coolant inlet temperature in the data center, causing an increase in electricity and water consumption. Ideally, areas of the cold plate in contact with hotspots of the chip need smaller channels to increase convective heat transfer, whereas areas with low heat flux may benefit from larger channels to compensate for the increased pressure drop. However, manual optimization of such a cooling design is challenging due to the nonlinearity of the problem. In this paper, we explore the usage of topology optimization as a method to tailor microfluidic cooling design to the power distribution of a chip to address the hotspot temperatures in high-power chips, using a platform called Glacierware. We compare the hotspot-aware, topology-optimized microfluidic design to straight channels of various widths to benchmark its performance. Evaluations of this optimized design show a 13% lower temperature rise or a 55% lower pressure than the best-performing straight channels, indicating highly competitive performance in industrial settings where both pressure and flow rate are constrained.

Generative Artificial Intelligence (GenAI) is being rapidly adopted by academic and policy leaders, often with a focus on its revolutionary benefits while overlooking its significant, undisclosed environmental costs. This paper deconstructs the physical footprint of GenAI, moving beyond the abstract cloud to analyze its resource-intensive infrastructure. It synthesizes current research into four key areas of concern. First, it reframes the energy debate, demonstrating that the operational inference (use) phase is the dominant long-term cost, with one study estimating its carbon footprint to be 25 times higher than the one-time training cost. Second, it reveals the hidden water footprint, which includes both direct Scope 1 consumption for cooling and the larger Scope 2 consumption from offsite electricity generation, a critical issue as two-thirds of new US data centers are in water-stressed areas. Third, it examines the systemic and material impacts, including the full Life Cycle Assessment (LCA) of hardware (e-waste, raw material extraction) and the rebound effect (Jevons’ Paradox), where efficiency gains increase overall demand. A new side effect includes AI tools propagating non-green code. Finally, the paper highlights the critical lack of corporate transparency, a black box of proprietary data that prevents effective governance. This paper argues that a full-cost accounting is

The rapid expansion of artificial intelligence (AI) is accelerating data center construction and creating downstream implications for households. This study examines how AI-era data center costs (comprising construction, energy, water, grid upgrades, cooling, and lifecycle management) move through service supply chains and affect household prices in healthcare, transportation, education, banking, and commerce. It also considers the productivity and welfare benefits that AI may transmit. This study identifies four pass-through channels: utility-rate socialization of energy costs, cloud-platform pricing, sectoral pass-through from AI-adopting industries, and indirect effects through supply chains and labor markets. It introduces the AI-inflated net good basket, defined as transmitted cost minus transmitted benefit, to show how AI reshapes the overall net cost of household consumption rather than simply inflating individual prices. The study develops the AI Infrastructure Net Cost Pass-Through Model (AI-NCPM), a four-layer conceptual framework tracing net cost flows from data center investment to sectoral allocation and household outcomes. The model’s parameters are analytically specified but not empirically calibrated; numerical examples are illustrative rather than representing estimated effects. Its main contribution is an integrative framework linking cost pass-through, infrastructure cost socialization, two-sided platform allocation, environmental externalities, and household expenditure incidence within a single net-cost account. Because these effects originate in the design, construction, energy and cooling systems, and lifecycle operation of data centers, the analysis connects AI infrastructure economics to the built environment. The framework suggests that low-income, minority, rural, older adult, and disability-affected households may face disproportionate net burdens, as costs fall heavily on essential services while benefits accrue more readily to affluent and digitally connected households.