What the US Movement Against Data Centers Can Learn From Abroad

The cost of energy for data centers could increase as companies build private gas plants, impacting operational expenses.

Data centers are becoming a significant issue in US midterm elections due to their substantial energy and water consumption. A new book by Paris Marx highlights global precedents for resistance against data center development, citing moratoriums in the Netherlands and Singapore since 2019, and growing concerns in Ireland over energy grid strain. In The Dalles, Oregon, Google's attempt to conceal its massive water usage, which accounted for 29% of the community's supply, led to public outcry. The explosion of generative AI, exemplified by ChatGPT's release in late 2022, is expected to accelerate the demand for data centers. Ireland's experience shows how industry pushback against grid connection limits can lead to the construction of private gas power plants, impacting climate targets and raising equity concerns. The article draws parallels to the current US trend of building private gas plants for data centers and discusses the potential for an AI bubble to burst, leaving behind underutilized infrastructure.

The proliferation of data centers for generative AI is creating a global tension between technological advancement and environmental sustainability. Countries like Ireland have seen the energy grid strained to its limits, prompting data center companies to lobby for private infrastructure development, including dedicated gas power plants. This move, while securing energy supply for AI, has led to concerns about meeting climate targets and equitable resource distribution, suggesting a potential for similar conflicts in other regions, including the US.

The prospect of an AI bubble bursting raises questions about the long-term viability of the massive infrastructure being built. If AI demand falters, the specialized hardware in data centers could become obsolete, leaving behind expensive, underutilized facilities and dedicated power plants. This situation echoes the dot-com bust, where excess capacity was eventually absorbed, but the rapid obsolescence of AI-specific components presents a new challenge, potentially leading to stranded assets and further environmental impact from unused infrastructure.