What the US Movement Against Data Centers Can Learn From Abroad
A new book by Paris Marx highlights how international communities, particularly Ireland, offer crucial lessons for the US movement against the rapid expansion of data centers, especially those fueling generative AI.
Intelligence analysis by Gemini 2.5 Flash
The article, an interview with author Paris Marx, explores the global history of resistance to data centers, arguing that the US is now facing issues that have already played out in other countries like Ireland and Chile. It emphasizes the significant energy and water demands of these facilities, exacerbated by the rise of generative AI, and warns of a potential 'AI bubble' with subst…
Imagine giant computer brains, called data centers, that help make AI work. These brains need a lot of electricity to think and a lot of water to stay cool, like a big factory. But when too many are built too fast, they can use up so much power and water that there isn't enough for the people living nearby, or it makes the air dirtier. Other countries have already learned this the hard way, and now the US is starting to see the same problems.
Analysis
The rapid expansion of data centers, largely driven by the insatiable demands of generative AI, is creating significant environmental and social challenges across the United States. However, as Paris Marx's book "Hyperscale" illustrates, these issues are not new and have been confronted by communities worldwide for years. The US anti-data center movement can draw vital lessons from these international experiences, particularly regarding energy consumption, water usage, and the potential for an economic bubble.
Ireland's Energy Grid
Ireland serves as a stark warning for what happens when data center growth outpaces grid capacity. Dublin implemented a moratorium on new data centers in 2021 after the grid operator declared insufficient capacity. This forced the industry to adapt, seeking to build facilities in other parts of Ireland and even developing dedicated gas power plants to bypass the public grid.
This shift led to lobbying efforts by data center companies to allow private entities to build electricity cables, raising significant concerns about equity and public control over essential infrastructure. The Irish Environmental Protection Agency also warned that these new gas plants would make climate targets much harder to achieve, a scenario now unfolding in the US with companies racing to build similar private power solutions.
The Dalles, Oregon
The issue of water consumption by data centers is another critical area where international and early US experiences offer insights. Google's operations in The Dalles, Oregon, became a flashpoint when the company attempted to conceal its water usage figures. It took community pressure and a year-long struggle to reveal that Google was consuming 29 percent of the community's water.
This incident underscores the lack of transparency often employed by tech giants and the significant strain their operations place on local resources. As more communities face similar proposals for data center construction, the experience of The Dalles highlights the importance of public oversight and demanding accountability for resource consumption.
The AI Bubble
Marx suggests that the current surge in data center construction, fueled by generative AI, might be an "AI bubble" similar to the dot-com boom and bust. The industry is planning massive build-outs of data centers and associated gas power plants, but faces challenges in securing necessary components like turbines and chips. This raises questions about whether these facilities will even be completed before a reassessment of AI demand occurs.
If the AI bubble bursts, there's a significant risk of stranded assets—expensive data centers and gas plants that become obsolete or underutilized. While some data center companies survived the dot-com bust by acquiring distressed assets, the rapid obsolescence of AI-specific chips means that even existing data center structures might not be valuable for future demand, leaving behind environmental and economic liabilities.
Key points
- The US anti-data center movement can learn from international experiences, particularly Ireland's, regarding energy grid strain and lobbying by tech companies.
- Google's water usage controversy in The Dalles, Oregon, highlights the lack of transparency and significant local resource demands of data centers.
- The current build-out of data centers for generative AI may constitute an 'AI bubble,' risking stranded assets and environmental liabilities if demand falters.
- Data center companies in Ireland lobbied for legislation allowing private companies to build electricity cables and are building dedicated gas power plants.
- The rapid cycling out of AI chips means that even if data centers are built, their specialized components may quickly become outdated, reducing their long-term value.
The article suggests that communities can learn from past experiences and potentially adopt more sustainable approaches, such as the community-led effort in New Zealand to create local cloud infrastructure for specific cultural needs, offering a model for more equitable and environmentally conscious data solutions.
The rapid, unchecked expansion of data centers for AI risks creating an "AI bubble," potentially leading to a glut of expensive, energy-intensive infrastructure like gas power plants that become obsolete or underutilized if AI demand is reassessed, leaving communities with stranded assets and exacerbated environmental issues.
