The Growing Concern Over Big Tech Data Centers
From an early stage in life, individuals develop a sense of proportion along with their understanding of right and wrong. It’s crucial to evaluate actions, even those perceived as good, within the right context. Unfortunately, this sense of proportionality seems to be missing among advocates of Big Tech as they push to establish extensive AI data centers throughout the country.
A recent article highlighted the situation in Archibald, Pennsylvania, illustrating why these data center projects present unprecedented challenges that could undermine the entire concept of generative AI. The locals are being asked to sacrifice significant resources for what can only be described as an uncertain and speculative promise.
Archibald, a small town nestled in the Pocono Mountains with around 7,000 residents, is contemplating the construction of six massive data centers. One of these facilities would take up roughly 14% of the town’s land.
Planned campuses would include 51 warehouses, each comparable in size to a Walmart Supercenter. Once fully operational, they would cover around 4.2 kilometers of land. This approach is unprecedented for the community, which has never undergone anything of this magnitude before. It’s occurring in various forms across numerous regions, contributing little for residents in return, especially when more efficient options like edge computing exist.
In recent weeks, several members of Archibald’s Borough Council and planning commissioners have stepped down, reflecting the growing turmoil surrounding these developments.
It’s important to note that Big Tech firms are acquiring and zoning vast amounts of land at an unprecedented pace. Supporters of this movement might label skeptics as anti-growth, but this situation differs from other infrastructure projects, particularly in terms of return on investment.
For instance, a gas power plant can power a region efficiently with a small footprint, while these massive data centers consume considerably more space and resources. Moreover, they contribute an alarming sound level of around 90 decibels to nearby residents—the whole endeavor rests on the shaky premise of advancing “artificial general intelligence.” It often feels like a mirage.
If Archibald’s proposal really required an extensive area to support a large language model, that might be one thing. However, the plan calls for super-sized facilities across countless locales nationwide.
To put things in perspective, consider the eight planned hyperscale data centers in Indiana alone. Together, they are projected to require 8,300 megawatts of power, which is roughly equivalent to the electricity needs of double the number of households in the entire state.
In Prince William County, a location already overrun with data centers, a new project entailing one of the world’s largest campuses—spanning 2,100 acres and housing 37 buildings—could be greenlit without significant community opposition. Given its size, it could accommodate far more than 50 gas plants.
In Box Elder County, Utah, plans are progressing for a massive data center on 40,000 acres of land that will demand 9 gigawatts of power when complete, an amount that dwarfs the four gigawatts currently used across the state.
The sheer scale of energy these giants require also necessitates an unnatural number of power lines, intruding upon private property and community landscapes. For instance, a proposed $1 billion high-voltage transmission line project from Virginia aims to impact nine counties with towering structures ranging from 135 to 165 feet.
The situation is notably dire in West Virginia, where residents are being compelled to clear land for power lines intended for a “data center array” situated in Northern Virginia, far from their own state.
It’s hardly surprising, then, that there’s a growing bipartisan backlash against this model, seen as absurd by many. Communities are being called to surrender substantial funding for an uncertain promise of returns.
In the end, it seems more likely that we’ll be left grappling with increased surveillance, deteriorating quality of life, and a problematic internet, rather than enjoying significant advancements in human achievement at the cost of land and resources.
What was originally thought to be a forward-thinking technology now appears archaic and depriving in terms of resources. It’s evident that the industry can’t sustain the operation of thousands of these data centers. The model based on large language model scaling is simply not viable for the future of AI. So, is it possible to halt this trend before most of our land is repurposed for centralized monopolies?
Keep in mind that land acquisition isn’t just an incidental outcome; it is a central aspect of this challenge.
