Field of (Server/Electric Sheep) Dreams

October 27, 2025.

(Excerpt from The Red Shoes…)

Build it and they will come. In this case, I am referring to elds full of servers, of buildings to house the servers, lled with air-conditioning systems to cool the servers, connected to the power grid, to power the servers. The elds of servers (aka “data centers”) must expand as the internet and its use expands. Tech insiders like Senior VP at Supermicro, Vik Malyala view this expansion as inevitable and unstoppable. According to Malyala: “With 5G networks and an ever-increasing number of IoT [internet of things] devices, it becomes clear why data centers are crucial for our world as data usage is only continuing to grow.”

And Steven Santamaria, writing for Dataversity, echoes this sentiment, calling data centers “indispensable”; and the “backbone of our digital lives”, serving our “appetite for digital consumption”; one Santamaria (I believe erroneously) calls a “need”, which is “growing exponentially in the age of articial intelligence (AI), machine learning (ML), remote work, IoT, and big data.”

In 2012, there were 500,000 data centers worldwide, and by 2020, this number jumped to over 8 million. This number will only grow, if nothing changes, if our world does not stop its insatiable hunger for data. The eight million (and counting) elds of servers consume enough electricity to power whole nations, spewing out carbon emissions on the level of the airline industry, and the electricity used comes to about 1.5% (in 2020), and 3% (in 2024), of global power consumption. By 2030, some estimates predict this number will jump to over 10%.

And as data centers “upgrade” hardware, they eliminate old hardware in the metric tons, generating approximately two million tons of e-waste annually, sometimes, ironically, in an effort to “go green” (with hardware demonstrating more energy efciency), as if data centers can ever be greener than the natural elds, or forests, they replaced.

Much of the e-waste generated from the data centers contains hazardous materials, such as lead, mercury, ammonia and chlorine. And, as with all other digital tech, mining of rare-earth minerals is needed for server components, and construction of vast data centers requires concrete and steel—all of which exacts a high environmental toll.

Worse yet, and most often overlooked, is the amount of water data centers consume for their cooling systems, which are water-based (utilizing “evaporative cooling”), because these use less electricity than non-water- based cooling systems, and because water is a cheaper utility for tech companies than electricity.

The waste water used in these systems is also dumped back to its source full of chemicals and at heightened temperatures, harming aquatic life. But since a large majority of data centers are housed in the southwest deserts of the US (due to the availability of cheap land), in the same parts of the country that are water-stressed (especially as an over decade-long drought continues), the average 3-5 million gallons of water, representing just one typical data center’s usage per day (the same amount a city of up to 50,000 people consumes), totaled 174 billion gallons of water in 2020. This amount of water allocated to machines, and away from human resources like drinking water and agriculture, should be considered criminal.

In Arizona, tech companies, like Facebook and Google, have been building new data centers in the past few years (which bring very few employment opportunities with them), drawing on water from Lake Mead and the Colorado river, bodies of water that have already dwindled to record low levels (and as the levels continue to drop, federal restrictions will kick in, forcing tech companies to use their issued “water credits” for groundwater supplies)—all so that you and can see who “liked” your latest social-media post or search for trivial information—which, let’s face it, comprises the bulk of internet use; trivialities, entertainment, and distractions.

Water disputes are not only surfacing in America’s desert regions, Google came under re in 2017 by conservation groups in South Carolina when the company requested a permit to draw 1.5 million additional gallons of water per day from an already depleted aquifer for its Goose Creek facility, which was already sucking up 4 million gallons of tap water, per day. In Iowa, in training AI models, cooling supercomputers draws even more water from the state’s watersheds. For each series of 5 to 50 prompts used in OpenAI’s ChatGPT, 500 milliliters of water is needed.

When it comes down to water for growing crops, water for drinking, and water for servers, who will be allocated the remaining drops? Companies with “water credits” to feed machines, or living humans and animals? How much data do we really need? And is it more important than survival?