A modern data center is strangely quiet. Steady, but never silent. A sort of industrial hum that, after a few minutes, becomes ingrained in your bones. Going through one is more like entering a power plant that has forgotten to introduce itself than it is like visiting a tech facility.
With cables draped like veins, rows of servers blink in amber and green. In order to combat heat that never completely goes away, cool air pushes upward from the floor. It’s difficult to ignore how physical everything is. Despite its abstraction, AI resides in structures like these—heavy, noisy, and incredibly real.
| Category | Details |
| Industry | Artificial Intelligence & Data Infrastructure |
| Key Players | Amazon, Microsoft, Google, Meta, OpenAI |
| Core Asset | Data Centers (“AI factories”) |
| Estimated Investment | $500 billion+ (Stargate Project and others) |
| Energy Demand | Rapidly increasing; AI queries use ~10x energy of search |
| Water Usage | Hundreds of millions to billions of gallons annually |
| Workforce Impact | Demand for ~500,000 electricians (next decade) |
| Growth Trend | Data center demand rising ~9% annually |
| Reference | https://www.iea.org |
For many years, algorithms, innovations, and software dominated the artificial intelligence narrative. However, the topic of infrastructure has recently come up, albeit grudgingly. In actuality, AI is not limited to using data. It is powered by a lot of water, electricity, and land.
It seems that executives and investors are more aware of this than the general public. In addition to hiring engineers, businesses like Google and Microsoft are also negotiating water rights, purchasing land, and obtaining energy contracts. The scale is more akin to an industrial boom, like railroads or oil pipelines, than a tech boom.
Think about the speed. Demand has skyrocketed since the introduction of generative AI tools into daily life in late 2022. Hundreds of millions of people use ChatGPT every week. Despite their apparent simplicity, each query necessitates a substantial amount of computation in the background. A single AI query may use up to ten times as much energy as a typical web search, according to estimates. Then multiply that by millions, and finally billions. All of a sudden, the math starts to feel awkward.
New structures are emerging outside of cities in areas that were previously considered peripheral. Large, windowless structures nestled into the periphery of suburbs or rising from farmland. Communities in Louisiana, Pennsylvania, and Texas are observing these projects with a mixture of interest and anxiety.
In a field that had been farmed for decades, one resident reported seeing survey stakes. Construction workers showed up months later. The land underwent rapid change. What was once used to cultivate crops is now being prepared to house servers, which are machines that never go to sleep.
Whether these advancements will yield the anticipated economic gains is still up for debate. During construction, job numbers can seem impressive at first—thousands—but once operations stabilize, they drastically decline. a few hundred permanent positions, occasionally fewer. The infrastructure remains in place in the interim.
The most pressing issue is probably the demand for energy. This type of load was not intended for the U.S. grid, which was primarily constructed in the middle of the 20th century. Some data centers now need as much electricity as small cities; they are no longer modest establishments. Within ten years, the power demand from these centers is predicted to double.
Utilities are obligated to provide that energy. Additionally, prices typically increase when supply becomes more constrained. Though few express it directly, there is a subtle worry that regular consumers might wind up paying for the infrastructure supporting AI.
Another problem that is equally urgent but less obvious is water. It is essential to cooling systems. Data centers use hundreds of millions of gallons a year in some areas. More than a billion in others. Businesses discuss sustainability and recycling, and some are actually taking action. However, it’s not always feasible to return water to its original ecosystem.
As you watch this happen, you get the impression that something fundamental is changing. AI is evolving from software to infrastructure. Something more like transportation or electricity. Maybe essential. However, it’s not equally distributed and demanding.
It appears that governments are keen to expedite the process. Permits are being streamlined, policies are being modified, and environmental reviews are being shortened. Competitiveness and winning a worldwide race are common themes in the language. “Build faster” has evolved into a sort of unofficial catchphrase.
However, speed has drawbacks. Communities are beginning to rebel, raising concerns about long-term effects, energy costs, and water tables. These are not impersonal issues. They are immediate, local, and challenging to overlook once construction gets underway.
There is a balancing act going on at the corporate level. Numerous businesses that are spearheading the growth of AI have also made bold climate commitments. targets for renewable energy, water positivity, and net-zero emissions. However, achieving those objectives is more difficult due to the size of AI infrastructure.
Efficiency gains—better chips, more intelligent algorithms—might relieve some of the strain. It is being worked on by engineers. However, demand continues to rise, frequently more quickly than efficiency improvements can keep up. At the core of the AI boom is this silent paradox.
We use the systems more and more as they get smarter. They need more infrastructure the more we use them. Additionally, it gets more difficult to overlook the expense as we construct more infrastructure. not only monetary. physical.
It’s evident that the future of AI isn’t floating in the cloud when you stand in one of those server halls and take in the steady hum and rush of cooled air. It is grounded. bolted to the ground. consuming water, drawing power, and changing landscapes in ways that are still in the early stages of comprehension.
