The AI Data Center Boom: An Investment Masterpiece or a Fear-Driven Market Bubble?

Electricity, water, and higher bills? Is this the price we pay for AI development?

Today, it is hard to find anyone who hasn't encountered artificial intelligence technology. Even if we have missed the headlines of business news outlets, we have certainly stumbled upon AI-generated videos on TikTok or Instagram. Behind this widespread digital noise, which might amuse us on a daily basis, lies a brutal and costly reality. In 2025 alone, tech giants allocated around $400 billion to AI infrastructure. This is no longer gradual expansion and development, but a complete all-in bet on a single card called "data centers."

Currently, more capital is flowing into the construction of these digital factories than into residential housing in many regions of the world. On paper, it looks like unstoppable growth, yet the market is preparing for a collision with physical limitations. The data center boom is raising—and for good reason—growing social concerns, ranging from gigantic energy consumption and water supply disruptions to a tangible impact on utility bills. Will this massive race ultimately improve our quality of life, or, on the contrary, will Instagram's AI slop become a costly burden for us?

What is an AI data center?

Before we evaluate whether billion-dollar investments will translate into higher electricity bills and a lack of running water, we must take a step back. To understand where this gigantic appetite of data centers for water and energy comes from, and why investors have rushed into this race, it is worth grasping what this digital colossus actually is and how it differs from the server rooms we have known so far.

Check out our products

Piano di marketing HVAC per un'elevata conversione

€4,99

Piano di marketing sui social media per social media manager e agenzie

€3,49

Modello di analisi PEST (16 pagine)

€1,79

Think of a traditional data center as a digital warehouse: it serves to store files, host websites, and maintain basic corporate software. It is functional, but it does not "think." An AI data center, on the other hand, is a factory of high-performance engines.

These facilities are densely packed with thousands of advanced graphics processing units (GPUs)—chips designed to solve billions of mathematical problems simultaneously using matrix multiplication. Because these chips generate massive amounts of heat and require gigantic voltage, these buildings resemble industrial power plants more than office spaces. If a regular data center is a library, an AI data center is a super-powered brain running at full speed, consuming vast amounts of electricity to transform raw data into intelligence.

Why did this topic explode only now?

The first true facilities dedicated entirely to machine learning began to emerge around 2016, when Google deployed its proprietary TPU chips, and Microsoft built a giant supercomputer with over 10,000 GPUs for OpenAI. For years, however, this infrastructure developed quietly, in the shadows and away from the public eye.

Everything changed with the debut of ChatGPT in late 2022. Suddenly, the technology ceased to be an abstract concept from technical laboratories and became a tool of everyday use for hundreds of millions of people. This avalanche of interest forced an unprecedented race. The AI Empire began to grow, the demand for computing power exploded overnight, and tech giants started massively constructing buildings with extremely high electricity and cooling requirements.

This rapid leap caused the topic to move beyond industry business portals and into public debate. People began to realize that every generated image or prompt sent to a language model leaves its very own physical footprint on Earth.

The bill for Artificial Intelligence: Water, electricity, and neighboring a giant

Today, social concerns are no longer just about the mere existence of artificial intelligence, but about the cost we bear as a community: the strain on local power grids, the consumption of millions of liters of drinking water to cool servers, and the potential impact of these factors on the increase in our daily bills. The question, "Is an AI center being built in my neighborhood?" is no longer surprising, as residents increasingly notice that the digital cloud requires thousands of liters of drinking water daily and massive access to the power grid.

The biggest concern is the fact that vast amounts of water are used to cool servers handling highly demanding AI tasks. With continuous intake, this leads to a drop in groundwater levels and the drying out of surrounding areas. Instead of promises of "modern jobs," local communities see the specter of dry taps and overloaded electrical grids.

This has caused tech giants to encounter a level of resistance they had never known before. Residents are beginning to actively protest and block investments:

• USA (Virginia) and the Netherlands: In the so-called "data center valley" in the state of Virginia and in cities in the Netherlands, pressure from residents and activists has led to the introduction of temporary moratoriums on the construction of new facilities due to the threat to the local water and energy balance.

Key data centers in the USA

Below is a compilation of 50 key data centers in the USA, prepared based on data from the datacente.rs platform:

What does the distribution of data centers in the US look like?

The map clearly indicates that American digital infrastructure is heavily consolidated around a few key technological and financial hubs. The East Coast, with the New York area and Northern Virginia, remains the absolute leader with as many as 322 facilities, followed closely by California (158) and Texas (170). Industrial and logistics zones in the Midwest also play a significant role, including the Chicago area with 127 facilities and Ohio with 148. It is evident that investments are concentrated in locations with strategic access to fiber-optic infrastructure and massive power reserves, bypassing less urbanized central states.

What does an AI center look like?

An AI data center is essentially a self-sustaining city. Although its primary purpose is computation, the facility is defined by its extreme auxiliary infrastructure. These are not open-space offices; they are high-security fortresses featuring tall fences, anti-ramming barriers, and multi-factor authentication (including biometrics and badges) at every entry point. Inside, thousands of servers sit within industrial-grade enclosed cages, connected by miles of high-speed fiber optics and powered by massive local energy distribution systems and battery banks that ensure "thinking" never stops.

The most striking feature of a modern AI data center is how it addresses the physical limits of hardware. As GPU density increases—with some server racks now demanding up to 600 kW of power—traditional air cooling is no longer sufficient. To prevent silicon from melting, these facilities have transitioned to direct liquid cooling, where coolant is pumped directly to the chips to dissipate heat efficiently. This shift has transformed these buildings into heavy industrial plants requiring high-capacity electrical substations and, in many cases, rows of backup diesel generators to guarantee uninterrupted power delivery under the continuous, massive loads required for 24/7 inference.

Which US data center is the largest?

The answer depends on how "size" is defined whether by power capacity or floor area:

• By computing power capacity (for an AI footprint): Currently, the largest complex dedicated strictly to artificial intelligence is xAI's Colossus in Shelby, Tennessee, with a target capacity exceeding 1.5 GW.

• By single building size: The title belongs to Switch's TAHOE RENO 1 facility in Nevada, spanning a massive 1.3 million square feet (approximately 120,000 m²).

Are we facing an AI Boom without profits?

Despite all the enthusiasm surrounding artificial intelligence, a troubling consensus is growing within the industry: most companies have still not figured out how to generate stable and predictable revenue from it.

At this point, the undisputed winners remain infrastructure providers and chip manufacturers. For them, the mere fact that tech companies are racing to purchase hardware is enough; they do not need the end applications to generate a return on investment (ROI). This creates a dangerous feedback loop: spending increases exponentially while the business case on the software side remains in the stage of promises.

Currently, the entire AI infrastructure operates within the tightly sealed framework of the so-called AI Bubble. The market values new ventures not based on generated income, but on pure market speculation and the fear of missing out (FOMO). Experts are increasingly referring to it directly as a classic investment bubble. Furthermore, allegations are emerging that the current AI model relies on multi-layered ties between tech giants and an unprecedented physical drain on valuable natural resources (energy and water), which so far has failed to deliver proportional added value to the economy or society.

The contradiction at the heart of the boom

Coupled with this financial speculation is another glaring contradiction. At the very moment giants announce new multi-billion-dollar investments, a significant portion of data center construction projects is experiencing severe delays or being quietly scaled back. Some never progress beyond the early earthwork stages.

If demand is supposedly exploding, why is the infrastructure failing to keep pace? Part of the answer lies in marketing. The capacity announced in the media sounds impressive, but headlines like "under construction" cover everything in practice from an almost finished facility to an empty plot with a poured foundation. The gulf between what was promised to investors on paper and what actually operates and generates revenue is wider today than ever.

Summary

The AI data center boom combines features of a digital breakthrough with market speculation, yet the scales are tipping increasingly toward an investment bubble fueled by the fear of missing out (FOMO).

On one hand, it represents a visionary technological leap that builds the foundation for future innovation. On the other hand, the current pace of expenditure is colliding with a harsh reality. The true constraint is no longer the availability of GPU chips, but physical, finite resources. The main bottleneck of the entire system has become energy infrastructure: transformer shortages, a lack of grid connections, and severely constrained electricity and water supplies. Consequently, fully constructed facilities stand empty waiting for the grid to catch up, while new projects are routinely postponed.

In parallel, uncertainties surround the profitability of the technology itself, and the actual business value of AI is growing increasingly uncertain. Instead of stable profits, the visible model is driven by massive resource consumption without proportional added value. This stands in sharp contrast to the concept of a circular economy: rather than heat recovery, resource reconfiguration, or energy reuse, the AI production line currently functions on a model of extensive, linear depletion. As long as the construction of "digital factories" outpaces power grid capacity, sustainability principles, and actual application revenue, the entire industry remains on exceptionally thin ice.

Sources:

"The $400B AI Boom is real - but power bottlenecks and overbuilding threaten to redefine the global data center landscape" - Global Data Center Hub, October 18, 2025

Data Center Expansion in Virginia: Closing critical gaps for informed water planning and permitting, 2024-2025

xAI Supercomputer - Memphis & The Colossus Data Center Cluster, Wikipedia

Global & U.S Data Center Directory & Technical specifications, continually updated dataset (2024-2025)