The AI boom is upon us, and it’s being powered by a global network of multi-billion-dollar facilities known as data centers. Outfitted with thousands of high-performance GPUs, sprawling storage arrays and hyperspeed networking equipment, these complexes have become the backbone of the artificial intelligence industry, enabling every chatbot response and AI-generated image. Now, as demand for computing power surges, Big Tech’s rush to expand their data centers’ capacity has reached code-red levels of urgency.
What Is a Data Center?
Data centers are the physical facilities that power the digital world. Ranging from a few hundred square feet to several acres, they store, process and move around vast amounts of data, driving everything from cloud services to artificial intelligence.
In 2025 alone, Amazon, Google, Microsoft and Meta are projected to collectively invest up to $320 billion in their AI infrastructure, up from the previous year’s $230 billion total. And that’s just to build. This technology requires enormous amounts of energy, with one ChatGPT query consuming up to ten times more electricity than a typical web search. The data centers supporting those queries also rely on high-capacity cooling systems that circulate ultra-purified water to keep servers from overheating. If current growth continues, these facilities could use up to 12 percent of the country’s electricity in the next few years — nearly triple its current share — putting a strain on neighboring small businesses and residents through depleted water supplies and rising utility bills.
Whether today’s systems can withstand these massive energy and water demands (or whether these investments will even pay off) remains uncertain. But for many stakeholders, that uncertainty seems almost beside the point. As local communities hit their limits, tech companies and policymakers are pressing ahead to stay competitive in the global AI race. The challenge now is to scale fast enough to keep pace without letting the infrastructure underneath buckle under its own weight.
Top U.S. Data Center Projects
A recent Business Insider investigation found that Amazon, Meta, Google, Quality Technology Services (QTS) and Microsoft lead the nation in large-scale data center expansion. The details of these facilities are often tightly guarded, but here’s what we do know about some of the biggest projects popping up across the country.
Amazon’s Ohio Expansion
- Estimated Cost: $23 billion
- Location: Fayette County, and other undisclosed sites
- Status: Under construction
Adding to its leading tally of 177 data centers and counting, Amazon announced a $23 billion multi-year project to build out more facilities across Ohio. The plan starts with a 590-acre campus in Fayette County and could include up to eight additional sites across the state to support growing demand for AI training and inference workloads, while expanding capacity and reducing latency for its cloud customers. All told, Amazon operates 28 data centers in the Buckeye State.
xAI’ Colossus 1 and 2
- Estimated Cost: More than $7 billion
- Location: Memphis, Tennessee
- Status: Operational and expanding
xAI is expanding its Memphis-based AI supercomputing campus with Colossus 2, the next phase of founder Elon Musk’s effort to build the world’s largest AI training cluster. The project follows the success of the first Colossus, a 100,000-GPU data center built in just 122 days to support the training of the company’s Grok models. Colossus 2 is expected to expand total capacity to nearly one million GPUs and 1.2 gigawatts of power, dramatically increasing compute performance across Musk’s tech empire, including xAI, X and Tesla-related AI projects. The facility is being developed on a 285-acre site, where xAI also plans to construct a water treatment plant to support the campus’ high cooling demands.
The QTS Atlanta Metro Data Center
- Estimated Cost: Undisclosed
- Location: Atlanta, Georgia
- Status: Operational and expanding
The QTS Atlanta-Metro Data Center spans 990,000 square feet across a 99-acre campus, making it the second largest data center in the world. The facility is powered by three on-site Georgia Power substations and offers direct fiber access to major carrier networks. With more than 120 megawatts of available power, QTS reportedly provides custom data center, colocation and cloud services to hyperscale clients like Google, Microsoft, AWS and Meta (although they do not publicly name names).
The Stargate Project
- Estimated Cost: $500 billion
- Location: Abilene, Texas; Shackelford County, Texas; Milam County, Texas; Doña Ana County, New Mexico; Lordstown, Ohio and more to come
- Status: Operational and expanding
The Stargate Project is a massive government-backed AI infrastructure initiative led by OpenAI, Microsoft, Oracle and SoftBank, set out to invest $500 billion in AI-first data centers across the country by 2029. So far, they’ve achieved seven-gigawatt’s worth of planned capacity (the equivalent output to seven nuclear reactors), anchored by its first operational site in Abilene, Texas, a one-gigawatt facility outfitted with Oracle Cloud Infrastructure and Nvidia GPUs. Construction is underway at additional sites in New Mexico, Ohio and a handful of other locations — all working toward a combined 10-gigawatt capacity target by the end of the year.
Microsoft’s Mount Pleasant AI Campus
- Estimated Cost: More than $7 billion
- Location: Mount Pleasant, Wisconsin
- Status: Under construction since 2024
Microsoft is building the Fairwater AI data center campus in Mount Pleasant, Wisconsin. With total investment exceeding $7 billion, the project includes two facilities, with the first — a 315-acre campus spanning over 1.4 million square feet — scheduled to come online in early 2026. The campus will house hundreds of thousands of Nvidia GPUs, use advanced closed-loop liquid cooling technology and a 250-megawatt solar project. Construction has employed over 3,000 workers, and once both campuses are fully operational, they are expected to create up to 900 full-time jobs.
Meta’s Altoona AI Data Center
- Estimated Cost: About $2.5 billion
- Location: Altoona, Iowa
- Status: Operational, with plans for expansion
Meta’s Altoona data center campus in Iowa is one of the largest facilities of its kind in the world by size, spanning over 5 million square-feet. Since breaking ground in 2013, the campus has expanded through several multi-billion-dollar projects, adding tens of thousands of AI accelerators and advanced liquid cooling powered by on-site renewable energy, all to support AI training and inference for Meta’s LLaMA models and recommendation systems. Fully operational, the facility supports over 400 full-time staff members, and has partnered with local schools and nonprofits to provide training and community programs. Most recently, Meta announced further investments to expand renewable energy capacity in Iowa, including a 225-megawatt wind project to support the facility’s growing AI operations.
Meta’s Kuna Data Center
- Estimated Cost: About $800 million
- Location: Kuna, Idaho
- Status: Under construction
Meta is building a 960,000-square-foot data center in Kuna, Idaho — one of the largest technology investments in the state — designed to support AI model training, content ranking systems and next-generation computer infrastructure. The company paused the project shortly after breaking ground in 2022 to switch gears to an AI-first redesign. At peak construction, the facility has employed 1,000 skilled trades workers and will support roughly 100 permanent jobs once fully operational. Using 100-percent renewable energy, Meta aims to offset the facility’s energy demand with a 200-megawatt solar project and a city-owned water and wastewater system designed to restore more water than it consumes.
Google’s Project Mica
- Estimated Cost: About $10 billion
- Location: Kansas City
- Status: Under construction
Google is planning a 500-acre data center campus in Kansas City’s Northland, part of what the city and developers are calling Project Mica. The project is expected to cost around $10 billion, and will include multiple data center buildings optimized for AI workloads with advanced TPU clusters and custom cooling. Google aims to scale operations for Gemini, DeepMind and cloud services, while also bringing an estimated 1,000 construction jobs and around 200 permanent positions to the Kansas City region once the site becomes operational. It will, however, require up to 700 megawatts of power (enough energy to supply roughly 400,000 homes) to operate.
Texas Critical Data Centers’ Ector County AI Campus
- Estimated Cost: Undisclosed
- Location: Ector County, Texas
- Status: In development
Spanning 438 acres, Texas Critical Data Centers — a joint venture between New Era Energy & Digital Inc. and Sharon AI, Inc. — is building its Ector County campus to support large-scale AI model training, with plans to eventually handle up to one gigawatt of power. The facility will integrate high-density computing with behind-the-meter gas-fired power generation and potential carbon capture technology. The first phase is slated to come online by December 2026, initially providing 250 megawatts of AI-ready capacity.
Apple’s Mesa Data Center Expansion
- Estimated Cost: $2 billion
- Location: Mesa, Arizona
- Status: Active and expanding
Originally a manufacturing site, Apple’s Mesa data center is being converted into an AI and iCloud hub, part of a $500 billion, four-year investment to ramp up domestic expansion and manufacturing that will span across nine states. The company has quietly invested billions to upgrade networking and power infrastructure to support new on-device AI and privacy-focused computing services, as well as advanced cooling systems and on-site solar farm for the 1.3-million-square-foot facility. Most recently, Apple completed a capacity expansion in April 2025, adding several new data halls to increase its computing capabilities.
Tract Data Center Campus
- Estimated Cost: Up to $20 billion (excluding server equipment)
- Location: Buckeye, Arizona
- Status: Planned, in development
Tract Data Centers’ massive 20-million-square-foot campus in Buckeye, Arizona is one of the costliest planned U.S.-based facilities on record so far. Spanning over 2,000 acres, the 1.8-gigawatt complex will house as many as 40 individual data centers to be built out in multiple phases over a 15-year period. With next-generation GPU clusters and high-density compute infrastructure, it’s being designed to take on the world's largest AI model training and inference workloads.
Who’s Paying for All These Data Centers?
Global data center spending is set to reach $7 trillion by 2030 — and most of the cash is coming straight from the tech giants themselves. Companies like Microsoft, Amazon, Meta and Google are pouring tens of billions of dollars into their cloud and AI infrastructure, while local governments sweeten the deals with tax incentives and favorable financing. Some projects also rely on private loans or partnerships with investment banks, spreading out the cost of land, construction and the specialized chips and cooling systems that keep these facilities running.
But the funding isn’t always as straightforward as it looks. Nvidia, for instance, pledged roughly $100 billion to OpenAI, supplying GPUs while simultaneously investing in the company’s infrastructure. Similarly, AMD struck a deal that lets OpenAI buy a chunk of its shares in exchange for chip contracts. Even Oracle has joined the party, selling $300 billion-worth of cloud capacity to OpenAI while investing $40 billion in Nvidia’s chips to power the service. The result is a tight circle of intertwined investments: Nvidia funds OpenAI, OpenAI buys cloud capacity from Oracle, Oracle buys Nvidia’s chips and Nvidia holds a major stake in CoreWeave — one of OpenAI’s key cloud partners. In effect, the same dollars are cycling through a small handful of companies, boosting one another’s valuations along the way.
Analysts warn this kind of financial choreography can make the industry appear far healthier than it actually is, likening it to the dot-com era of the late 1990s, when a small number of tech firms fueled one another’s stock prices until the bubble burst — only this time around the bubble is about 17 times as big, according to some estimates. Financial strategists also say these “related-party transactions” can blur the line between genuine revenue and circular accounting, creating an illusion of growth that’s built more on mutual reinvestment than on real economic output. Under this chain-link system, if even one player misses its growth targets or faces investor pressure, the house of cards can collapse, destabilizing the entire network.
Even so, the spending spree shows no signs of slowing down. Investors are betting that AI’s long-term payoff will outweigh any short-term risk. Whether companies are building colossal hyperscale campuses from scratch or retrofitting old ones, the endgame is the same: to scale fast enough to keep pace with AI’s explosive growth.
Who Runs These Data Centers?
Data centers often start as massive “build-out” construction projects. In the beginning, they’ll employ thousands of electricians, steelworkers and HVAC crews before scaling down to a couple hundred (or even a couple dozen) full-time roles once the facilities go live. Those permanent teams — typically drawn from the local workforce — are made up of network engineers, data center technicians and electrical specialists who keep the racks of GPU clusters and fiber connections running, along with industrial-scale cooling systems that circulate water through rows of heat-dense hardware.
While tech companies may own these sites on paper, they often rely on contractors who specialize in managing large, mission-critical facilities, to oversee day-to-day operations while their own in-house engineers monitor AI performance remotely.
Where Are These Data Centers Being Built?
Just fifteen states account for 80 percent of the country’s entire data center load, with Virginia alone making up for about 25 percent of that capacity. Home to “Data Center Alley,” the state has more than 500 facilities, thanks to its dense fiber networks, stable power grid and generous tax incentives. Generating up to 6,000 megawatts across 40 million square feet, this corridor hosts 70 percent of all internet traffic.
Most data centers are concentrated in a few key regions, where sprawling land, reliable and high-capacity infrastructure and favorable policies meet. Maricopa County, Arizona, has emerged as one such hub, accommodating dozens of hyperscale facilities despite being a water-stressed desert. When Google chose Mesa as the location for its Red Hawk data center, it secured more than $150 million in tax breaks and incentives spread over the course of 25 years.
Other emerging hotspots include Central Ohio, parts of Nebraska like Sarpy County, and the Dallas-Fort Worth area of Texas, where lower costs and various incentives attract tech companies looking to expand their AI empires. For example, Amazon has leveraged Virginia’s retail sales and use tax exemptions for data center equipment, while Google’s Project Mica in Kansas received $10 billion in revenue bonds to support its sprawling 500-acre campus.
Frequently Asked Questions
What does a data center do?
A data center is a facility packed with thousands of computers that store, process and distribute large volumes of digital information that keep the internet running. AI data centers take that a step further, running massive networks of GPUs to train and deploy artificial intelligence models that power everything from chatbots and virtual assistants to autonomous vehicles.
What is the biggest data center in the world?
Spanning more than 10.7 million square feet, the China Telecom Data Center in Hohhot, Inner Mongolia, is widely recognized as the largest data center in the world. The $3 billion facility houses six massive data halls, consumes 150 megawatts of power and supports global operations for major companies like Alibaba, Tencent and Baidu.
What is the most expensive data center in the United States?
Coming in at $20 billion, the most expensive U.S. data center currently planned is Tract Data Centers’ 2,000-acre campus in Buckeye, Arizona, though future projects could surpass it once construction begins.
