Artificial Intelligence is pushing U.S. data center construction into its hyperspace moment. In the Star Wars universe, hyperspace let starships cross the galaxy in hours instead of thousands of years, and AI is driving a similar acceleration in digital infrastructure buildout.
The United States already operates more than 3,000 data centers, and a recent Pew Research Center analysis finds more than 1,500 additional facilities in various stages of development nationwide. Surging demand for AI and machine learning compute is transforming data centers from traditional IT backbones into massive, gigawatt-scale industrial loads.
Deloitte projects that AI data center power demand alone will grow more than thirtyfold by 2035, climbing from roughly 4 gigawatts in 2024 to 123 gigawatts. To put that in perspective, several proposed 50,000-acre data center campuses could consume 5 gigawatts of power, more than the capacity of the largest nuclear and gas plants operating in the United States today.
From Texas and Virginia to Georgia and Pennsylvania, this wave of construction is reshaping the grid, land use, and connectivity demands in ways that have few historical parallels. For data center developers, operators, and their supply chain partners, it is creating unprecedented urgency around scalable power, resilient fiber infrastructure, and construction-ready connectivity solutions.
Data Center Construction Shifts to Rural Areas
Data centers are not only being built at a pace the country has never seen, but construction is shifting from dense metro hubs to rural counties, many of which have never hosted a facility before.
Pew Research found that:
- 67% of planned data centers are in rural areas, while 87% of existing data centers are in urban ones.
- 39% of planned data centers are in counties that currently don’t have any.
- Three-quarters of all planned data centers will be built in the South and Midwest, with the South alone accounting for nearly half (48%).
|
Region |
Existing Data Centers |
Planned Data Centers |
Increase from Current Total |
|
South |
1,209 |
754 |
62% |
|
Midwest |
655 |
419 |
64% |
|
West |
807 |
277 |
34% |
|
Northeast |
397 |
106 |
26% |
Virginia (287) and Texas (170) have the most planned data centers followed by Georgia (141), Illinois (123), and Arizona (86).
McKinsey & Co notes that hyperscalers are increasingly targeting secondary and emerging markets for new data centers as the existing data center hubs become saturated. A recent report from CBRE shows a record-low vacancy rate of 1.9% in primary data center markets, pushing developers into new geographies where power capacity, land availability, and high-performance fiber networks must often be built in parallel with the facilities themselves.
Power Demand Tells the Data Center Story
Perhaps the clearest signal of the current data center construction boom is the power story. U.S. data center power demand is projected to more than double in just two years, from 31 GW in 2025 to 66 GW in 2027, with an intermediate step of 41 GW in 2026, according to Goldman Sachs Research.
Over that same period, data centers’ share of total U.S. peak summer power demand is expected to jump from 4.1% in 2025 to 8.5% in 2027, tightening reserve margins across multiple regional grids and forcing utilities to rethink long-term capacity plans. This is not a gradual uptick; it is an industrial-scale load ramping up on a compressed timeline.
It is no surprise, then, that power has become the dominant constraint on new capacity. Deloitte’s 2025 AI Infrastructure Survey found that 72% of respondents view power and grid capacity as the primary challenge for data center buildout, ahead of supply chain disruptions (65%) and security concerns (64%). In other words, even if developers can secure land, materials, and financing, many projects still hinge on when and where they can get sufficient, reliable power.
The grid bottleneck is already slowing projects down. Deloitte reports that some requests for grid connection are facing wait times of up to seven years, which can push projects well beyond original schedules once permitting, construction, and commissioning are factored in. For operators racing to meet AI-driven demand, these delays are pushing them toward locations with faster interconnections timelines and toward technologies and partners that can make better use of every megawatt they can secure.
A Trillion-Dollar Investment Opportunity
The scale of the current buildout is staggering. McKinsey estimates that global data center infrastructure will require about $6.7 trillion in capital expenditure through 2030, with more than 40 percent of that investment flowing into the United States. Researchers there suggest that AI investment at this magnitude could boost GDP, create thousands of high-paying jobs, and accelerate innovation across multiple sectors.
|
Servers |
$3.5 trillion |
|
Storage |
$800 billion |
|
Electrical and mechanical equipment |
$800 billion |
|
Labor |
$600 billion |
|
Power generation |
$400 billion |
|
Shell and site |
$300 billion |
|
Network infrastructure |
$100 billion |
|
Land acquisition |
$100 billion |
|
Overall Total |
$6.7 trillion |
Here is how McKinsey envisions that spending breaks down:
About 70% of this spending is expected to come from hyperscalers building large campuses that can span hundreds of acres and draw industrial-scale loads, a concentration of capital that is reshaping site selection, power planning, and connectivity requirements across the country.
Texas: The Next Data Center Frontier
Texas is rapidly emerging as the next data center frontier, combining favorable geography, a business‑friendly regulatory environment, abundant land, a strong workforce, and a growing ecosystem of infrastructure partners such as Hexatronic.
A recent Texas Tribune analysis identified at least 248 planned data center projects in the state, underscoring the scale of development now in motion.
ERCOT President and CEO Pablo Vegas has described this as “an unprecedented change in the pace of growth”, noting that the grid operator received 519 requests to connect large electricity users over the past two years, compared with just 24 such requests the year before. Roughly 90% of those new requests are for data centers, most targeting an operational date of 2030 or earlier, which puts additional pressure on already tight power and transmission timelines.
Deloitte estimates that Texas now has more data center capacity under construction and planned (3.1 GW) than its current installed capacity (2.9 GW), reflecting how quickly AI and cloud demand are reshaping the state’s load outlook. While Dallas is already an established primary market, Houston and the Austin–San Antonio corridor have emerged as high‑growth secondary and emerging markets, attracting hyperscalers and large colocation providers looking for scale, connectivity, and speed to power.
Critically, Texas data center projects are not confined to metro cores. Of the 582 operating and planned facilities across the state, the Texas Tribune reports that nearly half are in suburban (214) and rural (62) areas, mirroring the national shift away from saturated urban markets and toward greenfield sites with room for multi-campus development.
Building the Backbone of the AI Era
For data center developers, operators, and their supply chain partners, this wave of construction is creating urgent demand for scalable power, resilient fiber infrastructure, and construction-ready connectivity solutions.
Every one of the 1,500-plus facilities now in development needs to be connected, and the push into rural and emerging markets means that connectivity infrastructure often has to be built from the ground up alongside the facilities themselves. That is where Hexatronic comes in.
From high-density hyperscale campuses in the Dallas-Fort Worth corridor to greenfield builds in rural Texas and beyond, Hexatronic delivers the fiber infrastructure solutions that keep the AI era moving at the speed of hyperspace.
Contact Hexatronic today to learn how our fiber connectivity systems and advanced services can support your next data center buildout.