AI factories, more commonly known as data centers, are as expensive as they are expansive. In recent months, a flurry of funding announcements from major tech firms has pledged billions of dollars to fuel new construction. These cash infusions grab headlines, but beyond the sheer scale of these facilities, what infrastructure supports them, and who is actually building them?
Companies like Vantage Data Centers, QTS Data Centers and Equinix specialize in developing the sprawling, often non-descript campuses that house the backbone of AI.
Vantage, for instance, has built 11 campuses across North America in 2024 alone. Its offerings range from “rack ready” spaces – where power infrastructure is in place and clients can simply install their equipment – to “whitespace” options, which give tenants control over power distribution and network connections. For those with more precise specifications, “build-to-suit” centers offer fully customized data infrastructure.
As demand for AI computing power surges, the business of building these digital fortresses is becoming as critical and lucrative – as the technology they support.
Massive arrays of air-conditioning units line rooftops or cluster alongside buildings, working around the clock to keep servers cool. Perimeter security can be imposing – high walls topped with concertina wire, surveillance cameras positioned to capture every angle, and security personnel patrolling the grounds.
Inside, these facilities house the hardware; racks of high-performing servers, storage systems, networking equipment and the electrical infrastructure that keeps everything running. Construction typically begins with electrical and plumbing work before the building shell takes shape. Cooling systems come next, along with the main switchgear linking external power sources to the facility’s internal grid. Nearly all large-scale data centers rely on backup generators to prevent outages.
Once fully operational, a data center can hold tens of thousands of servers, arranged in dense rows with diagnostic stations in between, constantly monitoring temperature, humidity and power flow. Some facilities even have their own water treatment plants as part of their cooling systems, and as energy demands grow, companies are exploring on-site power generation including dedicated substations, beyond existing solar, wind and battery storage.
The demand for more AI data centers is growing fast, according to a recent report from McKinsey & Company titled “AI power: Expanding data center capacity to meet growing demand.”
“Our analysis of current trends suggests that global demand for data center capacity could rise at an annual rate of between 19 and 22% from 2023 to 2030 to reach an annual demand of 171 to 219 gigawatts (GW).”
Revenue, innovative spirit and the desire to lead the world in AI development, particularly to stay in front of China, have motivated U.S.-based investment.
“In Fiscal Year 2025, Microsoft is on track to invest approximately $80 billion to build out AI-enabled datacenters to train AI models and deploy AI and cloud-based applications around the world,” said Brad Smith, Microsoft vice chair and president. “More than half of this total investment will be in the United States, reflecting our commitment to this country and our confidence in the American economy.”
The recent Stargate announcement at the White House, in which President Donald J. Trump introduced the heads of OpenAI, Softbank and Oracle, represents the largest plan to date. Those corporations have pledged to commit $500 billion to build up to 20 AI data centers in the U.S., primarily in Texas. Each center would be approximately a half-million square feet. Half of the buildings are already under construction. Microsoft, NVIDIA and Arm are also contributors.
Google is pledging $3 billion, and Meta is earmarking $1 billion, each in separate endeavors.
The Synergy Research Group, which provides market intelligence and analytics for the networking and telecoms industry, estimates that the number of large data centers operated by hyperscale providers increased to 992 at the end of 2023, and passed the 1,000 mark in early 2024.
“Meanwhile it has taken just four years for the total capacity of hyperscale data centers to double, as the number of facilities grows rapidly and their average capacity continues to climb. Synergy’s data shows that the United States still accounts for 51% of worldwide capacity, measured by MW of critical IT load, with Europe and China each accounting for about a third of the balance,” the report said.
Entrepreneur Kevin O’Leary, aka Mr. Wonderful on the Shark Tank TV show, announced last December a plan for the largest AI data center industrial park in the world, “Wonder Valley,” to be built in Alberta, Canada, through O’Leary Ventures. Mr. O’Leary is the chairman of O’Leary Ventures,
The entire industrial park would be off-grid, that is powered by natural gas and a geothermal power infrastructure. “We will engineer and build a redundant power solution that meets the modern AI compute reliability standards,” Mr. O’Leary said. “The first phase of 1.4 GW will be approximately US$ 2 billion with subsequent annual roll out of redundant power in 1 GW increments. The total investment over the lifetime of the project will be over $70 billion when considering the infrastructure, power, data centers and ancillary structures.”
