The global data center industry has crossed a symbolic and economic threshold that helps better understand where digital infrastructure is heading. According to the new 2026 Data Center Marketplace report from Colliers, worldwide investment in data centers exceeded $580 billion in 2025, a 27% increase from the previous year and, for the first time, surpassing the $540 billion allocated to new oil supply. Colliers attributes much of this jump to artificial intelligence pressure, with $445 billion coming solely from tech companies.
That figure is notable not just for its volume. It also reflects a much deeper change: data centers have ceased to behave primarily as relatively predictable tech real estate assets and are increasingly resembling large-scale energy infrastructure—capital-intensive, dependent on electricity availability, and sustained by increasingly aggressive financing cycles. The Colliers report summarizes this as a “structural change,” where business decisions are now driven not only by demand but also by which markets can finance, enable, electrify, and execute projects within realistic timelines.
AI Drives Spending but Also Changes the Rules
At the core of this new cycle is, above all, Artificial Intelligence. The report states that AI workload growth is increasing the average project size, multiplying density requirements, and forcing redesigns of digital campus construction logic. In North America, absorption reached 15.6 GW in 2025—double that of 2024—while over 90% of new capacity was leased before even going into service. In other words, the market isn’t building out of optimism but because demand outpaces real delivery capacity.
However, the same phenomenon fueling investment is also increasing risk. Colliers warns that the average project is already 40% larger than a year ago, with construction costs continuing to rise sharply. The report estimates a more than 45% increase in construction costs in 2025 and notes that 40-50% of many new electrical infrastructure projects’ budgets are allocated to energy components. In other words, energy is no longer just a component of data centers; it has become the dominant factor in their economic viability.
This shift explains why more promoters are now talking about substations, behind-the-meter generation, natural gas, or modular small reactors—almost as much as they used to discuss racks, fiber, or cooling. Competition is increasingly based not just on location but on the certainty of available power.
The Biggest Bottleneck Isn’t Land Anymore—It’s the Grid
One of the most revealing data points from the report is that utilities are tightening their entry conditions. In several major U.S. markets, the deposits required to reserve electrical capacity now range between $25 million and $75 million per project, often non-refundable until connection is actually activated. Colliers also notes that processing windows for these connections have stretched to 12 or 18 months in many cases, and in highly strained areas, timelines may be even longer.
As a result, there’s a geographic shift in demand. Mature markets remain attractive, but some growth is shifting toward locations with more headroom for securing power, permits, and land. The report highlights how the traditional distinction between primary, secondary, and emerging markets is eroding because capital increasingly values execution capacity over historical market labels. For example, Columbus grew 126% year over year and has already surpassed a gigawatt of commissioned capacity, while Dallas-Fort Worth absorbed 1,137 MW in 2025, solidifying its role as a major expansion hub.
Meanwhile, local opposition is intensifying. Colliers reports that over $64 billion worth of data center projects in the U.S. have been delayed or blocked since 2023, and more than 142 lobbying groups across 24 states are actively working to slow or regulate new facilities. Electricity costs, water use, noise, and land consumption are becoming high-profile political issues.
More Debt, More Private Credit, and More Questions About Profitability
The other pillar of this new cycle is financing. Colliers notes that hyperscalers issued over $120 billion in debt in 2025 to support AI infrastructure growth. Concurrently, private credit now accounts for between 60% and 75% of early-stage capital, accelerating projects but also concentrating risk in less transparent and less regulated structures than traditional banking.
This raises warnings. Moody’s recently warned that aggressive spending could lead to overbuilding and weak returns if demand, monetization, or funding do not keep pace with deployment. The agency estimates U.S. hyperscaler spending could reach around $700 billion in 2026—a massive figure that prompts scrutiny not only of growth but also of the financial sustainability of the model.
This doesn’t mean the cycle will abruptly stop. Rather, it signals a nuanced reality: AI is elevating data centers to a category of critical global infrastructure but also making them more expensive, complex, and dependent on energy, regulation, and capital markets. The key question for 2026 isn’t whether there will be demand but which operators can turn their plans into delivered megawatts without drowning in costs, delays, or debt.
In this landscape, data centers no longer compete solely with other real estate or tech assets. They now compete with large energy projects, new industrial chains, and a digital economy that’s increasingly measured not just by users or software but by electricity, land, debt, and execution capacity.
Frequently Asked Questions
How much was invested in data centers in 2025 according to Colliers?
Colliers estimates global investment exceeded $580 billion in 2025—a 27% increase from 2024 and surpassing worldwide spending on new oil supply.
Why is AI fueling the surge in data center investment?
Because training and inference of AI models demand more capacity, higher electrical density, increased cooling, and greater power availability per campus, which raises both project size and costs.
What is the main challenge for building new data centers today?
Electricity availability. Colliers highlights that energy and grid connection timelines now weigh more than location, with utilities demanding deposits of $25 to $75 million per project in several markets.
Is there a risk of overinvestment in AI data centers?
Yes. Moody’s warns that aggressive spending could lead to overcapacity and weak returns if actual demand, monetization, or financing do not match deployment plans.
via: colliers