Global investment in data centers surpassed $580 billion in 2025, a 27 percent rise year-on-year and the first time data center spend has exceeded investment in new oil supply, a new report from Colliers has found.
According to the report, Facilitating AI with Unprecedented Infrastructure, this record number was driven by AI investment, with $445bn coming from technology companies alone.
Colliers said this represented a “structural shift” under which data centers have become one of the world’s most “capital-intensive infrastructure categories.”
Forecasts from multiple independent firms predict global data center investment in 2026 will either remain at $580bn or surpass it entirely, with US spend alone reaching $500bn.
However, Colliers said the industry’s development will be “increasingly defined by which markets can finance, permit, power, and execute projects on realistic timelines,” as development economics have reset around power certainty and infrastructure delivery.
Several independent researchers are forecasting $3 trillion in investment by 2030 to meet projected global data center expansion. However, the sheer magnitude of investment required to fund the sector is bringing serious capital concerns, as fears of slow returns and considerable investment risk manifest.
Hyperscalers are raising significant funding for the data center buildout, but are, in many cases, unable to build fast enough. According to Colliers, hyperscalers issued more than $120bn in debt in 2025, more than five times the five-year average. However, more than $64bn of US projects have been delayed or cancelled due to power shortages, permitting delays, and supply bottlenecks.
Colliers also raised the issue of a significant power crisis underlying all of this, as utilities are now demanding $25-75 million in, often non-refundable, deposits per project. In Northern Virginia, new grid connections can take up to seven years.
As a result, power infrastructure is eating around 40-50 percent of total project budgets, Colliers said.
Debt risk
In the report, Colliers also examined the debt risk facing investors in the AI buildout. According to the IEEE, looking at more than 1,300 tech sector firms, total interest-bearing credit outstanding has reached around $1.3trn, with more than $1trn of this focused on over a dozen big tech AI-focused companies.
Private credit is now funding between 60-75 percent of early-stage development capital, Colliers said, which is concentrating exposure to powered land acquisition and early-stage infrastructure risk within private markets.
Colliers explained that, while this is accelerating project timelines and enabling rapid scaling, it is introducing new layers of risk.
“Capital is being deployed through opaque, illiquid structures with limited regulatory oversight, concentrating early stage development exposure within private funds rather than distributing it across regulated banking channels,” the report said, warning of “dislocation” across private credit markets and the broader capital stack if refinancing conditions tighten or demand expectations reset.
There are considerable fears that private sector investment is becoming more exposed to risk with new and sustained investment in the data center buildout.
Earlier this month, a report from ratings agency Moody’s suggested that aggressive spending in the sector could lead to overbuild and weak returns.
The ratings agency said that, given the capital-intensive nature of AI and the requirement for upfront investment before revenue is realized, a data center will take between 12 and 24 months between initial spending and revenue generation.
Moody’s also warned that higher capital intensity and debt levels for hyperscalers could lead to a “reassessment of creditworthiness” if profit growth fails to materialize.
Comments