Money & Finance › Markets of the world › Commercial real estate markets: office, industrial, retail, apartments
Data centers: the AI-driven building boom
AI boom since 2023· Asset or instrument
Fact, law or dataA rule, statute, definition or measured number.
Buildings packed with computer servers, power and cooling, now the fastest-growing property type because of artificial-intelligence demand.
See it on the Money & Finance map →How it works
Data centers are sized in megawatts of power, not square feet. Main tenants are 'hyperscalers' (Microsoft, Amazon, Google and Meta) plus AI firms, often on 10–15-year leases. In CBRE's eight primary North American markets, vacancy held at a record-low 1.4% in mid-2026, with 7,481 megawatts under construction, about 80% preleased. Northern Virginia is the largest hub. Power supply, transformers and local approvals are the main limits. US investors reach the sector through data-center REITs such as Equinix and Digital Realty.
Why it matters
They tie the AI stock boom to physical real estate, power grids and debt markets; their scale now shapes utility rates and electricity planning.
Facts, rules and data
CBRE and JLL data center reports, Big Tech capital-spending disclosures, UNCTAD investment data, and utility interconnection filings.
Debates and risks
Whether AI demand will justify the building wave or leave overbuilt capacity if AI spending slows, and who would bear those losses, is debated.
Connections (3)
▸Data centers: the AI-driven building boom is the physical side of The AI stock boom and Big Tech concentrationMainstream view
The connection
AI companies' stock gains rest on spending huge sums on chips, and those chips need buildings, power and cooling. Microsoft, Alphabet, Amazon and Meta spent hundreds of billions of dollars on capital projects in 2025, much of it on data centers. Each dollar of that spending becomes construction, leases and electricity demand.
When and why
Since ChatGPT's launch in late 2022; major clusters in Northern Virginia, Texas, Arizona, Ohio and Louisiana.
How solid is the link?
Company earnings reports and capital-spending guidance, CBRE construction data, and utility load forecasts.
Debates and risks
Whether AI revenue will justify the spending; if growth disappoints, stock prices and building plans could fall together.
▸Data centers: the AI-driven building boom are increasingly funded by Private credit fundsFact, law or data
The connection
Data centers cost billions each, more than tech companies want on their own balance sheets. Private credit funds and insurers lend against the long leases instead. In October 2025 Meta and Blue Owl formed a joint venture for the Hyperion campus in Louisiana, about $27 billion of costs funded largely by privately placed bonds bought by PIMCO and others.
When and why
US, 2024–26; the Hyperion deal, signed October 2025, was described as the largest private-credit transaction on record.
How we know
Meta and Blue Owl announcements, S&P's A+ rating of the bonds, and coverage in Bloomberg and The Wall Street Journal.
Debates and risks
Whether keeping this debt off tech balance sheets hides risk, and what lenders face if AI tenants leave early, is debated; Meta guaranteed a minimum value.
▸Data centers: the AI-driven building boom made up over a fifth of 2025's greenfield Foreign direct investment (FDI) vs. portfolio flowsFact, law or data
The connection
When a company builds a new facility abroad, that is 'greenfield' foreign direct investment. In 2025 tech firms announced data centers worldwide, from Europe to the Gulf and Asia. UNCTAD found data centers made up more than a fifth of all announced greenfield project value, over $270 billion, even as project counts fell overall.
When and why
Worldwide in 2025; reported in UNCTAD's Global Investment Trends Monitor of January 2026.
How we know
UNCTAD greenfield data, built from announced projects in the fDi Markets database; announcements are plans, not completed spending.
Debates and risks
How many announced projects get built, given power shortages and the risk that AI demand forecasts prove too high.