Datacentre REIT economics

Some of the oldest and largest owners of datacentres are real estate investment trusts. They own the buildings, the power connections and the cooling, and rent them to the companies that fill them with computers. They do not own the GPUs. That single fact shapes their economics: they are exposed to demand for space and power, and to the cost of capital, but not to the risk that a generation of chips loses its value.

This piece explains how a datacentre REIT earns money, the measures it reports, how it funds new building, and what the AI build-out changes. Our coverage includes three: Equinix and Digital Realty, which file with the SEC, and Keppel DC REIT, listed in Singapore, which publishes an annual report.

What a REIT is

A real estate investment trust is a company that owns income-producing property and, in return for paying out most of its taxable income as dividends, pays little or no corporate tax on it. The rules differ by country, but the consequence is the same everywhere: a REIT keeps little of its cash. Growth has to be funded from outside, by issuing shares, borrowing, selling assets or bringing in partners. A REIT's cost of capital is therefore as important to its growth as the demand for its buildings.

Two business models

Datacentre REITs earn money in two broad ways, and most do some of both.

• Retail colocation and interconnection. Many customers each rent a cage or a few cabinets, and pay for power and for cross-connects, the direct cables that link them to carriers, cloud providers and one another. The more customers in a building, the more valuable each connection becomes, so interconnection revenue tends to be high-margin and loyal. Equinix built its business on this model.
• Wholesale and hyperscale leasing. A few large tenants each lease a whole hall or building, for many years, often fitted to their own specification. The margin per megawatt is lower, but the scale is large and the leases are long. Digital Realty spans both wholesale leasing and colocation.

Keppel DC REIT, sponsored by the Keppel group, owns datacentres in Asia-Pacific and Europe.

How the revenue is built

A datacentre REIT's revenue starts with rent, usually priced per kilowatt per month or per cabinet, and power, which is either billed on use or included. Leases usually carry annual escalators. When a lease ends, the REIT either renews it, re-lets the space or loses the tenant, and the difference between the old rent and the new one, the renewal spread, is watched closely because it shows whether pricing power is rising or falling. Churn is the share of revenue lost when customers leave. Signed leases that have not yet started form a backlog that shows the revenue already secured for the years ahead.

The measures they report

Because real estate is depreciated in the accounts even when its value holds or rises, net income understates what a REIT earns in cash. REITs therefore report funds from operations, or FFO: net income with property depreciation added back and gains or losses on property sales taken out. Adjusted funds from operations, or AFFO, goes further and subtracts the recurring capital spending needed to keep the buildings competitive, along with other non-cash items. AFFO per share, and how much of it is paid out as dividends, is the measure most investors anchor on.

Alongside those come operating measures: occupancy or utilisation of the space, growth in the same buildings from one year to the next, the size and pre-leasing of the development pipeline, and leverage, usually net debt divided by earnings before interest, tax, depreciation and amortisation.

The economics of building

A REIT creates value when it builds at a yield on cost, the stabilised operating income divided by the cost of building, comfortably above its cost of capital. The spread between the two is the margin on development. Securing power, pre-leasing the space before it is built and keeping construction on time all protect that spread.

To fund large hyperscale projects without issuing too many shares, REITs increasingly bring in private capital through joint ventures that own a share of a development, and they recycle capital by selling stakes in finished, leased buildings and reinvesting the proceeds.

Where AI changes the model

The AI build-out has changed the demand for datacentres in three ways. The scale is larger: Microsoft, Alphabet, Amazon and Meta paid

$201.9B

SEC EDGAR 10-Q/10-K (XBRL): purchases of property and equipment, cash-flow statement (calc.)As of 30 Oct 2026Source

(as of 30 Oct 2026) for property and equipment in the latest quarter, and part of their capacity is leased rather than built. The specification is different: AI halls draw far more power per rack and need liquid cooling, as our piece on power and cooling explains. And power has become the scarce input, so land with an energised connection is worth more than ever.

Those changes cut both ways for a REIT. A portfolio with power available can lease it quickly and on good terms. But air-cooled halls built for cloud tenants may need costly refits to take the newest racks, the largest tenants concentrate the risk, and REITs now compete for AI demand with power landlords and neoclouds that build their own halls.

Compared with neoclouds and power landlords

It helps to place the three kinds of owner on one line. A datacentre REIT owns diversified buildings with many tenants, on long leases, often funded with investment-grade debt, and carries no GPU risk. A power landlord owns energised shells leased to a few operators, again without the GPUs, but with more concentration. A neocloud owns or leases the building and owns the GPUs, and carries the risk of both. The returns, and the volatility, rise along that line.

Sources

• Equinix and Digital Realty: annual and quarterly reports (10-K and 10-Q, SEC EDGAR).
• Keppel DC REIT: annual report.
• Microsoft, Alphabet, Amazon and Meta: quarterly and annual reports (10-Q and 10-K, SEC EDGAR), purchases of property and equipment; how we compute it.
• AI Infrastructure Primer 2026, Compute Terminal: the chapters "The value chain" and "Financing the buildout".