Data Centers Are the New Warehouses: The Hottest Trade in Real Estate

Every Cycle Gets a Darling

Each real estate cycle anoints one property type as the place institutional money must be. In the 2010s it was the warehouse. Online retail needed logistics space near every major city, vacancy ground down toward 3 percent by 2022, rents compounded, and a category investors once dismissed as boring boxes became the best performing property type of the decade. The 2020s have handed the crown to the data center, the specialized industrial building that houses the servers behind cloud computing and, increasingly, artificial intelligence. The pattern is the same one, a technology shift creating structural demand for a physical shell, and the numbers this time are bigger.

Seven Hundred Billion Dollars of Demand

The demand side is dominated by the hyperscalers, the handful of companies operating cloud platforms at global scale. The six largest, Microsoft, Amazon, Alphabet, Meta, Oracle, and Apple, are projected to spend roughly 700 billion dollars on capital expenditure in 2026, nearly six times what they spent in 2022, and the bulk of the increase is AI infrastructure, buildings, power equipment, and the chips inside. Training and running large AI models requires computing density that old facilities cannot serve, so demand lands on new construction, and 2026 is on track to set another record for leasing. Tenants now routinely commit to buildings years before they exist, a practice called preleasing, which tells you everything about who holds negotiating power.

Vacancy Barely Exists

United States data center vacancy sits around 2 percent, a record low. Northern Virginia, the largest data center market on earth, is effectively full at 0.3 percent, and Atlanta runs near 1 percent. For comparison, a healthy office market was considered tight at 10 percent vacant. Pricing follows scarcity, rents are at all time highs and rising across every major market, and a telling detail has flipped, large tenants taking 10 megawatts or more historically negotiated volume discounts, and today they pay premiums for scale instead, because guaranteed capacity is worth more than a lower rate. When your biggest customers stop asking for discounts, you are no longer in a normal market.

Megawatts, Not Square Feet

The industry does not size deals in square feet, it sizes them in megawatts of power capacity, because electricity, not space, is the scarce input. A grid connection for a large campus can take 24 to 48 months to secure, longer where new transmission lines or generation are needed, and that queue is now the industry bottleneck. Of the roughly 16 gigawatts of United States capacity slated for 2026 delivery, about two thirds had not broken ground as of spring, and analysts expect 150 to 200 billion dollars of planned spending to slip into 2027 and 2028 for exactly this reason. The shift to giant AI campuses of 500 megawatts and up has pushed construction schedules into multi year territory and turned land with secured power, so called powered land, into the industry's most valuable commodity.

How the Landlords Get Paid

Two business models split the industry. The hyperscale or wholesale model leases entire buildings or campuses to a single giant tenant on long term triple net leases, meaning the tenant pays taxes, insurance, and maintenance on top of rent. A fifteen year lease to Microsoft is closer to a corporate bond than a building, and investors price it accordingly, at yields below most other real estate. The colocation model, run by companies like Equinix, rents smaller spaces to many tenants and charges for interconnection between them, more operationally complex, more diversified. Public investors access the sector through REITs like Equinix and Digital Realty, while private giants and infrastructure funds own much of the rest. The richest profits sit in development, building at a high yield on cost and watching the finished asset get valued at a much lower stabilized yield, a spread that has minted the sector's fortunes.

The Bear Case

The risks are real and worth stating plainly. Demand concentration is extreme, a handful of tenants drive most leasing, so the sector's fate is chained to hyperscaler capex budgets, and every capital spending boom in history has eventually overshot, including the warehouse boom, where industrial vacancy roughly doubled off its 2022 lows once the wave of new supply landed. Technology risk cuts uncomfortably, rising chip power densities and liquid cooling keep changing optimal building design, so a facility built for 2020 workloads can age out of the top tier fast. And the speculative fringe is growing, land bought without secured power, capacity built ahead of signed leases. The mitigant for existing assets is the lease structure itself, long commitments from some of the most creditworthy tenants on earth, which parks the real risk in the development pipeline rather than the standing buildings.

The Bottom Line

Data centers are the defining real estate trade of this cycle, near zero vacancy, record rents, 700 billion dollars of annual hyperscaler capex, and a bottleneck that has shifted from money to megawatts. The analogy to warehouses runs in both directions, a technology wave can make a boring building type the best investment of a decade, and every wave eventually crests. For a student of the industry, the two numbers that matter are hyperscaler capital spending and the interconnection queue, demand and supply for the only asset class where the landlord's real product is electricity.