The Economics of Owning
To understand why a hyperscaler would sink billions into concrete it could rent, you have to look at the full cost of a data center over its life, not the sticker price of a lease. Owning wins on economics for one core reason: the operator captures the developer's margin instead of paying it, and it does so across a twenty-year asset life.
Break the cost of a facility into its parts. The single largest capital line is not the building shell at all; it is the electrical and mechanical fit-out, the transformers, switchgear, uninterruptible power supplies, generators, and cooling plant that turn a warehouse into a data center. The shell and land are a minority of total cost. This matters because when an operator leases a wholesale shell, it very often still owns and installs its own power and cooling gear inside. So the lease-versus-own distinction is frequently about who owns the building envelope and the land, while the expensive electrical infrastructure is the tenant's either way.
When an operator owns the whole stack, the recurring cost is essentially depreciation plus the electricity bill plus maintenance staff. Depreciation on a self-owned facility runs over a long life, and operators have been extending assumed server and facility lives, which lowers the annual charge. There is no external landlord earning a spread. When the same operator leases, it pays a rent set to give the developer an acceptable yield on the developer's invested capital, a return that stacks on top of the underlying construction cost. Over fifteen years, that stacked return is real money, often a double-digit percentage premium on the effective cost per megawatt.
Owning also hands the operator three forms of control that have their own economic value. First, power: an owner negotiates directly with the utility or independent power producer, can co-locate generation, and can size the interconnection for future expansion rather than today's need. Second, design: an owner can deploy its own cooling architecture, whether that is direct-to-chip liquid cooling for dense accelerator racks or an evaporative approach tuned to the local climate, without a landlord's constraints. Third, expansion: an owner that bought a large parcel can add halls as demand grows, capturing the option value of adjacent land it already controls. A leaseholder must renegotiate or relocate.
There is a scale threshold beneath which none of this pays. Building requires fixed organizational capacity, land acquisition teams, power-procurement specialists, construction and commissioning staff, that only amortizes across a large, steady program. A company adding a few megawatts a year cannot justify that overhead and is right to lease or colocate. A company adding gigawatts a year absolutely can, and for it the self-build economics are compelling. This is why the very largest operators are the most build-heavy: their volume clears the threshold with room to spare.
Standardization compounds the advantage above that threshold. A company that builds continuously develops a repeatable reference design, the same hall geometry, the same electrical topology, the same cooling approach, deployed campus after campus. That repetition drives down engineering cost, shortens construction schedules through learning-curve effects, and lets the operator negotiate directly with equipment makers for transformers and switchgear at volumes a one-off leaseholder cannot match. Owning the design also means owning the roadmap: when a new generation of dense accelerator hardware arrives needing direct liquid cooling, the self-builder can revise its standard and roll it out, while a leaseholder is constrained by what its landlord will permit. So the ownership economics are not static; they widen as the program scales and the design matures, which is another reason the biggest operators keep tilting toward building.
Capital structure is the counterweight. Owning ties up enormous capital in illiquid, single-purpose assets. Even a company with a fortress balance sheet has a cost of capital and competing uses for cash, so it weighs the roughly ten-percent-and-up lifetime saving of owning against the flexibility of keeping that capital free. This is exactly why off-balance-sheet joint ventures have proliferated: they let an operator capture most of the ownership economics, control and long-run cost, while an infrastructure partner supplies the capital and absorbs the construction debt. The operator gets the economics of owning without the full balance-sheet hit, which is why you should read a large build-to-suit JV as an owned build wearing a lease's clothing.