Data Center Markets: Why Location Decides the Rent
Published rents per kW per month differ by a factor of three between markets. What drives the gap, what CBRE's figures show, and why a project's location is an investment variable rather than a detail.
Two data centers with identical engineering can earn very different rents depending on where they stand. Published asking rates differ by roughly a factor of three between the cheapest and most expensive markets CBRE tracks. If you are assessing a project, its location is not background colour — it is one of the two or three numbers that decide the outcome.
What the published figures show
CBRE put the global weighted average asking rate at $217.30 per kW per month in Q1 2025, up 3.3% year over year on a weighted inventory basis, and the global weighted average vacancy rate at 6.6%, down 2.1 percentage points year over year (CBRE, Global Data Center Trends 2025, published 24 June 2025).
Underneath that average, the same report gives these market ranges:
| Market | Asking rate, per kW per month | Note |
|---|---|---|
| Singapore | $310–470 | The most expensive CBRE tracks; a long-standing moratorium on new capacity |
| Tokyo | $190–355 | Slightly down from $200–370 a year earlier |
| London | $180–215 | Up from $160–195 a year earlier |
| Sydney | $140–215 | Declined over the year |
| Phoenix | $190 | Stable |
Two things to take from this. First, the spread is enormous — a project's location can matter more to its revenue than its construction cost does. Second, the direction is not uniform: London rose while Tokyo and Sydney softened, in the same quarter, in the same global shortage. "Data center rents are rising" is true on average and false in particular places.
North America is the tightest, not the dearest
CBRE's North American numbers describe scarcity rather than high prices: vacancy at an all-time low 1.6%, and 74.3% of capacity under construction already preleased. Northern Virginia alone recorded 538.6 MW of net absorption and an 80% surge in under-construction capacity (CBRE, North America Data Center Trends H1 2025). CBRE's outlook for the region describes pricing holding "comparable to 2011–2012 levels at $200+ per kilowatt (kW) per month".
That last phrase deserves attention. It says today's pricing is back at levels last seen fifteen years ago — a recovery from a long period of oversupply, not an unprecedented boom. It is a useful corrective to marketing that presents current rents as a permanent new normal.
What actually drives the gap
Available power. The first-order variable. A market where a utility can deliver tens of megawatts within a planning horizon has supply; one where it cannot has scarcity and higher rents. See why power is the constraint.
Planning and political consent. Singapore's pricing reflects deliberate restriction of new capacity. Several busy markets have become politically contested, and an objection is as effective as a transformer shortage.
Network density. Interconnection-rich markets — where many networks meet — command a premium for connectivity rather than for space, which is why a hub city prices above an industrial site with equal power.
Latency to users. Inference serving users needs to be near them. Training does not, which is why training capacity can be sited where power is cheap and land is available, and why rural sites are plausible for one workload and not the other.
Power price and climate. Where the tenant pays for electricity, the local tariff affects what it will pay in rent. Cooler climates allow more free cooling and lower operating costs.
Tax treatment. Several US states offer sales and use tax exemptions on data center equipment, which can move a siting decision on its own.
Primary and secondary markets
Primary markets — the established hubs — have the deepest tenant demand, the most competition for sites, and the figures above. A project here competes with sophisticated operators who hold land and utility relationships.
Secondary and greenfield markets offer cheaper land and sometimes faster power, which is exactly why new projects go there. The trade is tenant demand: hyperscalers lease where their networks and their people already are. Preleasing of 74.3% describes credible capacity in established markets. It does not describe a greenfield site with no tenant.
For an investor, this is the practical test: if the project is in a secondary market, what makes a tenant choose it? Cheap power is a real answer. Fiber to a nearby hub is a partial one. "The market is short of capacity" is not an answer, because the shortage is concentrated where the tenants already are.
Using this when you read a deal
- Which market, precisely? A county, not a region.
- What rate does the model assume, and how does it sit against the published figures for a comparable market? Remember these are asking rates, averaged across deployment sizes, and large wholesale leases commonly price below small retail requirements — a sanity check, not a benchmark.
- Why would a tenant come here rather than to a primary market?
- What is the power position, and how does the local utility's record on large loads look?
- Is latency relevant to the workload the sponsor expects to host?
Next: how data center leases work for what the rate actually buys, and the risks for what happens when the tenant does not arrive.