Compare the U.S. colocation markets that can serve you.

How to choose the right market

The right market is the one that meets the workload's latency, connectivity, risk, staffing, compliance, and budget requirements with a credible delivery path. Market rank alone should not decide it.

1. Map users, applications, cloud regions, partners, and network dependencies.
2. Define required power, rack density, cooling, and growth by date.
3. Set disaster-recovery distance, climate, utility, and operational risk requirements.
4. Compare existing capacity separately from planned or utility-dependent capacity.
5. Normalize facility, network, migration, support, escalator, and renewal costs.
6. Run at least two markets in parallel when the delivery date or power block is critical.

The four largest markets side by side

Market choice is usually a bigger lever than provider choice. These are the published figures that distinguish them, on one basis, so the comparison is real rather than impressionistic. The emerging four are not in this table because published research does not cover them on the same basis, and inventing the cells would defeat the purpose.

Sources: CBRE Global Data Center Trends 2026, CBRE on Northern Virginia, CBRE on Atlanta and the CBRE Dallas-Fort Worth profile. Rents are per kW per month for a 250 to 500 kW requirement. Where CBRE publishes no numeric figure for a market, the cell says so rather than carrying an estimate. Power and queue figures are market reported and vary by substation. Full sourcing on the pricing benchmarks page.

Which market fits which requirement

If more than one row applies to you, the requirement is probably two deployments rather than one, and splitting network-facing capacity from bulk compute is often cheaper than forcing both into a single market.

The rest of the markets we cover

Seven markets sit outside the two groups above, either because published research does not cover them on the same basis or because they answer a narrower question. Each has a full guide, with the facilities we track and, where a collector reaches them, the public filing record.

Markets we cover on one layer of data rather than four

Twelve more markets have a full guide, and they carry a caveat the fifteen above do not. No permit collector reaches them and our facility inventory does not extend to them, so what you get is the interconnection record, the announced capacity we could verify in its own source, and the regulatory position. That is thinner, and saying so is the point: five of these twelve are markets where a jurisdiction has stopped taking data center applications.

Three markets we cover on the announced record alone

These three are newer than most of our collection. No permit collector is pointed at them, the facility inventory does not extend to them, and their cities are not yet on the interconnection list, so they are counted nowhere rather than counted wrongly. What each guide does carry is the announced capacity, verified the same way as everywhere else, and the thing that differs most between the three and is least visible in a quote: how electricity is actually bought. One has no retail supplier to shop, one buys from a federal corporation through a municipal utility, and one splits the bill between a regulated wires company and a competitive supplier. That single difference changes what an operator in each is able to promise you.

Three markets that arrive on the permit record

These three carry three of our four collections: the public construction record, the interconnection register and the announced capacity, each verified the same way as everywhere else. What they do not carry is our facility inventory, which does not extend to them, so none of them shows a building count and none shows a zero standing in for one. They also have something in common that matters more than any figure below. In all three, the largest number attached to the market belongs to something nobody can lease, and the market a buyer can actually reach is the smaller one underneath it.

Three more markets, and what each of them is actually short of

These arrived on 3 September 2026 and they are deliberately uneven. One is an interconnection market with no construction, one is a construction market with no interconnection, and one has almost nothing a buyer can lease at all. None of them carries our facility inventory, so none shows a building count and none shows a zero standing in for one. Reading what a market lacks is usually more useful than reading what it has, so each card below leads with the gap.

What no market comparison can tell you

Market-level figures set expectations. They do not decide a deployment, and treating them as if they do produces confident plans that fail at the facility level.

• Vacancy is not availability for your requirement. A 2.4% vacancy market may still have nothing that fits 40 kW racks with liquid cooling on your date.
• Asking rents are not your quote. They are a published band for one requirement size, before density, term, ramp and redundancy.
• Pipeline is not inventory. Roughly 80% of North American construction is preleased before completion.
• Market power cost is not your power cost. Tariffs and pass-through mechanisms vary by substation and by contract.
• A market queue is not your queue. A provider with energized capacity is unaffected by the market average, which is exactly why the distinction matters.

Use these figures to choose which two or three markets to test, then verify at the facility level. See how we verify capacity for what that confirmation involves.