Eight primary North American markets had 7,481 MW of data center capacity under construction at the end of June 2026, a record, and 80.4 percent of it was already preleased. That leaves under 1,500 MW open for preleasing, which CBRE puts at roughly six months of demand at the current absorption rate. Vacancy across those markets is 1.4 percent. JLL’s midyear report reads the same way from a wider lens: vacancy near 1 percent for a third year, available space limited to small fragmented blocks, and most tenants committing today for deliveries in 2028.
A buyer who needs usable capacity in 2027 is competing for the unleased fifth of a pipeline that is spoken for as it is built, plus whatever existing operators and owners can free up. This guide sets out where that capacity still comes from, which markets are effectively closed for a 2027 start, and how to screen each route before a term sheet.
Read the arithmetic before the brochures
The six-month figure is the one to carry into planning. It describes the primary markets as a group, so a specific requirement can be easier or harder to place than the average suggests, but it sets the tone of every negotiation: the operator has more buyers than blocks.
Two other numbers from the same CBRE data belong in the model. Asking rates rose across every deployment size in the first half of 2026, led by an 8.3 percent increase for requirements between 3 MW and 10 MW. And minimum utilization requirements in new leases now run from about 60 percent to 85 percent of allocated capacity. A proposal for 2027 delivery should be read with both in mind. The rate is rising, and so is the share of contracted capacity a buyer pays for whether or not it is drawing power.
Separate the markets that are closed for 2027 from the ones still open
Market reports describe conditions at a point in time, and the picture by mid 2026 was uneven.
Northern Virginia had committed every megawatt scheduled to deliver during 2026, with preleasing reaching into 2027 and beyond. Vacancy was 0.2 percent, and available capacity was 10.8 MW in a market with roughly 4.5 GW of inventory. Chicago had no contiguous block of 5 MW or more available during 2026, and the next large contiguous projects were not expected until late 2027 at the earliest. In Atlanta, build-to-suit projects were securing tenants about two and a half years ahead of delivery. Dallas-Fort Worth had more than 765 MW under construction, about 95 percent of it preleased, and Hillsboro’s 170 MW under construction was more than 90 percent preleased. Figures are from CBRE’s H1 2026 report, as summarized by Data Center Frontier.
JLL counts more than 66 GW under construction across North America, and 77 percent of it is being built in frontier markets rather than the established hubs. Put the two reports together. The conclusion for a 2027 operating date is plain: a requirement of 5 MW or more in a primary market is, with few exceptions, a build-to-suit or a secondary-market conversation. Smaller requirements still find pods and suites in existing buildings, at the rates above, and often outside the top eight markets.
Screen the five routes that still produce 2027 capacity
Each route carries a different risk. The first question to ask differs with it, and so does the trade-off a buyer ends up pricing.
| Route | Where it exists | What to verify first | What to price |
|---|---|---|---|
| Unleased blocks in the current pipeline | Later phases in secondary and frontier markets; a small number of primary-market phases delivering late in 2027 | The utility date and the conditions attached to it; whether the phase has its own energization milestone | Delivery risk sits with construction, so remedies matter as much as rate |
| Existing buildings with power headroom | Operators with pods coming off contract; buildings energized below their utility allocation | Energized capacity today versus allocated capacity; cooling limits for the intended density | Density and cooling may cap the deployment; retrofit scope and timing |
| Phased build-to-suit or powered shell | Sites with a first phase energized in 2027 and later phases following | Which phase the 2027 date refers to; what the buyer takes on for fit-out | Longer term and larger commitment; the buyer carries more of the schedule |
| Enterprise-owned facilities with spare capacity | Corporate buildings running below design load, available through sublease, partnership or sale | Rights to transact under the ownership or lease; utility service terms and maintenance history | Older electrical and mechanical plant; the owner’s exit terms |
| Re-marketed capacity | Blocks returned when a tenant re-scopes a commitment | Why it came back; whether power and cooling match the original design basis | Short decision windows; the operator may hold out for a larger taker |
Routes two and four are the ones most buyers skip, and they are the ones most likely to produce a 2027 date in a primary market. The colocation, retrofit, or development comparison covers how to set one brief that all five routes can answer.
Price the delay alongside the kilowatt
A late start has a cost that rarely appears in the rate comparison, and it is rarely small. It includes the workload that waits, the interim capacity bought to cover it, and any renewal or exit terms the delay triggers elsewhere. Build that number first. Then keep it next to the rate when the proposals arrive, because a cheaper kilowatt that lands a quarter late is often the more expensive one.
Then put the same four questions to every provider. What milestone defines ready? What evidence supports that date? How does the ramp schedule sit against the minimum utilization floor? What happens commercially if the date moves? The answers belong in the term sheet, and the remedies belong in the contract. Ask counsel to assess the actual agreement rather than a summary of it.
Run the evidence screen before the term sheet
A 2027 date is only as good as the documents behind it. Dated utility correspondence, study results, phase schedules and commissioning plans say more than a marketing deck, and the power-delivery evidence checklist sets out what to request for each question. The Data Center Power Evidence Worksheet is the same checklist in a form a team can fill in.
Sort what comes back into documented, reported, assumed, estimated and missing, then build one dependency schedule that works backward from the required operating date. A proposal whose date is reported rather than documented is incomplete, which is a different finding from unreliable. Record which documents would need to arrive for that conclusion to change, and when the provider expects to have them.
Prepare for a site screen
For a capacity inquiry with a 2027 operating date, provide the acceptable locations or markets, the usable IT load at the first milestone and at full deployment, rack density and cooling needs, the earliest and latest operating dates, any flexibility to phase, and which of the five routes are acceptable. Identify which provider documents are already in hand; confidential records can be handled through an agreed process.
Explore data center capacity sourcing and the Site Screen. If an acquisition is one of the routes under consideration, review the acquisition advisory service.
Read the evidence and editorial methodology for the distinction between documented, reported, assumed, estimated, and missing information.