Ask a colocation broker in Northern Virginia how much space they have left for a 10 MW deployment landing in the next six months, and the honest answer is usually some version of “not here, and maybe not in the market you want.” According to CBRE’s North America Data Center Trends report, vacancy across the primary North American markets fell to a record low of 1.4% in the first half of 2026 — even as total power supply grew 33.7% year-over-year to 10,903 MW and a record 7,481 MW was under construction. More supply, less availability: that’s the paradox buyers are now negotiating against, and it is quietly resetting how far ahead a capacity decision has to be made.
The scarcity math, in plain numbers
The headline number is vacancy, but the number that should worry a buyer more is preleasing. CBRE found that 80.4% of all capacity currently under construction is already preleased, up from 74.3% a year earlier. Run the subtraction: of that record 7,481 MW being built, less than 1,500 MW is actually available to a buyer who hasn’t already signed. Net absorption came in at 1,456.2 MW for the period, up 11.7% year-over-year, which means the market is on pace to absorb essentially everything still unspoken for almost as fast as it’s built. “It is unlike any other industry in my lifetime, where you can triple supply, yet still have vacancy rates going down,” said Pat Lynch, CBRE’s global executive managing director for the data solutions group, in the firm’s August 2026 release of the report. That is not a sign of a market cooling off — it’s a sign that supply and demand are now moving on the same curve, and buyers who assume construction will eventually catch up to demand are planning against a trend that isn’t holding.
Why “buy capacity now” has become “buy capacity for 2029”
Gordon Dolven, CBRE’s head of Americas data center research, has been blunt with clients about what this means operationally: start leasing conversations roughly three years ahead of when the capacity is actually needed. That’s a dramatic shift from the 12-to-18-month lead times that were standard even three years ago, and it changes who should be in the room for a capacity decision. A real estate or procurement team negotiating a lease the way it negotiated office space — shopping multiple markets, comparing term sheets, deciding close to the need date — is structurally late in 2026. Dolven’s own framing of where the market is headed is worth sitting with: “The next phase of growth will be defined less by demand and more by where new supply can realistically come online.” For a buyer, that reframes the entire question from “which provider offers the best rate” to “which markets can physically deliver power and shell space on my timeline at all,” which is a site-selection and power-queue question as much as a commercial one — closely related to the time-to-power dynamics already reshaping site selection.
The market-by-market picture isn’t uniform — and that’s the opportunity
Northern Virginia, still the largest data center market in the world, is effectively sold out: CBRE put vacancy there at roughly 0.2%, pushing new deals into 2027 commitments by default. Chicago isn’t much better, with 2.2% vacancy and no contiguous block of 5 MW or larger available anywhere in the market through the rest of 2026. But capacity is not disappearing — it’s relocating. Atlanta overtook Northern Virginia in construction volume, with roughly 2,882 MW under construction, up 52.3% year-over-year, while Dallas-Fort Worth has about 765 MW under construction and is already around 95% preleased. Pricing is following the same divergence: CBRE recorded a 19% year-over-year rate increase for 10 MW-plus deals in the New York-New Jersey market and 14.5% in Atlanta, while 3-10 MW deals — the size band most enterprise and mid-market buyers actually need — saw the steepest increase of any category at 8.3%. The takeaway for a buyer isn’t that there’s no capacity; it’s that the capacity that exists is increasingly in secondary markets, priced higher than a year ago, and gone quickly once it’s listed. CBRE’s report also flagged that this pricing divergence is compounding: because the smaller 3-10 MW band is seeing the fastest rate growth, mid-market enterprise buyers — the ones least likely to have a dedicated real estate team tracking quarterly vacancy data — are absorbing the steepest cost increases in the market, often without realizing it until a renewal or expansion quote comes back well above what they budgeted.
Need Expert Guidance?
Talk to a Data Center Expert
21+ years of hands-on experience in data center design, operations & infrastructure. Book a quick discovery call to discuss your project.
📞 Book a Discovery CallA counter-signal worth weighing before you panic-sign
It would be a mistake to read the vacancy numbers as proof that every gigawatt in the pipeline is a sure thing. Wood Mackenzie’s tracking shows newly announced data center capacity actually fell to 36 GW in the first quarter of 2026, down 19% quarter-over-quarter and the third consecutive quarterly decline in new announcements. At the same time, capacity sitting in “advanced discussions” jumped from 37 GW to 107 GW in that same quarter — which reads less like demand cooling and more like the funnel maturing, with speculative announcements giving way to deals that are actually getting signed. Separately, Currence’s February 2026 analysis flagged that 30 to 50% of the roughly 16 GW slated for 2026 delivery is at risk of slipping into later years, versus a 26% slip rate in 2025, though SemiAnalysis has pushed back that this estimate is too pessimistic. The honest summary for a buyer: don’t assume every announced megawatt will land on schedule, but also don’t assume a slowdown in headline announcements means more room to negotiate — the deals that matter are moving to signature faster, not slower, which is consistent with the build-versus-buy tradeoffs explored in how the build-buy-colocate math has shifted in 2026.
What this means if you’re the one buying
Three practical shifts follow from this data. First, treat lead time as a hard constraint, not a negotiating lever: if Northern Virginia and Chicago are functionally sold out and secondary markets are preleasing at 80%-plus before completion, a request for proposal sent six months before you need power is already behind the market, not ahead of it. Second, widen the market search before you widen the budget — Atlanta’s and Dallas-Fort Worth’s construction growth suggests secondary and tertiary markets will keep absorbing demand that primary markets can no longer physically host, and locking in there earlier is cheaper than waiting for primary-market space that may not reopen. Third, build contract flexibility for schedule risk into the deal itself, given the real spread between slip-rate estimates; a capacity agreement that assumes a fixed delivery date without remedies for delay is underwriting someone else’s construction risk. Buyers who move on this data now, rather than after the next vacancy report confirms it, are the ones who will actually get the megawatts they need on the timeline their business requires.
None of this means the market is frozen or that every deal requires panic. It means the planning horizon itself has changed, and the cost of discovering that late is no longer a minor scheduling headache — it’s a 2027 or 2028 go-live date for a workload that was supposed to launch next year. Treat the current vacancy and preleasing numbers as a planning input the same way you’d treat a chip lead time or a construction cost index: something to check quarterly, not something to assume has stayed the same since the last time you signed a lease. For a structured way to work through site selection, power timelines, and contract terms before you’re negotiating from a position of scarcity, see the TechInfraHub Data Center Buyer’s Toolkit.
Written by
Raajeev Ratra
Data Center Infrastructure Expert | 15+ Years in DC Design, Operations & Project Management
Raajeev is a seasoned data center professional with hands-on experience in hyperscale facilities, colocation design, power & cooling infrastructure, and global DC operations. He shares practical insights to help engineers and IT leaders build better infrastructure.