The Capacity That Isn't Actually Capacity
A headline vacancy rate is a lot like a restaurant's posted wait time. It tells you something true about the building. It tells you almost nothing about whether the specific table you actually need is available. I think the data center market just produced the clearest example of that gap I have seen all year, and it did not come from a report. It came from one advisor's actual phone calls.
What James Mercer Actually Found
James Mercer, principal of Metro Colo Advisory, recently searched for 5 megawatts of capacity on behalf of a client requiring 140 to 160 kilowatts per rack, with direct liquid cooling. He contacted more than ten North American operators. Nearly every one told him they had space. Almost none of them could actually support the density and cooling his client's workload required. Mercer agreed to be quoted on the record about the search, though he declined to name the specific operators, citing ongoing channel relationships with them.
JLL's own research backs up what Mercer found the hard way. North American data center vacancy has held at roughly 1 percent for three consecutive years, and according to Andrew Batson, who leads JLL's data center research, only a very small percentage of even that thin 1 percent can actually support high-density AI deployments. A separate industry survey puts real numbers on how narrow that slice is. Eighty-two percent of existing facilities support racks below 30 kilowatts. Only 9 percent can handle 50 or above. Mercer's client needed nearly three times that ceiling.
Why This Is LENS² and Not Just a Tight Market
I built the LENS² framework specifically to answer a question PACTS was never designed to ask. PACTS tells you whether a site can work, generation, transmission, approvals, connectivity, thermal capacity, schedule. A vacancy rate is essentially a PACTS-level number, a rough proxy for whether physical square footage and stated capacity exist somewhere in a market. What PACTS does not tell you, and what a 1 percent vacancy figure cannot tell you either, is whether that capacity actually fits this specific client's workload, at this specific density, with this specific cooling requirement.
That is exactly where the Workload times Uptime multiplier at the end of LENS² does its real work. A client running standard enterprise workloads at 10 kilowatts a rack reads a 1 percent vacancy market completely differently than a client running dense AI training at 150 kilowatts a rack. The first client is competing for a genuinely scarce resource. The second client is competing for a resource so scarce it barely shows up as a rounding error inside that same 1 percent figure. Both clients are technically shopping in the same market. They are not remotely shopping in the same market in any way that matters.
Why the Construction Pipeline Doesn't Fix This Either
More than 66 gigawatts of capacity is currently under construction across North America, which sounds like relief on the way. Seventy-seven percent of that pipeline sits in frontier markets, not necessarily the markets or the density profiles a specific client actually needs, and most tenants securing space right now are contracting for deliveries in 2028. A client with an urgent, high-density requirement today is not meaningfully helped by capacity that will not exist for two more years, in a market that may not even be the right one.
Why This Failure Mode Is So Easy to Miss From the Outside
Nobody publishes a headline number for real, usable, density-matched capacity. Vacancy rates are easy to report because they are a single clean figure. The actual answer to whether capacity fits a specific client's workload requires calling more than ten operators individually, the way Mercer did, because no aggregated market statistic captures it. That is precisely the failure mode LENS² exists to catch before a client burns weeks discovering it the hard way, one phone call at a time.
What This Means for Anyone Searching for Capacity Right Now
Do not let a market's headline vacancy rate stand in for an actual answer about whether usable capacity exists for your specific workload. Run the client's real density and uptime requirements against a market's actual, verified, density-matched availability before starting a search, not after ten calls have already confirmed what the vacancy number never could have told you.
If you want a specific client's requirements run against a market's real capacity before your own search burns the weeks this one did, I would rather do that work with you now.
If that's where you are, whether you're the one running the search or bringing this to a board that's about to trust a headline number, grab 15 minutes on my calendar, and I'll walk through what I would actually want to verify
PLUS:
Grab the guide. Before advancing a data center site, there are 12 questions I make sure I can answer. I wrote them up here: [The 12 Questions Every Real Estate Professional Should Ask Before Advancing a Data Center Site].
