Data center growth is outrunning the people who run it

The data center buildout has been underwritten against shells, switchgear and megawatts, the items on a capex line and in a lender's model; the figure that may actually govern delivery is smaller and more awkward: headcount. DCD Intelligence's new workforce survey finds more than two-thirds of developers and operators staffing below what their operations require, with nearly a third running at less than 80 percent of demand.

The shortfalls run through every critical function the survey examined, and they fall hardest where uptime is decided: facilities and physical infrastructure, power and cooling systems, and IT operations. DCD warns that gaps in those areas risk reduced performance and increased downtime—the finding that matters most for an asset class whose leases and debt are written against availability. Financing a shell is by now a well-understood exercise; staffing one is not, and the two do not run on the same clock.

Recruitment is where the stress starts: nearly half of respondents name a lack of technical knowledge among prospective hires as their biggest barrier, and the report also flags low awareness of the industry as a drag on the pipeline. Together those findings describe a labor market that does not produce candidates the industry recognizes and does not attract enough people to become them—a problem capital cannot solve by paying more for the same finite set of engineers.

The survey's more revealing half is where firms are sourcing: organizations are recruiting increasingly from adjacent technical infrastructure and utilities, with telecoms cross-hires ranking third. The skills driving the fastest workforce expansion are fluid dynamics and power, which sends recruiters into power and water utilities and the oil and gas industry. That is the bench the grid buildout is bidding for too, and data centers are now competing for power and cooling engineers against the utilities whose interconnection queues they depend on.

Power rights have, as this publication has argued, become a distinct asset class in which the queue, the permit and the connection trade before the electron does; the survey prices the other half of that trade. A queue position is worth only what a developer can convert into energized, cooled, running capacity, and conversion runs through the skills DCD locates as the sharpest shortage, power and cooling first among them.

DCD's prescription is hard to argue with and hard to execute: structured training, standardized onboarding, upskilling on AI and whatever else arrives next. The operators best positioned to capture the growth, the report concludes, are the ones that treat workforce development as core infrastructure—a word that carries weight, because infrastructure is what a firm builds ahead of demand, at multi-year lead times, on the theory that the capacity will be needed. The survey finds organizations expanding education-based pathways toward that end through internships, apprenticeships and university programs.

The economics still push the other way: a laterally hired engineer closes a gap this quarter and appears in the org chart, while an apprenticeship pays back over years that may outlast the current development cycle, and the firm that funds one cannot stop a competitor from hiring the graduate. That asymmetry explains why this kind of report tends to be agreed with in public and deferred in private, and why the shortage DCD describes is likelier to widen than close. The report says as much, warning that without a shift, skill gaps and recruitment shortages continue and could widen as the pace of technological change accelerates—which makes the AI upskilling line both the easiest to endorse and the hardest to deliver.

The same capital hierarchy that governs which projects get built probably governs who can afford to fix this: hyperscaler-anchored assets command infrastructure pricing while everything else fights for capital, and that is a prediction rather than a finding, but it follows from the structure. An operator with long-dated commitments from investment-grade tenants can fund a training pipeline across a cycle, while a merchant developer leasing quarter to quarter has every incentive to buy the engineer rather than grow one—and if the split holds, the labor constraint compounds the capital constraint instead of easing it.

An industry that has spent three years learning that consent, not capital, decides what gets built should recognize the shape of this. Permits and people are both constraints that no amount of debt can clear, and both are settled years before the first rack is energized, by institutions the developer does not control: planning boards in one case, training programs and technical colleges in the other.

Follow that logic into credit and the survey stops reading like a human-resources document. If availability and service commitments are what protect an asset's cash flow, then staffing depth belongs in diligence next to the tenant roster and the power contract; the reasonable inference is that merchant data center debt, where lenders are underwriting the forecast rather than the customer, is missing a line. The next operator that misses an availability commitment because it could not staff a hall will push that question out of the interview and into the covenant.

The sub-80-percent cohort is the line to watch in the next iteration of this survey, though the survey is self-reported and carries no dollar figure for the shortfall, so the magnitude is softer than the direction. If that cohort is no smaller when the next tranche of shell space is energized, the industry will have answered its growth with poaching rather than a pipeline, and the schedule will turn on whoever is available to commission the hall, not the megawatt.