NextFin News - The U.S. data-center boom is hitting a familiar but increasingly expensive constraint: labor. As artificial-intelligence buildouts push power demand, land use and cooling requirements higher, developers are finding that the biggest delay risk is often the workforce needed to wire, commission and energize the buildings themselves. The market is still expanding, but the gap between demand and delivery is widening.
CBRE’s Q1 2026 Global Data Center Trends report said vacancy in the top four U.S. data center markets fell to an all-time low even as inventory rose 33% year over year. Northern Virginia, the largest of the four, posted vacancy of 0.3%; Atlanta was at 1%; Dallas-Fort Worth at 1.8%; and Chicago at 2.2%. In the same period, Goldman Sachs Research said U.S. data center power demand is expected to climb from 31 gigawatts in 2025 to 41 gigawatts in 2026 and 66 gigawatts in 2027, with total capacity rising to roughly 95 gigawatts by the end of 2027.
Those numbers frame the labor issue. A data center is not a conventional building that can be handed over once the shell is complete. It requires high-voltage electrical work, mechanical systems, controls, testing, cooling, commissioning and utility coordination before it can generate revenue. Each project therefore depends on a highly specialized set of trades, and those trades are now being pulled in multiple directions at once by data centers, utilities and other infrastructure projects.
The pressure is showing up at the same time as the industry’s pipeline is getting larger and more power-hungry. Goldman Sachs Research said yearly U.S. capacity additions are expected to reach 13.6 gigawatts in 2026 and 36.3 gigawatts in 2027, compared with realized additions of 6.4 gigawatts in 2024 and 8.5 gigawatts in 2025. That is a steep acceleration in a short period, which makes execution capacity a central issue rather than a side effect.
The latest CBRE data suggests the market is already operating with little cushion. Inventory across the four largest North American data center markets increased 33% in Q1 2026, but demand was strong enough to keep vacancy near record lows. CBRE also said global power availability and grid infrastructure constraints are affecting development timelines and site selection. In practice, that means a project can be delayed by any one of several linked bottlenecks: utility access, permitting, labor or equipment lead times.
The labor constraint matters because it compounds those other limits. If a project has land and financing but lacks electricians or commissioning crews, it still cannot turn demand into operating capacity. If power is available but there are not enough skilled workers to install the systems that connect the site to that power, the facility remains a stranded asset. That is why the current cycle is less about headline demand and more about the ability to convert demand into delivered megawatts.
The workforce challenge also reflects how specialized the current wave of construction has become. AI-oriented facilities are larger, denser and more power intensive than earlier-generation data centers. That raises the number of trades required per project and increases the importance of synchronization across electrical, mechanical and testing teams. In a labor-tight market, those coordination requirements become a source of delay and cost inflation.
In that sense, the labor crunch is not a separate story from the data-center boom. It is part of the same story. The more the industry grows, the more it needs scarce technical workers, and the more those workers become a constraint on growth itself.
Vacancy Is Tight, Power Demand Is Rising, And Delivery Is Becoming Harder
The first important point is that the sector is not short of demand. It is short of deliverable capacity. CBRE said the top four U.S. markets — Northern Virginia, Atlanta, Dallas-Fort Worth and Chicago — all posted very low vacancy in Q1 2026 despite a 33% year-over-year increase in inventory. That is a clear sign that new supply is being absorbed quickly, not sitting idle.
Goldman Sachs Research’s forecast reinforces the point. It said U.S. data center power demand should rise from 31 gigawatts in 2025 to 41 gigawatts in 2026 and 66 gigawatts in 2027. It also projected that the share of total peak summer U.S. power demand consumed by data centers would rise from 4.1% in 2025 to 5.3% in 2026 and 8.5% in 2027. Those figures imply that the sector will be a larger part of the electricity system almost immediately.
That matters because the labor shortage does not operate in isolation. When power demand is rising so fast, every delay in construction has a larger opportunity cost. A contractor who cannot staff a project on time does not just miss a schedule milestone. It can miss a leasing window, a utility interconnection slot or a broader market moment when capacity is most valuable.
The result is a market where growth remains strong but less elastic. Developers can still announce new campuses and operators can still sign large deals, but the path from announcement to operation is getting longer. That makes skilled labor one of the industry’s most important hidden constraints.
The Real Bottleneck Is Specialized Work, Not Generic Headcount
The second point is that data centers need a narrow mix of workers. They require electricians, HVAC technicians, controls specialists, testers, commissioning teams and project managers who understand mission-critical environments. That is a very different labor profile from ordinary commercial construction.
Because of that specialization, the industry can have nominally enough workers in the broader construction market while still lacking the exact trades it needs most. A shortage in one critical role can delay the handoff of an entire building. In data centers, that is more damaging than in many other property types because the asset does not earn its return until the full power and cooling stack is in service.
The problem is also regional. The same workers are being courted by data-center developers, utility companies and transmission contractors. When those groups compete for the same technical labor pool, wages rise and delivery schedules stretch. The market is therefore not just dealing with a shortage of people; it is dealing with a shortage of the right people in the right places at the right time.
CBRE said global power availability and grid infrastructure constraints are influencing both timelines and site selection. That is important because it means labor shortages are arriving alongside other execution hurdles. A developer that wins land but loses labor may still miss the delivery window. A developer that wins labor but lacks utility access may still not finish on time. The bottlenecks are linked.
“Global power availability and grid infrastructure constraints are impacting development timelines and site selection,” CBRE said in its Q1 2026 Global Data Center Trends report.
That sentence captures the environment better than any broad growth narrative. The sector is still expanding, but the physical path to expansion is becoming harder to traverse.
Why This Cycle Is Harder To Execute Than The Last One
The third point is that today’s data centers are more complex than the ones built during the first cloud wave. Earlier facilities could often be delivered at smaller scale with less demanding cooling and electrical requirements. The current AI cycle is different. It requires more power, generates more heat and demands more sophisticated infrastructure on every project.
Goldman Sachs Research’s capacity forecast shows how quickly the scale is changing. It said U.S. data center capacity is expected to reach roughly 95 gigawatts by the end of 2027, more than doubling the level at the end of 2025. It also said yearly capacity additions are scheduled to accelerate to 13.6 gigawatts in 2026 and 36.3 gigawatts in 2027. Those are not incremental changes. They imply a construction market under sustained pressure for a prolonged period.
That matters because labor supply does not scale as fast as demand when the jobs are specialized. Training electricians, commissioning teams and controls technicians takes time. Certification and experience matter, and those cannot be manufactured quickly. So even when pay rates rise, the supply response is gradual.
For contractors, that creates a difficult combination: higher demand, tighter schedules and a thinner labor pool. For developers, it means the cost of missing a delivery target rises. For operators, it means a larger share of the backlog may be trapped in work-in-progress rather than translated into active megawatts.
Goldman Sachs Research said U.S. data center power demand is expected to climb from 31 gigawatts in 2025 to 41 gigawatts in 2026 and 66 gigawatts in 2027.
The jump from 31 gigawatts to 66 gigawatts in two years is the kind of growth that exposes every weak link in the construction chain. Labor is one of those links, and it is often the one that becomes visible only after other constraints have already been priced in.
Who Gains, Who Feels The Squeeze
The biggest beneficiaries are likely to be the developers and contractors that secured power, land and labor early. Scarcity tends to reward firms that already control the hardest parts of the build process. Those with established trade relationships, large project teams and utility access can continue to convert demand into delivery.
The most exposed players are late movers and speculative builders that must assemble every part of the project in a tighter market. If they need to source labor, secure power and coordinate equipment all at once, delays are more likely and costs are harder to control. The market is not short of projects. It is short of execution capacity.
For the broader market, the implication is straightforward. Data-center demand can stay strong even while the supply response slows. That can keep existing capacity valuable, but it can also increase the gap between announced expansion and actual delivered megawatts. In a sector this constrained, the timing of delivery matters almost as much as the size of the pipeline.
The next test is whether contractors, utilities and operators can add enough skilled labor to keep pace with the power buildout. If they cannot, the bottleneck will not stop the AI infrastructure cycle, but it will make the cycle more expensive, more delayed and more selective about which projects get finished first.
Explore more exclusive insights at nextfin.ai.
