TMIAn electrical contractor outside Charlotte wins a subcontract on a 200MW data center build. Serious work, serious money - $14M scope, 18-month timeline. Their foreman has 22 years in commercial construction, knows how to run a crew, knows how to hit a schedule. He shows up to the first site walk and spends the next two hours realizing that almost nothing on this site works like anything he has done before. The voltage levels are different. The redundancy requirements are different. The commissioning process is different. The tolerance for a one-hour delay is different - on a hospital project you can manage around a half-day slip; on a live data center campus where the owner has hyperscaler SLAs, a half-day slip in mechanical or electrical work can trigger penalty clauses that wipe out three months of margin.
That contractor is not incompetent. Their foreman is not incompetent. They are experienced professionals who stepped into a work environment that was purpose-built to different specifications than anything their experience had prepared them for, and they found out the hard way rather than the planned way. They are not alone. Power generation and AI infrastructure is the fastest-growing project segment in construction right now - WSP lists it as a primary growth driver in their Q1 results, Turner and Gilbane are chasing the same work - and the crews building it were mostly trained for something else.
What makes this work different
The hazard profile on a data center or power generation project is not harder than commercial construction. It is different in ways that matter, and different in ways that generic safety training does not cover.
High-density electrical systems run at voltages and amperage levels that are uncommon on standard commercial work. The arc flash risk on a 480V bus feeding a dense server pod is not the same as arc flash risk on a typical commercial distribution panel, and the PPE requirements, clearance distances, and lockout procedures are calibrated differently. A journeyman electrician who has never worked above a certain voltage tier can pass every safety certification on the books and still not know what they do not know when they step into a high-voltage data center switchgear room for the first time.
Cooling systems on hyperscale builds are another departure from standard commercial work. These are not conventional HVAC systems - they are precision cooling infrastructure running continuous loads, with redundancy built in at every level, and a failure mode that can cascade from a leaking pipe fitting into a rack shutdown in under four minutes. The mechanical contractor who knows commercial chiller work cold can still create a serious problem if they do not understand how the data center's DCIM system monitors and responds to cooling anomalies, and why a pressure test procedure that is perfectly appropriate on a standard commercial loop can trigger an automatic shutdown on a mission-critical system.
And then there is commissioning. On a standard commercial project, commissioning is a final phase. On a data center build, commissioning is a continuous parallel process that starts before the structural work is done and never really stops. Crews are working in live-adjacent environments from early in the build. The tolerance for a tool left in a cable tray, a fitting left slightly loose, a temporary power connection that was correct for the construction phase but not labeled for the operations team - all of these have downstream consequences that are larger and faster than on any other project type your crews have worked before.
The operators winning data center work in 2027 are already solving the crew problem in 2026. Everyone else is going to learn it the expensive way.
What Construction Safety Week got wrong
Construction Safety Week 2026 had a real theme this year - communication, multilingual safety, getting information to workers in the language they actually speak. That is a legitimate problem and the industry needed to address it. But it is a universal problem, not a data center problem, and the event messaging mostly treated all construction jobsites as equivalent environments with equivalent hazard profiles.
Turner's new jobsite communication app and Gilbane's OSHA partnership both came out of the same week, both framed around generic safety communication. Neither addressed what is different about the work that both companies are actively growing into. You do not solve a data center safety problem by translating the same safety message into more languages. You solve it by having crews who understand why this environment operates differently - and by having systems that capture and enforce those differences at the task level, not just at the orientation level.
The industry's safety infrastructure was built around the hazard profile of traditional construction. A lot of it is excellent. None of it was designed for a world where 40% of new large-scale construction starts are data centers and power generation projects that run on fundamentally different operating assumptions. The gap between the safety training most crews carry and the environment they are walking into is not a gap you close with an app.
Building a crew that can actually do this work
The contractors who are moving into data center and power generation work successfully are not waiting for the industry to build a training standard around it. They are building their own. The pattern looks roughly the same across the ones doing it well: identify the five or six crew members who have touched this type of work before, in any capacity, and build the competency base from them outward. One electrician who spent two years on a colocation facility build is worth more to you right now than ten certifications.
The second thing they do is treat the pre-construction walkthrough differently. On a standard commercial project, the foreman walks the site, reviews the drawings, notes the constraints. On a data center project, the pre-construction process needs to include a conversation with the commissioning manager about how their process interacts with the construction schedule, which systems will be live while other work is ongoing, and what the owner's escalation path looks like when something goes wrong during active commissioning. That conversation is not happening on most bids. The contractors who have it before they sign the contract are the ones who can actually price the risk correctly.
When the operational system is built to capture crew certifications and project-specific qualifications at the task assignment level - not just at hiring - the foreman does not have to hold all of this in their head. The system surfaces the right crew for the environment, flags when someone is scheduled for a task that requires a qualification they do not carry, and builds the institutional knowledge about what this type of work actually requires before the next bid goes out. The contractors who figure that out in 2026 will be the ones writing the scopes of work for data center projects in 2028, because they will be the ones with the track record to prove they can do it without killing the owner's uptime SLA.
The market is not waiting. WSP's Q1 numbers show where the money is moving. The question is whether your crews are moving with it or watching it from the sideline while someone with better-prepared people takes the work.