Commercial construction has historically been a business of one-offs. A hospital is different from an airport, and one office tower may bear little resemblance to the next. Designs change, subcontractors change, and site conditions change. The knowledge accumulated on one project does not always transfer cleanly to another.

Data centers begin to break that pattern. Hyperscalers may construct multiple facilities based on similar designs, and contractors can build the same type of infrastructure repeatedly. That makes it possible to take actual information from completed projects, including labor productivity, unit costs, and construction outcomes, and use it to plan the next one.

If a contractor has estimated and delivered five data centers, Mundt explained, the final numbers from those projects can be fed back into its estimating systems, such as Ediphi before bidding the sixth. The contractor begins the next project with a much clearer picture of what work actually costs and how long it takes.

The goal is not simply to make construction cheaper. It is to make it more predictable.

That creates a feedback loop more characteristic of manufacturing than conventional construction. Each facility can inform the next, shortening preconstruction and allowing contractors and their customers to forecast schedules with greater confidence.

The economics of data centers are also changing what contractors are willing to experiment with.

On most commercial projects, a new technology must typically justify itself through lower construction costs. Data centers introduce another variable: the value of time. Bringing a facility online even a few weeks earlier can allow a hyperscaler to deploy expensive computing equipment sooner.

“Working in data centers is a little bit different, because it’s all about speed optimization,” Mundt said.

That changes the calculation around technology. A system does not necessarily have to be cheaper if it can demonstrate that it meaningfully accelerates delivery. Clients can be less concerned with additional cost ” as long as you can prove to them that there’s a benefit from a speed perspective ” according to Mundt.

Robotics is becoming one of the clearest examples.

Layout work has traditionally required workers to move through large spaces with drawings, chalk lines, and tape measures, marking exactly where equipment should be installed. Robotic systems like Rugged Robotics can increasingly ingest a digital drawing and autonomously mark those locations across a data hall.

For a conventional project, the calculation might focus on how many labor hours the robot eliminates. For a data center, the more important question can be how much sooner the facility becomes operational. If work that once took a month can be completed in two weeks, those two weeks have considerable value to an owner trying to bring additional computing capacity online.

Diverge VC is seeing robotics spread into other parts of the jobsite. Nguyen said his teams have worked with robots for layout and are piloting machines that clean, grout, and perform drywall-related work. He believes advances in generative AI could accelerate the transition from machines that execute individual programmed tasks toward systems capable of interpreting schedules and production requirements with considerably less human supervision.

Neither Nguyen nor Mundt expects that evolution to eliminate construction workers anytime soon.

” The golden age for construction robotics is going to be a world in which robots are operating alongside humans, not replacing them,” Mundt said.

The foreman does not disappear because a robot performs layout. Instead, the robot completes the repetitive work while the foreman spends time on tasks requiring greater judgment and coordination.

Nguyen sees much the same dynamic. With contractors already struggling to fill positions, he questions whether a robot performing an unfilled role can meaningfully be described as taking someone’s job.

“There’s not a human there ready for that job,” he said.

There is also a safety argument. Some of the first tasks likely to be automated are precisely those workers would prefer not to perform.

“Who wants to tie rebar as their career?” Nguyen said. “Who wants to drill overhead, 15 feet up in the air?”

For Nguyen, those are natural applications for machines because they combine repetition, physical strain, and risk.

Technology is also being used to address the shortage before a worker reaches the jobsite. Mundt pointed to renewed interest in augmented and virtual reality training as contractors search for ways to compress years of accumulated field knowledge into shorter training cycles. The challenge is not simply getting inexperienced workers productive more quickly. They must do so safely on sites where schedules are already moving at extraordinary speed.

The industry’s labor problem, in other words, is creating pressure at both ends. Contractors need more people entering the trades, while simultaneously finding ways for each worker already in the field to accomplish more.

Prefabrication offers another way around the constraint.

Instead of assembling every component at the construction site, portions of electrical, mechanical, and structural systems can be manufactured in controlled environments and delivered ready for installation. Mundt said trade partners are already moving in this direction, assembling components off-site and delivering them as kits of parts. Data centers, with their relatively repetitive designs, offer a particularly attractive environment for the approach.

Nguyen expects the shift to accelerate. “In the next five years you’re going to see a ton more,” he said of prefabrication and modular construction.

Yet even an increasingly industrialized construction process still encounters bottlenecks that are stubbornly difficult to automate.Commissioning is one area where the construction industry remains ripe for greater automation. Before a building becomes operational, its electrical, mechanical and building systems must be tested, verified and coordinated. The process can be painstaking, particularly on large and technically complex data center projects, creating opportunities for technology to streamline workflows and reduce manual work.

“Commissioning, I mean, that’s such a painful process,” Nguyen said. “Any automation that you can put towards commissioning and workflow in that regard, I think it’s huge.”

Supply chains represent another vulnerability across construction. Projects often depend on large quantities of specialized materials and equipment, and a delay in a single critical component can bring an otherwise fast-moving project to a standstill. The challenge became particularly visible during recent supply-chain disruptions, when equipment such as electrical switchgear became difficult to procure.

“That will debilitate a job or a project,” Nguyen said.

Ultimately, however, both conversations return to a constraint that cannot be solved simply by building faster: electricity.

Power availability is becoming one of the largest constraints facing new data center construction. The scarcity is influencing decisions throughout the development process, from cooling systems and building layouts to the overall efficiency of a facility.

It is also changing the role contractors can play in the development process. Rather than simply executing a set of plans, contractors are increasingly being brought in earlier to help optimize facilities around power constraints, with a focus on more efficient building designs, layouts, modular systems and cooling technologies.

As data center demand continues to grow, efficiency is becoming increasingly important. The measure of a best-in-class facility is shifting beyond speed and scale to include how much computing output it can deliver relative to the space and electricity it consumes. Energy efficiency and computing output per square foot are likely to become increasingly important measures of performance.

Both ultimately depend on access to reliable, affordable power. With grid capacity increasingly constrained in many markets, the pressure is already pushing the industry to think beyond the grid. Mundt expects more discussion around power generation colocated with data centers and believes small modular nuclear reactors could eventually play a role if permitting and deployment challenges can be resolved. The attraction is straightforward. A data center operator able to secure its own reliable generation becomes less dependent on an electrical system that is increasingly struggling to accommodate new loads.

“If you can control your own destiny from an energy creation perspective,” Mundt said, “that’s a huge unlock for a lot of these data centers.”

Taken together, the experiences of Diverge VC and Suffolk Technologies suggest that the ai boom is beginning to change construction itself. Buildings are becoming more repeatable. Historical project data is feeding back into future designs and estimates. Work is moving from jobsites into factories. Robots are beginning to work alongside tradespeople. Training is being compressed. Contractors are being asked to participate earlier in decisions about efficiency and constructability.

The industry’s traditional equation is changing. For decades, contractors competed primarily around cost, schedule, and quality. Hyperscale computing has introduced another measure: how quickly a building can begin producing compute.

The race to build AI infrastructure is therefore becoming more than a race to pour concrete and erect data halls faster. It is pushing construction toward a more industrialized model, one built around repeatability, automation, and continuous learning. Yet even that transformation ultimately runs into the same physical boundary confronting the rest of the AI economy. The industry may learn to build data centers faster than ever before, it still has to find enough people and enough power to make them run.