A building can be structurally complete. The steel is up. The skin is on. The switchgear is installed. The generators are set. The cooling systems are piped and charged. The controls are wired. From the outside, it looks like a data center. From the inside, it looks like one too. Walk the floor and you see racks, power feeds, cooling lines, cable trays — everything in its place.

‍

And then you try to turn it on.

‍

Commissioning is the process of proving that a data center actually works. Not that it looks right. Not that it was built according to the drawings. That it performs — under load, under stress, under failure conditions — the way it was designed to perform. It is the most critical phase of the entire construction lifecycle, and it is the one the industry consistently underestimates, under-resources, and treats as an afterthought.

‍

The consequences show up in ways that are invisible to everyone except the people who have to live with them. A facility that was not properly commissioned does not announce itself as a problem. It reveals itself slowly — through nuisance alarms, through unexpected switchover behaviors, through cooling asymmetries that surface only at full load, through control sequences that work in isolation but conflict with each other under real operating conditions. The tenant discovers these issues over months. Each one is individually fixable. Collectively, they represent a facility that was never truly validated.

‍

Why Most Commissioning Is Theater

‍

There is a version of commissioning that is common in the industry and that does almost nothing useful. It looks like a series of checkbox exercises performed by teams that are not qualified to evaluate what they are testing, under conditions that do not represent real operation, documented in reports that nobody reads until something goes wrong.

‍

This version exists because commissioning is treated as a schedule item rather than an engineering discipline. It gets compressed when the construction timeline slips — and the construction timeline always slips. By the time the building is ready for commissioning, the project is already behind schedule, the tenant is demanding a delivery date, and the pressure to "get through Cx" as fast as possible is immense. Testing gets abbreviated. Failure scenarios get simulated at partial load instead of full load. Documentation gets filled out after the fact based on what should have happened rather than what was observed.

‍

The people performing this abbreviated commissioning may be perfectly competent construction professionals. But commissioning is not construction. It is a separate discipline that requires a different skill set — the ability to understand how multiple complex systems interact under dynamic conditions, to identify when a test result looks right but is not right, and to make the call that a system is not ready for operation even when every stakeholder in the room wants to hear that it is.

‍

The qualified commissioning professionals who can do this work — really do it, not just check the boxes — number in the low thousands in the United States. Against a pipeline of hundreds of large-scale projects, the arithmetic is not encouraging. The demand for commissioning talent outstrips the supply by a factor that grows every quarter.

‍

What Happens When You Skip It

‍

The consequences of inadequate commissioning do not manifest as a single dramatic failure. They manifest as chronic underperformance.

‍

A cooling system that was not tested at full load under peak ambient conditions may deliver its rated capacity at 72 degrees ambient but derate at 95 degrees. The tenant discovers this during a heat wave, when the facility is at peak utilization and the cooling system cannot maintain the required supply temperature. The options at that point are bad: shed load (turn off GPUs), temporarily augment the cooling (expensive and disruptive), or accept elevated temperatures (which degrades GPU performance and shortens equipment life). None of these options were part of the plan. All of them are expensive.

‍

An electrical distribution system that was not tested through its full failure cascade may perform flawlessly under normal conditions but behave unexpectedly when a real fault occurs. A generator that starts and synchronizes correctly in a standalone test may not transition properly when the transfer switch has to coordinate with other generators, utility feeds, and load management systems simultaneously. The individual components all work. The system does not, because the system was never tested as a system.

‍

A controls platform that was not thoroughly validated may operate correctly under steady-state conditions but generate false alarms, incorrect switchover commands, or conflicting control actions under transient conditions. These issues are maddening to diagnose after the fact because the root cause is not a failed component — it is an interaction between components that was never exercised during commissioning.

‍

Every one of these scenarios is preventable. Not easily, not cheaply, but preventable. They are prevented by commissioning processes that test the actual operating conditions the facility will face, performed by teams that know what they are looking for, documented with enough rigor to be useful when something eventually needs to be diagnosed.

‍

The Talent Problem Behind the Process Problem

‍

The commissioning workforce challenge is distinct from the broader construction labor shortage, though they overlap. The electricians and mechanical contractors needed to build a data center are in short supply across the economy. But they are at least a category of worker that the training infrastructure understands how to produce. Apprenticeship programs, trade schools, and union training centers have been developing these skills for decades.

‍

Commissioning engineers are different. There is no standard pipeline. Most of the experienced commissioning professionals in the industry learned their trade through years of practice on increasingly complex projects, under the mentorship of other experienced professionals. It is an apprenticeship-based discipline in practice, even if it does not carry that label formally.

‍

The result is that the supply of commissioning talent grows slowly — one person at a time, through years of supervised experience — while the demand is growing exponentially with the number of facilities entering the construction pipeline. You cannot train a commissioning engineer in six months any more than you can train a surgeon in six months. The knowledge base is too broad, the judgment calls are too consequential, and the experience base required to recognize subtle problems is too deep to shortcut.

‍

This creates a structural bottleneck that will not resolve quickly. The industry can build facilities faster. It can procure equipment on shorter lead times. It can compress construction schedules through modular methods and parallel workstreams. But it cannot compress the development of the human judgment required to validate that the finished product actually works. That takes time, and there is no technological substitute for it.

‍

What the Best Look Like

‍

The projects that get commissioning right share a few characteristics that have nothing to do with budget and everything to do with philosophy.

‍

They start commissioning planning at the design phase, not the end of construction. The commissioning team reviews the design documents and identifies testing requirements before the building is built, not after. This means the building is designed to be testable — with isolation points, measurement access, and control capabilities that facilitate comprehensive validation.

‍

They maintain continuity between the commissioning team and the construction team. The people who will test the systems understand how they were built, because they were present during construction. They saw the decisions that were made in the field, the substitutions, the change orders, the compromises. They know where the bodies are buried, and they know what needs extra scrutiny.

‍

They test at full load and under real failure conditions. Not partial load. Not simulated failures. Actual full load, with actual utility disconnections, actual generator transfers, actual cooling system switchovers. These tests are expensive. They consume time. They sometimes reveal problems that require rework. But they produce a facility that the tenant can trust — and trust, in this market, is the asset that compounds.

‍

They document everything — not to satisfy a compliance requirement, but to create an operational record that the facility's team can use for the life of the building. A commissioning report that sits in a file cabinet is worthless. A commissioning record that informs how the operations team manages the facility, plans maintenance, and troubleshoots issues is an asset that pays dividends for decades.

‍

The Market Will Sort This

‍

There are going to be a lot of data centers built over the next five years. The capital is committed. The demand is real. Some of these facilities will work beautifully from day one because they were commissioned by teams that knew what they were doing. Others will look beautiful and perform poorly because the commissioning was treated as a formality.

‍

The market will learn to tell the difference. Tenants will share intelligence. Reputation will matter. The developers and operators who invest in rigorous commissioning will build a track record that earns them the best tenants and the best terms. The ones who cut corners will spend their margin on remediation and their credibility on excuses.

‍

Commissioning is not the glamorous phase. It does not have the drama of deal-making or the scale of construction. But it is the phase where the building either earns its right to operate or reveals that it was never ready. Every building gets commissioned one way or another. The only question is whether it happens before the tenant moves in, or after.