Paul: There is a lot of discussion these days about data centers, particularly AI data centers. It is similar to the rail build-out many years ago and the telecom build-out. This next build-out the world is undergoing is related to AI cloud computing.
Today, we are going to talk about what is involved in getting all of these AI data centers online and operational so they can provide value to the world. We have a great discussion today, and I’ll turn it over to David to introduce the group.
David: Thank you, Paul. It is a privilege, as always, to be here. I am very happy to support ICXA’s initiatives, particularly this one on data centers.
This is a key initiative for ICXA’s technical committees: joining perspectives from around the world. That is precisely what we are trying to do in the technical committees that I have the privilege to chair. We want to interconnect knowledge that is currently fragmented across geographies and solve problems together as a global community.
We try to solve problems through different lenses, with permanent contributions from professionals around the world. This allows us not only to advance knowledge and research in the profession, but also to develop world-class best practices and global standards. Those standards help us better understand the risks ahead of us in the transition to operations and make sure assets are sustainable, safe, and profitable for organizations.
As you mentioned, Paul, today we are discussing a highly relevant topic for the global execution ecosystem. Data centers are expected to account for approximately $800 billion in cumulative capital expenditure by the end of 2026. The main contributors are Alphabet, Microsoft, Meta, and Amazon, contributing around $600 billion.
This is growing. By 2030, we are expecting approximately $7 trillion in cumulative investment, with around $1.7 trillion to $1.8 trillion per year from that point until we reach the projected figure from PwC of $30 trillion cumulative by 2050.
This is a large and highly relevant industry. With the AI supercycle we are living through right now, this seems to be the direction of capital expenditure across the globe.
To discuss this today, we have two experts joining us from Latin America and Europe. Rodrigo joins us from Costa Rica, and Steve Hedley joins us from the UK.
Let’s take a moment to introduce yourselves. Steve, can you give us a little background on what you do and your expertise in data centers?
Global Perspectives on Data Center Commissioning Hurdles
Steve: It is a pleasure to be on such a great panel. My name is Steve Hedley. I am a global commissioning director. I have been operating in the data center space for about six and a half years, and I have delivered data center projects across the UK, Europe, India, and Japan.
I am currently working for an owner-operator. We build and operate our own data centers in Northern Europe, the Nordics, and the UK.
David: Excellent. Good afternoon to you. Rodrigo, you are in Costa Rica.
Rodrigo: Good morning. I am from Costa Rica. My background is in electrical engineering, and I have been working in the data center industry for the last 18 years. In the last six years, I have focused more on commissioning in the data center industry.
David: Excellent. Thank you both.
Without further delay, we will start with the first question. We will begin with regional perspectives and the challenges you are seeing. From your region, what are the biggest challenges today in commissioning data centers and successfully transitioning them from construction to live operations? What makes these challenges different in your region, country, or market? Let’s start with you, Steve.
Steve: It is probably fairly consistent globally, but certainly in the markets where I am currently operating across Europe, the Nordics, and the UK, tight timelines are a major challenge. That is the same across most industries. In my previous life in oil and gas, it was the same.
One of the biggest things is demand. Projects are springing up everywhere. Everyone is struggling with skilled resources: good mechanical resources, electrical resources, BMS resources, and the ability to line all of that up across design, construction, pre-commissioning, commissioning, and the transition to operations.
Grid availability and energy costs are also huge issues at the moment. Demand for data centers and the power associated with them continues to rise, and that puts a lot of strain on local infrastructure, power supplies, grid power, and backup power. Those are probably the biggest issues we are seeing.
If we go deeper, we also start looking at design changes. Liquid cooling is a major topic now. It has been significant in the United States for some time, and I have spoken with quite a few people in Asia who are doing it. In Europe, I would say it is still relatively in its infancy.
That comes back to resources. There are many people globally who are familiar with liquid cooling, commissioning, handover, and operation, but we are starting to see that transition into Europe now. That is a major area of focus.
David: Particularly with the regulatory schemes in Europe and the initiatives to get to net zero by mid-century, that adds another layer. Absolutely.
Rodrigo, tell us about Costa Rica and the broader region. What challenges are you seeing?
Rodrigo: More than Costa Rica specifically, I think the challenge for Latin America is that we have two different types of clients.
First, we have companies whose core business is not data centers. These are companies with enterprise data centers, and many of them do not understand commissioning. It is very difficult to help this type of client understand the importance and the full process of commissioning.
On the other hand, we have clients whose business is data centers. They understand commissioning, but, as Steve mentioned, the biggest problem is that during construction, project teams sometimes do not account for the time needed to complete a proper commissioning process according to standards.
Construction companies sometimes see commissioning as a barrier to construction progress, especially because it affects the time needed to develop and deliver the project. Many people think commissioning is only the testing that happens during construction. They do not always understand the full process: the documentation, the staged approach, and everything that must happen before the test stage.
David: Thank you. Coming from Latin America myself, I would add that the social and political environment does not always support a smooth flow of information across regions. Paul, I’ll hand it to you for the second question.
From Construction to True Data Center Operation
Paul: Rodrigo, you mentioned several challenges, including interfaces and the integration of systems into functional, performing assets. In your view, which construction-to-commissioning interfaces create the most problems during the transition to operations? Let’s start with you, Rodrigo.
Rodrigo: I think controls are the most constrained area. In many industries, teams only verify that individual equipment works by itself. For example, if you test an electrical generator, you only test that the generator starts when the power goes out.
In the data center industry, there are many interfaces between systems that must be tested together. It is not usual in other industries for the cooling team to have to work closely with the electrical team to test and prove that systems work together as designed.
You also have failure scenarios. You have to test all systems in the event that something fails. This is part of the process that regular industrial teams from other sectors may not know or understand. They may not understand that we have to test between systems. For me, that is the most complicated part of commissioning a data center.
The failure modes you observe can be totally different from what you predicted during design or even during simulation. One of the key things I see is the level of improvisation. I have had the unfortunate experience of witnessing what can go wrong in projects when improvisation appears. In some projects, fortunately not here in Canada but in other parts of the world, I have seen unfortunate fatalities when improvisation occurs. You are trying to react to a live system without fully understanding how it behaves.
To me, transition to operations is the most critical stage in a project. We talk about value and early engagement, but it is only when you see the transition in action that you realize how important all the front-end actions were and why you should not reduce the effort required to ensure that the transition and residual risks are properly identified. In essence, Paul, this is risk mitigation. What we are doing here is mitigating risk and increasing predictability in operations.
Paul: We would love to get your thoughts on this too, Steve.
Steve: Rodrigo set it up very nicely. It is the interfaces between systems.
If we talk specifically about construction and commissioning, the challenge is getting a clean handover. You may have an electrical room, for example, with switchgear and UPS systems. Because everything is normally redundant, that room is probably going to interface with another room. If construction does not get those rooms ready in the sequence we need to commission them, it can cause real problems.
On some of the really big data centers we are seeing now, much of the delivery is phased. You might have an 80 MW data center split across two or three phases. How do you continue construction on phases two and three without affecting phase one? Even within commissioning itself, how do you commission systems on phase two without affecting phase one?
There are real challenges there, and they are global, not just European.
Ownership and scope gaps are also worth discussing. Many people see commissioning as a late-stage gateway activity: something that needs to be done at the end. That is a problem because commissioning needs to be involved early. It needs to be involved with the design and early construction so the team is prepared.
Once you get into energization and dynamic commissioning, you need to be ready. There is a mindset issue as well. I am seeing a shift where people are starting to recognize that commissioning is not just something to get us over the line. It is a key part of the whole process, from design through construction and into operations.
David: That is very interesting. One thing I would add, based on both of your perspectives, is that even at a regional level, the operating philosophy, sparing philosophy, and redundancy philosophy can change from region to region because they are driven by capital investment strategies.
Each region has its own challenges, and those challenges cascade down into the design. From working across different regions, I can appreciate that the operating philosophy itself changes region by region.
When you mention sparing and redundancy, that brings me to the third question, which is really the backbone of this discussion: completion versus readiness. Both project and operations perspectives are valid, but this is the core of many of the problems we face.
When can we confidently say, from a data center perspective, that a facility is ready for operation beyond construction completion and individual system testing? What evidence should exist before the facility is considered truly ready to operate? Let’s start with you, Steve.
Steve: For data centers, it comes down to the integrated systems test, or Level 5 testing. That is where you take everything that has been done before and prove it together.
This varies from person to person, but it generally follows the ASHRAE guidelines. Level 1 is generally factory testing. Site delivery and site installation may also sit there. Level 2 covers interfaces, pipework, cable connections, and dead testing. I view that as commissioning QA/QC. It is where you confirm that what you have is what you ordered, that it is where it is supposed to be according to the design, and that it is ready and safe to energize.
Then you move into Level 3, which is pre-startup and vendor site acceptance testing. Level 4 is integrated system functional performance testing. Then you reach the IST.
The building or phase, depending on the project, is not ready to operate until that IST is signed off. We normally have a commissioning agent on board who signs off and says, “This is ready to go.”
The documentation leading up to that is enormous. At each step, from Level 1 through Level 4, you have gateways with documentation checklists to sign off. You need construction completion documents, dead testing records, installation reports, RAMS, live testing results, BMS interface evidence, screenshots, and more.
All of that typically goes into a commissioning system. There are several systems in the market that people use. They feed into checklists and functional performance scripts, all of which are clearly documented and signed off by general contractors, vendors, commissioning agents, and other responsible parties. All of that funnels into the IST.
Once the IST is complete and everyone signs off, that is operational readiness. That is the handover of the building, the phase, or a certain number of data halls, depending on the project requirements.
David: Excellent. The integration and the common view of the risks ahead are what form the transition to operations. Rodrigo, how about you?
Rodrigo: I agree with Steve. I think the most important part of the transition is that, when people finish Level 5 testing, they often say the data center is ready to operate. But they do not always think about training the people who will operate it.
In a data center, we need time to train the operations team. Operating under normal conditions is not the same as operating under emergency conditions. The operations team has to know how the system works during an emergency or abnormal condition.
Many people think, “The test is complete, everything works, so we are done.” But they do not think enough about the next stage: training people, preparing the common documentation, and getting everything ready to deliver to the client.
Sometimes certification is also part of the construction project. In those cases, the data center may not be handed over for operation until the certification process is finished. That requires time, and people do not always consider that time when they create project schedules.
David: That is interesting. From both of your perspectives, the transition is really about full integration of people, not only the technical asset. It includes testing, people, systems, and the organization itself.
We need an integrated view of how to transition, and we need to differentiate construction completion from the true readiness state of the organization and its capability. Paul, I’ll hand it over to you for the next question.
Paul: There are a lot of systems involved in putting these complex facilities together. Where do you see the biggest system integration challenges or emerging risks? Is it complex power systems, electrical power, cooling controls, or emergency systems? What is the biggest system challenge in this transition? Let’s start with you, Steve.
Steve: For me, it has to be the phased delivery that I touched on earlier. Five years ago, people were building 10 MW or 15 MW data centers, and they were delivered as a single building. Now we are talking about 1 GW campuses. You cannot deliver all of that in one piece.
If you have a chilled water system, for example, with a certain number of chillers spanning a central cooling loop across four or five phases, commissioning needs to be involved early with the designers.
Designers are designing a building for operation, and Rodrigo will probably corroborate this, but sometimes there is a gap between how the building will operate and how we actually get it to operation. It is fine to say, “This is how it will work once operational,” but we still need to build it and test it.
I have seen several cases where teams look at how many chillers are needed for phase one or how many power strings are required. The same applies to electrical systems: generators, switchboards, and UPS systems. Once phase one is complete and a customer is in, you may still have another two, three, or four phases to commission without affecting phase one.
It is important that commissioning has early buy-in and sits with the designers to ask, “This is what the building will do at the end, but how do we get there? How do we move from point A to point B? How can we test it thoroughly? How do we collect the evidence? How do we do it without disrupting the live phase?”
Personally, I think you almost have to treat each phase of a 50 MW or 60 MW building as its own project, while still looking ahead to the next move. Commissioning needs that interface and buy-in with construction teams, designers, and operations teams.
If you hand phase one over to operations, they do not want you coming in and interfering with it while you are working on phases two, three, four, or five. For me, phased delivery continues to be the biggest challenge. We touched on grid reliance and power, which are also major issues, but for a project-specific challenge, phased delivery is always the biggest.
Paul: That is exactly why commissioning needs to be involved early. Phased handover and working in a live environment are definitely challenging, especially when you need to keep everyone safe, make boundary isolations crystal clear, and avoid disrupting facilities that are already in service.
Rodrigo, what are your thoughts?
Rodrigo: Every piece of equipment is important and critical for the data center, but I think the most important area in the last stage of the process is controls.
Right now, we have a lot of redundant equipment. You have to prove that the control system is ready. For example, if you have four generators and one is in maintenance, the system is degraded. If a failure occurs, what happens? Is the controller ready to assume the power requirement with one generator unavailable? The same applies to the cooling system.
What happens if you have a CRAC unit or other cooling equipment inside the clean room or white space in maintenance? What happens if a chiller or another piece of equipment is unavailable? What does the control system do in that situation?
In data centers today, we need increasing certainty that control systems are ready to take action in the case of failure. We have maintenance people and human operators, but human response time is not always fast enough to maintain online operation. Every day, we have to trust the control system more. For me, that is the most critical part of the data center commissioning process right now.
David: Excellent. Hearing you talk about this transition, I want to focus more on the tension between operations and project teams. Let’s focus on the project-to-operations interface.
Building on what Rodrigo just said, where do you see the greatest disconnect between project teams and operations during the transition itself? Are operations teams involved early enough? Is the disconnect in testing or training? Where do you see the biggest tension during the transition window? Rodrigo, let’s start with you.
Rodrigo: This is common because many people think the data center industry is an IT industry or an office-type environment. It is really an industrial project.
It is hard to change the way project teams think about maintenance and operations. I think operations teams are not involved as early as we would like. In many cases, they start their training process and become involved in construction very late, even though they need to know critical things for operations.
This is an issue in Latin America. Another issue is recording knowledge. People come and go in every company. Sometimes you train people and then they leave, so the training process has to start again. Managers may think they will have one or two people who know everything about the data center, but that is not how things work now.
We have to record everything so that other people can come in if the people with knowledge change. We need processes, protocols, and training that make the operations team confident when responding to emergencies.
When operations are normal, everything is fine. The real test of the people is when there is an emergency in the process.
David: Resource availability and knowledge transfer are key, particularly in Latin America. Steve, what do you see in your region? What are the distinctions between project and operations?
Steve: For me, it varies from project to project and customer to customer. I always think that if the first involvement of operations is during handover, it is already too late.
Most places have a gateway during Level 4 commissioning, when you are doing functional performance testing. Companies and consultants I have worked with normally include a gateway there to involve operations in witnessing and vendor training. That is quite common, and it does happen in most places.
The other side is that if you bring operations in too early, you are adding another stakeholder during the project. I have seen operations teams come in with the best intentions but start snagging items that are not necessarily snags or punch-list items. They may simply be unfinished work. That can add complexity to an already challenging project.
It is about balancing the need to bring operations in early enough, while making sure they are familiar with the design and the intent of the building. In commissioning, we do not invent anything. We do not create anything. We take the design and what has been built, and we make sure it works.
Whether it is a data center, oil and gas project, or pharmaceutical facility, that principle is the same. As I said earlier, commissioning needs to be involved early. It is also important to engage the operations team during design so they know exactly what they are getting, and then involve them throughout the commissioning process.
It should not be a case of arriving at IST with a big stack of documents and saying, “Here is your building.” No operations team wants to be put in that position.
David: Excellent. That is perfect. Paul, do you have another question?
Paul: Absolutely. This is a global industry. Data centers are being built all over the world and in many countries. Let’s talk about local challenges you may be experiencing on your projects.
What local factors are affecting successful commissioning and the transition to operations in the regions where you work? How are projects or companies in your regions responding to those challenges? Rodrigo, let’s start with you.
Rodrigo: As I mentioned earlier, in the region we need to educate and develop more people in commissioning because we have a serious lack of specialists in this area.
Many people think commissioning is only a testing process. Sometimes they think it is the same as startup. They do not think about the other types of testing or the five stages Steve mentioned, from Level 1 to Level 5.
In Latin America, many people are not familiar with these stages. We have a lot of work ahead of us to certify people in commissioning and expand knowledge of the whole commissioning process, both across industries and especially in the data center industry.
For a data center, it is critical that commissioning clears the issues that may occur before operation starts. Once a client is in the data center, nobody wants to shut down. You have to be very sure that all conditions and all scenarios that can be tested are tested before the first client puts information into the data center.
Paul: Steve, we would love to get your thoughts on some of the challenges you are seeing in the UK and Europe.
Steve: For me, it is the interface between global standards and what customers want. Big end users, some of which David mentioned earlier, have their own requirements. Those requirements may have been generated in the United States, for example, and then the company wants to build data centers in Germany.
All of the tests, specifications, local rules, and local regulations need to be taken into account, and they do not always align. It can be difficult to manage expectations depending on who and which companies you are dealing with.
It also goes back to what we said in the first question about resources. How you test a data center does not really change. I have done data centers across Europe and Asia, and the testing itself rarely changes. But there are nuances in documentation requirements, water testing requirements, electrical testing, and whether a project needs to meet NETA, for example.
Many companies want to work to NETA, but Germany does not work to the NETA standard. You are not going to get an electrical contractor in Germany who is familiar with NETA. As commissioning personnel, our job is to conduct the right scope gap analysis and identify which standard is higher, whether NETA or a local standard.
Then we have to balance what local contractors and vendors can provide against what a customer or commissioning agent may require for sign-off. That is a real challenge.
In Europe, there are many differences even between neighboring countries such as France and Germany. The fundamental testing stays the same, but the route to get there can be very different in France, Germany, the Nordics, Spain, or Portugal. That challenge can be tedious.
There are also practical problems: supply chain, long-lead equipment, and resource availability. Getting the right people in the right place at the right time is challenging. If I had to identify one main issue, it would be balancing global and regional testing standards, specifications, and regulations. It can be a real thorn in the side of commissioning teams.
David: You are touching on a very important topic, Steve: local and global standards. This is the value of conversations like this. We understand common practices and how different professionals see the same problem from different perspectives and with different regional challenges.
That allows us, as professionals, to understand the common themes and how the industry is moving ahead, while recognizing that everything must be adapted and tailored to local social, technical, regulatory, and infrastructure conditions.
This is one of the key values of these discussions. With contributions from professionals like you, we are reaching that level of global standards. All of this is about predictability and looking forward. The ultimate goal of determining these challenges is to make the transition from projects to operations more predictable, and to bring projects to life.
That brings me to the last main question. If we want to improve project predictability in data center projects at a global level, which commissioning and transition-to-operations principles should be consistent everywhere? Which aspects must be adapted to local regulations, infrastructure, grid availability, resources, and workforce? Rodrigo, let’s start with you.
Rodrigo: The typical phrase we hear in our region is, “We have always done it this way.” We have to change that thinking and start thinking about a global industry and a global process.
There are many different standards. Some standards focus only on the electromechanical part of the data center. Other standards are more integrated and consider electromechanical, security, architectural, structural, and other areas together for a successful project.
On the other hand, there are specific standards for different sectors and companies. Meta, Amazon, and Google, for example, have very tight standards.
When we think about commissioning, we have to prepare more people in our regions to be adaptable to the differences in how companies do things. We need a global standard that covers the details some standards miss and others include. That is the challenge for me: creating a global standard that covers the important details consistently.
David: The most important thing is preparing professionals through contributions like this. I always say there is no silver bullet. There will never be a silver bullet. But there can be common ground. A projected $30 trillion by 2050 speaks for itself. The risks ahead of us are significant, and we need to prepare professionals through the advancement of knowledge.
Steve, how about you?
Steve: Rodrigo touched on some very good points, and so did you, David. If you look at a global standard, I would be surprised if most companies did not already have one. With a global standard, a business can define what it wants commissioning to look like.
You can put your processes, expectations, and tagging guidelines into that standard. For example, how do you want to manage red, yellow, green, and blue tagging? It generally follows ASHRAE, but different companies have their own specific ways of doing it. All of that can go into a global standard.
You want the standard to be detailed enough to protect you and give you a strong baseline for how you want to deliver projects. At the same time, it needs to be flexible enough to work dynamically across different regions.
How you test a data center in Japan is fundamentally the same as how you test it in Germany or France, in terms of equipment and what you want the data center to do, based on the design. But you cannot lose sight of regional specifics: environmental requirements, electrical safety, electrical testing standards, grid conditions, and climate. The climate in Dubai is very different from the climate where I am in the north of England, so you cannot lose sight of local requirements.
You have to balance global and regional requirements. You need global standards that define how you want to commission data centers or facilities. Then you break that down into regional standards: this is how we do it in Germany, this is how we do it in France, this is how we do it in Norway, and so on.
Each project, regardless of region, will also have its own dedicated commissioning implementation plan. That is where you can take the global standard and say, “This is how we want to build the data center, but for this specific project in this specific region, these are the exceptions, deviations, and local requirements.” It is a real balance between global and regional needs.
David: That shows the importance of industry-agnostic standards that provide a starting point and allow professionals to adapt to local conditions. Paul, do we have any questions?
Paul: We do have a few questions. We have about 15 minutes left, so let’s get into some of these.
The first question is from Darren Sangster. There has been a lot of public discussion about excessive water usage in data centers. Can the group give some context on this topic? Steve, let’s get your thoughts.
Steve: Some older data centers used open-loop cooling with cooling towers and a lot of adiabatic cooling, which uses more water. Most of the designs I am looking at these days are closed loop. Once you do the initial fill of the system and circulate it, other than topping it up for regular maintenance, they are not using nearly as much water as older systems did.
I saw something recently that made me laugh. Someone said that farming avocados, which are very popular these days, uses more water than all the data centers in the world combined.
Water usage is a concern, and of course we always consider the environmental impact of projects. There is always a better way to do things. But I find it ironic that many people complain about water and energy usage in data centers while doing it online. Where is that data coming from? It is going through a data center.
It is a good point. My personal opinion is that data centers do not use as much water as some people think they do. Certainly, many newer designs are much more efficient, and they have to be in order to meet updated compliance requirements.
Paul: It is similar to the golf industry. Golf also uses more water than data centers, but we are not usually talking about limiting people’s golf every day.
The next question is for Rodrigo. Timely completion of grid projects and the design of redundant electrical supply associated with data centers are crucial for sustainable and reliable data center systems. Do you have any thoughts?
Rodrigo: Of course. We have major issues with power capacity in many countries. One constraint for data centers is how much power they need to consume from the grid.
There are hubs, such as Queretaro in Mexico, where power is not available right now for all the data centers being planned. Some projections in the United States state that by 2030, data centers could absorb a significant share of grid capacity.
One interesting point is that we cannot think about the data center as an isolated project. This year, in Virginia, there was a grid disturbance. Around 3.8 GW of data center load transferred to generators. When power returned to the grid, the grid effectively saw the loss of 3.8 GW of load. That can create problems with frequency and voltage stability.
We have to be sure we comply with the requirements of the companies to which we connect our power. We also have to confirm whether the grid has enough power for new projects in a specific location.
Paul: Very good. Our next question is from Marlin. Marlin says the important part of data center commissioning is redundancy and failover testing. We simulate conditions such as utility failure, generator failure, UPS failure, cooling unit failure, and loss of power path to verify that standby systems automatically respond and maintain the critical IT load.
That is important testing for sure. Steve, is there anything you want to add?
Steve: That is the fundamental aspect of commissioning a data center. Coming from an oil and gas background, one of the biggest surprises for me when I moved into data centers was how much focus is put on redundancy testing.
In oil and gas, you generally have redundant systems. You may have pump A and pump B, or compressor A and compressor B. In commissioning, you test each one, complete your A sheets and B sheets, and make sure the system works. Then the priority is to get the oil or gas out of the ground as quickly as possible.
The major shift in data centers is redundancy. You have different tier designs: N, N+1, 2N, N+2, and so on. It depends on the design of the particular project or facility.
Everything we do in data center commissioning is about asking, “If that fails, what happens?” It is about protecting the white space and keeping it powered and cooled at all times.
It can be challenging, but it is interesting to see the difference between oil and gas, where the focus is often on production, and data centers, where there are constant “what if” scenarios. If this goes wrong, what does the next system do?
There are times when scenarios become extremely unlikely, and you reach the point where you have proven redundancy as far as reasonably possible. If you lose additional systems beyond that, you may be dealing with a much bigger problem than cooling alone. But commissioning a data center revolves around that core statement: keep the white space cool and energized.
David: That is where ALARP, “as low as reasonably practicable,” changes from industry to industry and region to region. Coming from the energy sector and oil and gas, redundancy plays a very different role in data centers, and it can also vary by region.
Paul: Joseph has a question related to safety. Safety is important on all projects, including data centers. Rodrigo, what are your thoughts on safety as the facility comes live with energized sources, electrical power, and pressure systems? Are there particular safety issues addressed in your data center commissioning processes?
Rodrigo: Safety is very important. Before returning to safety, I want to add something to Steve’s point about redundancy. It is important to consider that some systems need frequency synchronization. For example, some UPS systems must be synchronized at the output because you need to be ready to transfer load from one piece of equipment to another. Some equipment, such as static transfer switches, may have a very small synchronization margin, such as 30 degrees or less.
Sometimes people do not think about synchronization because the equipment works on its own. But when you have to transfer loads from one system to another, you need to think about it.
Returning to safety, every data center has to comply with local regulations or international regulations, such as NFPA 70E for arc-flash studies.
The principle in data centers is that you have two or three paths. You should be able to shut down one path so maintenance can be performed safely. The last thing you want is to work on energized equipment. Redundancy in a data center is not only for the IT load. It also allows you to shut down equipment while the IT load continues to operate.
In Latin America, one issue is that people are not always well trained in working on energized equipment or applying the studies needed to use appropriate PPE or shut down equipment safely. Clients or owners sometimes require equipment to remain operational even when it is energized. We have to be very careful with that.
Paul: Last question, and then we will close it out. This one is for Steve. Operations should receive complete as-built documentation, test records, operating procedures, and proper training. This makes the transition from project to operations much smoother and safer. Anything to add?
Steve: That is a very good point, and it is something we touched on earlier. Every data center project I have worked on has followed the commissioning steps I outlined earlier. It is all about gateways and sign-offs.
Your red tag stage includes factory testing, site delivery, installation QA/QC, cable installations, pipework, flushing, testing, balancing, live tests, dead tests, and more. All of that goes through the Level 1 through Level 5 gateways.
You have sign-offs at each step, along with punch lists and issue resolution logs. You should not move forward if a critical issue remains open. For example, if you red-tag a UPS and it has critical punches, you should not sign it off and move into yellow tag, then green tag, then blue tag.
You need these processes as barriers. That is how we use the Level 1 through Level 5 commissioning process to manage risk.
Documentation is produced throughout each stage and verified and signed off by the commissioning specialist, whether from the general contractor or commissioning agent. It all goes into a system. There are several in the market, such as Facility Grid, CxAlloy, In360, and Omega 365.
That documentation forms the handover from a commissioning perspective. Your checklists, metadata, and asset information are fed into the gateways until you reach the IST. Then that information goes to the operations team.
It is very important to get that right. It does not only go to the operations team. It could also go to the customer, depending on how the project is structured. Everything Marlin described needs to be done well, signed off, and accepted by everybody at each stage, not only at the end.
David: I want to use the last question to thank you both. We are now over time. Thank you for this great contribution to intercontinental knowledge. It will help ensure projects are transitioned safely and that the transition leads to profitable projects.
We have already incurred around $800 billion in data center investment by 2026, with trillions of dollars expected in the years to come. The numbers speak for themselves. As professionals and as a global community, we need to protect that investment by identifying risk and, more importantly, protecting the people executing these projects.
Thank you both for your contribution. Paul, I’ll leave it to you for the last words.
Paul: Thanks, David. Thanks, Rodrigo. Thanks, Steve, for sharing your expertise.
For anyone watching, if you are interested in checking out the ICXA standards and seeing how you can apply this expertise and guided process to your projects, visit icxa.net. There is a preview there, and you can get a copy of the standards. You can start applying this industry expertise to your projects as well.
Lots of great questions and a great discussion. We appreciate everyone’s participation and look forward to seeing you again on the next live webinar. Have a great day, everyone.
Visit icxa.net for more information.