← Risk in Context Podcast
Loading summary
Transcript33 lines
- [00:01]
A
Welcome to Risk in Context, which features conversations with Marsh colleagues, risk professionals and others intended to help you better understand key risk, build more effective insurance programs, and think creatively about risk. I'm Mike Matthews, Marsh's Global Digital Infrastructure Practice leader. Digital infrastructure projects represent some of the most capital intensive undertakings in today's economy, with estimated trillions of dollars expected to be invested in global digital infrastructure projects by 2030. But for many owner developers as well as tenants, one of the most consequential decisions relates to selecting the right site, with questions related to availability of power, water and talent, human capital, as well as connectivity, natural catastrophe, exposure and insurability. Considering the scale and value of these assets involved, getting site selection right is critical for developers, tenants, lenders and insurers. In today's episode of Risk in Context, I'm joined by three colleagues from across Marsh who are in daily contact with our digital infrastructure clients as well as the team at Marsh. Derek Wishmeyer, Senior Advisor with our Global Digital Infrastructure practice, Lash Magoo, partner at Oliver Wyman and Maglem Tiwari is also a partner at Oliver Wyman. We will discuss some of the key questions data center owner, operators and developers as well as tenants should be asking and share practical actions they should consider as they make critical site selection decisions. I'd like to introduce my fellow panelists today for the podcast. Lash, why don't you introduce yourself?
- [01:39]
B
Thank you Mike. Hi, this is Laksh Magoob, partner with Oliver Wyman, based out of New York. I spend most of my time serving clients within the digital infrastructure value chain. So everything from hyperscalers to the telecoms to the equipment providers and data center developers, and last but not the least, the private equity players and the private capital players funding this massive build out.
- [02:06]
A
Thank you Lash. Derek, it'd be great to hear some of your background and what you're seeing today.
- [02:12]
C
Yeah, sure. Derek Wishmeyer My former background is general counsel for data center business and leading into that general counsel of a fiber business. So I bring an operational risk aspect and overview to all of our clients, navigating contractual risk, vendor risk, lease risk and site selection risk. In my perception especially the data center world and digital infrastructure is ever changing and the risks, uncertainties and some of the questions that our clients are facing are ever evolving from site selection to power procurement, you name it. And all of our clients are facing those issues day to day.
- [02:49]
A
Thank you Derek. And last but certainly not least, Mangalam, be great to hear some of your background and what you're seeing today.
- [02:55]
D
Yeah, thank you Mike. My Name is Mangalam Tiwari. I'm also a partner based out of sf, work very closely with Laksh as well. And I work across the entire value chain with the digital infrastructure, helping private capital on investment thesis, helping hyperscalers and some of the frontier labs on investment thesis and site selection. And Mike, I think Derek called it out clearly as well. I mean this industry is blowing up. We expect the total compute capacity to be somewhere around 80 gigawatts in the next three to five years. The tenancy is really high, the vacancy is really low. So in that regard, site selection becomes a key and it really sits at the intersection of, as you said, par, availability of water and availability of related infrastructure. So sites given, given all the trends in this area, I think site selection is key and this is where our clients need the most help. And I've been working in the space for the last few years and a lot of changing and evolving dynamics happening here, which I'm sure we'll talk more about.
- [04:06]
A
Well, thank you. As we talk about site selection, it's been a massive shift in data center strategy over the last 20 years. As you look at just say the Fortune 500 organizations, they used to own a lot of their own data centers 20 years ago. And as technology migrated to the cloud and for any other reasons around investment and proper use of capital, they have partnered with data center owner operators. So you have two sides of these site acquisition, site selection fence. And we tend to spend a lot of time talking about the owner, operator, developer side of site selection. But whether you're building your own data center or you're selecting a site for your data center partner and operations through a third party operator, site selection is a key component. Whether it's latency, availability of power, cloud, compute demand, physical location, human capital. There's so many instances or so many key components to the site selection selection process. We also look at owner operator. Developers need to know the insurability of a site, they need to know the finance ability of a site and they also want to look at the leasability of a site. So if you're building a new multibillion dollar data center, why would a tenant select you over another one? What can you provide? Whether it's physical security, availability of power, cloud compute demand, fiber density in the area, there's so many things that go into it. And Marsh as an organization has built a very, very strategic team around this site selection process. One of the key aspects of site selection is power. Eric, maybe you can talk about some of the speculation, the interconnect agreements and how the complexity is shifting for our owner operated developers that we're supporting today.
- [06:06]
C
Yeah, there's a myriad of changes, Mike, that are facing the industry, especially as the power shift goes from on meter to behind the meter. What we're seeing is not only an industry changing and what the infrastructure looks like, but the risks involved and whether it's a natural gas powered turbine or other nuanced sources of that back behind the meter power. Let's just start with what behind the meter means and what we're seeing in the industry. So traditional data centers, Mike, as you had referenced 10 years ago, effectively all used power that was off of the grid. So utilizing those utilities that provide power to homes, residences, town centers and malls, they were utilizing the same utility off of the meter. And now with the demand for increased power as well as short supply of power on the grid, including NIMBY and NYMI issues in certain jurisdictions, data center developers are being pushed into a corner where they are have no choice but to build power facilities and infrastructure that creates and generates their own power on premises with the data center and we call that behind the meter. There's a number of issues that are growing as the industry evolves to the behind the meter model, including employment difficulties in finding the talent to not just design build that data center as well as power plant, but also to operate it. We're also seeing difficulties navigating the procurement side of sourcing those gas turbines for the natural gas generation of power. Behind the meter.
- [07:35]
B
Makes sense I think to better frame the whole behind the meter argument and this trend, I think it's important to also look at what's happening behind the entire powered land. Right. Like a couple of years back when we were talking with our clients and working with everyone from developers to hyperscalers, powered land was the phase or the phrase which feels now a little bit outdated.
- [08:01]
A
Right.
- [08:02]
B
Which is the focus was hey, how much megawatts or gigawatts does this piece of land or parcel that we are selecting have access to today? How that's evolved is that people have realized that the nameplate megawatts or gigawatts are no longer equal to interconnectable tenant usable megawatts. Along with that, the arms race to get to the megawatts for be it the hyperscalers or the frontier AI labs or the GPU as a service Neo cloud developers. That arms race has also led to a variation in what people are willing to the risk profile and the customer fit. That was before and that's sort of evolved. So a site, what I mean by that is a site that used to work for everybody doesn't necessarily work for everyone. Now a Neo cloud or a Frontier AI lab may choose a site that they're okay with for training workloads that may not work for an enterprise colo or a cloud availability zone demand. So there is a difference in their risk appetites and that's creating this need for. For a differentiated way to get the fastest time to market or fastest time to revenue. And a lot of these players are very much open to now behind the mirror which was not the norm a few years back. And that's to your point is where we are seeing or why we are seeing this ramp towards behind the meter. And we estimate from our own models this to be about anywhere from 10 to 15 gigawatt. That at least is short on the lower end and can be even more. That has to. To meet that demand we have to go towards behind the meter.
- [09:56]
A
That's a great, great point. Lash. One of the things as we take two or three steps back we have large global publicly traded data center owner operators shifting to acquiring building their own power for the very first time. On the opposite end of that scale we have first time data center developers building large campuses for single tenant simultaneously developing behind the meter power. So you have two very large infrastructure projects happening at once full of risk. We also have power developers building their first data centers. So you have this ecosystem ranging from globally large established performers to new performers and new power producers. It's just full of risk and execution, all backstop to contractual performance and deliverables. So we talk about this tightly wound ball of risk. This is actually a pretty big avalanche of risk happening all at once. All being backstopped by insurance products and lots of sophisticated investment dollars. So we're seeing this momentum build and we have to get creative around risk solutions. Contract advisory, continuous advisory specific to site selection because. Because you have to get it right from the beginning.
- [11:37]
B
I agree. I think the. This reminds me of. I heard one of my clients say this and I loved it was the not in this age and in this era cheap land can be the most expensive capital in the stack if power permitting and the demand is not bankable. Right. And that's what. And those are the risks that you're. That you're solving for now.
- [12:03]
A
Yeah.
- [12:04]
C
We're also seeing geographic availability of contractors that have experience or are willing to take on the work. As the massive push into data centers grows, the number of contractors with experience building large or small scale Data centers is dwindling and their schedules are being filled up. So even the geographical risk of having a contractor available becomes one to bring
- [12:26]
D
my consulting had in the play. When I look at the energy for data centers, I mean typically we look at seven things. A availability, B reliability, 3 speed to par, 4 cost, scalability, sustainability and also local market fit. And especially that. You just mentioned the contractual structure on all of that. Right. The biggest shift as correctly pointed out, is that par has moved from being a utility input to being a strategic constraint. And in many markets demand for AI computer is clear. But the bottleneck is really whether you can secure the deliverable megawatts or not. But the other nuance that I want to point out, that not all megawatts are equal. Right? Data centers need power that is firm, redundant, scalable and economically viable. And renewal power is essential, but intermittent power has to be firmed. Grid power is ideal, but interconnection queues can be multi year. And you mentioned behind the meter and bridge power. They can accelerate deployment. Absolutely. But they introduce fuel emissions permitting and operational complexity. So the winners will not just be the operators with land or capital, as Derek was pointing out, but the winners will be those who can integrate power procurement, grid strategy, backup generation, cooling contracting and sustainability into one coherent infrastructure plan. In the AI era, cloud strategy and energy strategy are absolutely inseparable.
- [13:56]
C
That's a great job of highlighting all the facets that go into the data center construction piece. I think stepping back where all the risks is born is following the revenue and the business model and that is stemming from the lease agreements, especially tailored to the hyperscale clients and tenants. Those lease agreements include a number of issues that are becoming conditions that are harder and harder for data center developers and power vendors to meet. They come from in two flavors. It's the operational uptime after the data center is up and running and those are guaranteed. We're seeing more and more four nines of available uptime of power and cooling. But the other side of the risk is delayed startup related whereby there are strict timeframes for the data center with delivered power and the correct environment to be delivered. And if not there are very sharp penalties, financial penalties for the data center developer if those data centers are not turned up timely. And those risks then flow through that to all your vendor agreements, making sure your long lead time equipment is on time, delivered, operational, the power is sourced and the building's completed. To avoid those harsh penalties, I'd like
- [15:08]
A
to state the four mission statements of any data center owner, operator, developer, mission Statement number one, acquire land and power. When you, when you achieve that mission statement to is to acquire a tenant. Mission statement number three is to develop, develop and deliver the operational facility without liquidated damages. Mission statement number four is 101% uptime. It's that clear. And if we look at mission statement number one, acquiring land and power, we're starting to get requests for insurability as a marketing tool, operational marketing tool. And that is a exercise or a strategy to acquire tenants. So maybe Lash, are we seeing our developer operator community starting to use our advisory to acquire tenants?
- [16:08]
B
Absolutely. I think the what's interest and it's going in two ways. One is often the developers are identifying who they want to work with and what's the target and picking parcels of land based on the target itself. So the thinking about who's the tenant goes from day zero and those are typically the most successful ones because as we were saying the requirements and the zones can be completely different. So that's one second is the other way is how they take into having bankable customer demand and taking their needs into account. And the timing of the power based on the customer's needs is also critical. So the early some of these developers are picking their partners is critical. And where we are working with them with developers has been helping them go to market with some of these, be it the Frontier AI Labs or the Neo Clouds or even the Hyperscalers and providing different kind of compute, be it inference focused or sometimes training focused. But inference is becoming a large chunk of the conversation now.
- [17:29]
C
Mike, to add on to Laksh's comment, sourcing the right location for data centers is becoming more of an art and a science. What we're seeing, the value and the diligence that we provide is also first of all it's requested very often and it's twofold. So because power is becoming more difficult to source then our developer clients need to find property with power and find out what they can do to mitigate that from an insurability and construction risk, whether be still construction to mitigate waters and floods, additional wall strength or ceiling strength for wind. These are things that our developer clients are anxious to know ahead of time as they're getting pushed into locations that have power, but maybe not the best natural category or natural catastrophe risk exposure. So they can mitigate that from the outset in their design. But what we're also seeing is with the triple net lease function, you know where the tenant will, the hyperscale tenant will absorb the building's responsibilities for taxes, insurance, and maintenance. They're utilizing our diligence reports as an advertising mechanism to show what the lowered capex ongoing for those customer clients and tenants will need. And we're seeing a lot of traction and demand for that now because of
- [18:42]
A
that, we touched a little bit on some of the labor component. But part of that site selection is saying Talking about day 300, day 900 as we look at the operational life cycle of a data center, whether your Data center is 20 years old or day one brand new, in five years there will be a major retrofit to that facility. So we look at data center life cycle as a constant flowing chain of events. Renewal of technology, renewal of services and getting skilled labor not only for the development but the operation on a 24 by 7 basis in some remote areas of the country, remote areas of the world has become a big challenge for our owner operator developers. So Magla Melach, how are we helping clients with that human capital element from an advisory standpoint during the site selection process?
- [19:43]
D
It's a great question Mike, because the labor availability is absolutely critical. But if I take a step back and look at holistically how we are thinking and talking to some of these operators and site selections, I would think about this and maybe eight to nine lenses like first starting back from a customer back monetization, like a specific name customer archetype that the stated workload would fit, having that visibility in mind, keeping in mind the power ramp certainty, the substation and interconnection reality, also the bridge part feasibility and regulatory and political bankability, campus expansion option value including the surrounding nearby DC ecosystem and then specifically the buildability labor and construction. Right. And I think the last piece that I would touch is also resilience and insurability as you alluded to. But now if I double click especially on the labor and construction part, it's been a key ask for my clients just understanding what the availability is, who are the top vendors, what are the typical rates? They can negotiate them with that because given the labor scarcity. But we are helping clients think through construction labor by looking at the wage rates, looking at the union jurisdiction records, looking at the Dodge Construct Connect Pipeline, EPC capacity surveys, also the OEM equipment real time labor data to come up with a comprehensive view of what labor is available, what is the shortage and what typical rates they should negotiate at to ensure labor availability.
- [21:28]
B
If I have to just put that in in terms of the kind of work Mike that we've been doing, right. I think there are three categories of things that we're Doing as part of this baking this into a parcel selection and site selection. So often we do we work how do we bake basically this labor into site selection, not afterwards is is really building a trade by trade heat map of be it electricians, mechanical commissioning and controls within the radius of the site. That's one second thing we do in that is we really overlay competing other competing mega projects that are nearby, be it data center fabs, other infrastructure programs to see future crowding. Because at the end of the day that all eats into some, some of the same linemen and the same same electricians and so on. We want to understand do we have sustainable supply of people. We also work with our clients to lock in delivery partners before the land closes. So can we for example line up preferred EPC and a couple of backup contractors during the diligence phase and do a soft reserve of some of these critical trades and commission and commissioning teams and have a bit of a early modular build strategy to reduce completely the local labor dependency and then we can do it a bit in a modular function that gives us a credible construction plan that the ICs now want to see. So the investment committees for some of our clients actually want to see these labor plans and remediations upfront before they approve the entire funding.
- [23:10]
A
That's fantastic. Information team, thank you for all that. I think we'll tackle this next subject by committee here. As we look at some of these sites, whether it's in floodplains, the deep southwest, the center of the country, remote parts of the world, we have to start taking a look at natural catastrophe. Certainly these facilities are engineered to withstand a whole heck of a lot. But from floodplains, windstorms, earthquake, hail, wildfires, we can't select where cloud compute is needed or where the demand is. So the data centers have to be there. How are we helping clients with that natural catastrophe aspect of site and development? I know that whether it's engineering, it's really the putting both the tenants and investors at ease for the development and operation of these facilities. But what are some of the things we're doing from an advisory standpoint?
- [24:15]
C
Mike, I'll jump in. And it's a good point. It's one thing that having joined Marsh, coming from industry I think we excel at and that's getting ahead of the tough questions with developers and power generation clients is jumping out of the gate to ask them have they reviewed where they're located, what the site selection process was for them and if they're and to help them review those CAT risks right out of the gate so that we're able to help them solve for it in their engineering, their design and their construction.
- [24:45]
B
That makes sense. I think the question Mike, you're asking is how do we really help our clients? Not just answer can we build here but really ask the question can this site actually be insured, financed and trusted by the hyperscaler or NEO cloud or an enterprise colo through a one in 100 year event? And the key part there is really about when we are working with some of these clients it's about how do we get or remove bad sites early which is parcel level, hazard scenes and screening through. You mentioned floodplains or wildfire or the risk of hail. Really understanding some of that. And if any reaction here is if any of these things impairs tenant sla, we identify that upfront and we heavily discount that parcel and take into account there. There is also of course from a marsh lens how the ability to quantify, I would say the insurance premiums and deductibles and the capex premium that for example, how much if it's a floodplain, the elevated pads is a capital premium that needs to be engineered into the design. The roof hardening needs to be done, the drainage needs to be thought through differently. So those are capital premiums that needs to go as part of the engineering work for some of these sites upfront.
- [26:10]
A
Thank you Magwa, many thoughts.
- [26:12]
D
In my experience, natural catastrophe catastrophe risk is becoming one of the most under appreciated parts of digital infrasite selection. Right. Historically developers looked only first at powerland fiber tax incentives and maybe speed to market. And those still matter, don't get me wrong. But the next level of sophisticated sophistication is asking will this site be operable, insurable, financial through the climate and catastrophe conditions for the next 20, at least 10 to 15 years, right. Like for example, a floodplain is not just a mapping issue, it is a capital allocation issue. Right? So an uptime issue, an insurance issue and a timely customer confidence issue. So in a lot of my client conversations we actually build this as pointed out, like what is the premium required based on what are the probability metrics out there? And we include that metric, which is a key one as we think about site selection because data centers will have to exist. It's just that what are the payments that we're looking at? What would be the risk and insurance products associated with that? Right? So the real question is can I keep power available through a flood, wildfire, heat wave, hurricane, ice, storm, drought, or even a grid emergency. And some of these questions are being answered up front through probabilistic models.
- [27:36]
A
Well, thank you. I think we all have our thoughts on this ecosystem and if we had one key takeaway, you know, focused on site selection, I would say for myself I'll start. I think that we need to take a look at the human capital aspect of this because of that 24 7, 365 functionality, that five year life cycle of complete retrofit. This is a massive demand on a very stressed part of the labor force. So as I look at that, that site selection and moving people to different parts of the world, different parts of the country, I think that is a human challenge that needs a lot of attention today. Maybe. Derek, any other things that come top of mind?
- [28:27]
C
There is an entire menu of items that do become relevant. I think Mike, you're right. Those are the most pressing at the moment. But availability of infrastructure as I've the telecommunications and fiber providers are getting stretched then now as well. They're coming to the table and requiring upfront payment for construction of the lateral lines by way of either an IRU as we know it as an indefeasible rights of use, which is a long term lease for fiber. So requiring upfront construction payments is something that's trending now to build to these data centers. And that's another capital concern that some clients had, didn't imagine, you know, a year ago. But it's becoming a pressing issue now for them. So in the site selection, making sure availability of not only one fiber provider, but perhaps redundant fiber providers are available and knowing what the cost of having it built to the facility is makes sense.
- [29:19]
B
Fully agree. I think if I have to then summarize a bit of our discussion. Right. And takeaways has been around. So site selection is not about finding land with power, given all the things that we're seeing with powered land. But all the queues is really about how do we underwrite a specific parcel to deliver bankable power that's available today and customer usable or tenant usable megawatts and gigawatts on a timeline with a clear path to scale. And that includes scale includes the fiber. The water's becoming a bottleneck too now in some cases. So how do we make sure the power permitting and increasingly water are taken care of?
- [30:01]
D
Mike, I really want to hone down on the labor availability issue because that's absolutely critical given all the demand for new data centers. And really in my mind the labor issue in data centers is more nuance than simply Saying that there are not enough construction workers, right? Because you need the right kind of worker, the right skilled workers. And really today the real shortage is with specialized mission critical labor, right? Like the electrical contractors, the mechanical contractors, commissioning agents, control specialists and like other experienced providers who understand high density, high reliability environments. That matters because data centers are increasingly being built in markets where power is available. Because everybody power is top of mind for them. They're solving for power. But not a lot of people are thinking through this skilled labor availability. So you are exchanging one bottleneck for the other, right? Think about Phoenix, Columbus, maybe even Atlanta. These are high growth markets. They are attractive but also require labor mobility, contractor depth, local management, housing, travel models and strong safety execution. We are working with, I mean we've worked with a lot of private capital players who are also thinking through some of these maintenance or MEP providers, as we say, like the maintenance and hardware providers and their availability and supply across these markets. And really at the end, my thesis is that labor will become a front end site selection criterion, right? And not a backend procurement issue. The best developers will build labor heat maps alongside power maps. This is some of the work that we are helping our clients do, leveraging all the data sets that are out there. They will lock MEP capacity early, create preferred contractor partnerships, use off site prefabrication to reduce field labor intensity and then really orchestrate the crews around it leveraging like AI tools and data alongside with it. So I think really that would be the competitive advantage going forward.
- [32:09]
A
Thank you. Thank you. It was great insight and great commentary. That's it for our edition of Risk in Context. We hope you enjoyed our discussion and thank you for listening. You can rate, review and subscribe to Risk in Context on Apple Podcasts or any other app you're using. You can also follow Marsh on LinkedIn or X. In addition to your podcast feed. You can find more episodes of Risk and Context and more insight from Marsh on our website marsh.com thank you from the Digital Infrastructure team here at Marshall Sam.