Name: Building the Machine: Infrastructure for the Intelligence Era
Uploaded: 2026-09-11
Duration: 20 min 45 s

Every advance in artificial intelligence depends on a complex industrial backbone: the data centers, networks, power systems, and water resources that make large-scale computing possible. As demand for AI grows, the challenge is not simply to add capacity, but to design and build infrastructure that is reliable, resource-conscious, and thoughtfully integrated into the cities and regions that host it. Success depends on smart choices in data center site selection, architecture, grid access, cooling, and the relationships among developers, governments, and communities. Two leaders in data-center development and design examine what it takes to build at the scale AI requires, and with the public trust on which it depends.

Kevin Guan, Chief Investment Officer & Executive Vice President, Bridge Data Centres 
John Taylor, Global Critical Facilities Leader, Principal , Gensler 
Nicholas Gordon, Editor, Asia, Fortune 

Building the Machine: Infrastructure for the Intelligence Era

Conferences

All right, why don't, why don't we get started, um, so I, I wanna start with you, with you, Kevin, um, you know, you see so many headlines about, you know, new data center projects, huge sums of money being thrown around whenever I travel around the region in every country, people talk about data centers and how they wanna be a home for data centers, rightly or wrongly. Um, but why don't you help us kind of go behind those headlines and those big figures a little bit and tell us kind of what's actually going on in this AI data center boom here in Asia? Sure, just maybe to level set a little bit, um, so British Data Centers BDC were one of the pioneers and largest operators of hyper scale data centers in Southeast Asia. Um, and, uh, we're, you know, 70% owned by Bank Capital, which is an American private equity firm, um, to, to answer your question, um, you know, if you look at where demand is from a global perspective, is certainly robust, is no longer just about cloud services, is no longer about trainings, it's actually about, uh, AI adoption at scale, right, and. Uh, but ultimately all that demand needs to be supplied by critical digital infrastructure data centers and the physical limits to how quickly you can build data centers at scale, right? And so we're kind of seeing significant delays and challenges with delivery in the US. We're seeing, um, you know, demand far outceed supply here in Asia as well, but if you look at Asia as a whole, it's a collection of very. Different tutonic plates. Japan is different than Malaysia. Malaysia is very different than Australia, right? And what we're seeing now is that there's an incredible opportunity to address a short term, I would call, you know, supply constraint in some of the more um. Developed markets such as the US or Europe, right? But longer term, ultimately Asia has 50% of the global population and adoption for AI should take off and catch up relative to the US. And so we talked to our hyperscale customers, they're looking to us to resolve near term supply constraints in other parts of the world, right? But they're also more importantly, Investing for the medium to longer term for the demand potential that there is here in Southeast Asia and in APAC and beyond, um, and we're gonna talk about some of those constraints and those timelines in a second, but I wanna turn to John next and you know, as someone who's involved in the design and construction of data centers. You know what actually goes into that process? Is it more than just a big box with server racks in it? Like how should we think about what these data centers actually are? Well, I think we're well past the stage of them being a big gray box, um, that no longer fits the bill. Um, increasingly data centers are moving into urbanized areas. They're no longer single story, uh, large, sprawling campuses. They are in parts of the world, but in Asia where we're somewhat land constrained, they're often up to 1011 stories high. Increasingly they're coming into areas which are urbanized, sometimes in close proximity to communities, residential communities. And so we have to build sympathetically. Um, the facade, uh, the footprint. Obviously there's concerns around noise as well, there is. Some sensitivity there may be a bit overplayed, but we need to move data centers from being an industrial, perceived as industrial infrastructure to being something which is part of the civic infrastructure which serves a community, you know. There's a lot of pushback at the moment in parts of the world, and I think we're all having to respond, and it's only right that we do so. These sensitivities are, are real. And, and I want to pivot back to Kevin. I mean, because, you know, a lot of data centers are being built in hot climates. You're people are considering building them in kind of urban areas. When you're thinking about opportunities to kind of build new data centers in certain markets, I mean, how are you approaching these many different markets that are here in, in APAC? Yes, um, you know, I think to, you know, follow up on the constraint question, right, we're increasingly seeing more and more of that, um, everywhere, um, no matter what the underlying geography is and across sort of the valley chain, the bottlenecks of constraints are showing up, right, um, and so, um, how we think about resolving these constraints, how we think about, you know. Um, embracing opportunity is really through innovation and also through close partnerships with the community. I'll give you a few examples, right? Ultimately, we're in the business of converting power to compute, right? And so if you start with power. We've been working with innovative partners such as SK and Keppel on new energy solutions, right? We've been, uh, working with strategic partners on, uh, on behind the meter solutions, so you're not dependent and only drawing on the national grid, right? And then um the designs that go into our data centers where we co-develop with our hyper scale customers and our, you know, um, supplier partners, right, so we try to make them as efficient as possible so every NBA of power you can e-draw, you can generate more, you know, what we call IT capacity, right. And when it comes to water, for instance, right, in, in Malaysia, um, water is becoming more of a constraint. So instead of using and competing with local residents for clean water, what we've done is we've signed partnerships with the local municipal, and we've taken municipal wastewater, right? And we've built a wastewater treatment plan on our campuses. And we're recycling municipal wastewater using a portion of that for our data centers and what we don't use we return to the community in the form of clean water and it goes on and on all the way through operations where you know we work pretty closely with uh local um schools and communities to really train um workers for the next generation of uh. Work and careers in digital infrastructure, right? So I think innovation and close partnerships with local communities is what it takes to address increasing constraints. So I, I, I do wanna follow up on that and I also wanna get John's views after yours, but you know, the, the conversation about constraints have largely focused on power and water. I think water in particularly in the US, but I mean power is important too. Are those still kind of the main constraints on data on data center construction, or have we moved on to a whole new set of constraints? Those are still the constraints, um, but they're increasingly more innovation to resolve those constraints, right? So for further elaborate on water, we've developed waterless solutions, right? So you know, um. We, we can take closed loop water, circulate those what we call directed chips, cool the chips, and instead of evaporating the water, right, we can use other ways to cool down that water, right? So, so that's what we call water solutions. There's, uh, you know, besides recycling wastewater, there's more efficient ways to use water without wastage. But I think what's under the radar and what's becoming more of a constraint. Is sort of this competition for talent, right, and how do we retain, train, and, um, elevate talent and then what's actually becoming an even bigger constraint is capital, right? To give you a perspective, Bridge, we operate, you know, 600 megawatts this year will be about 1 gigawatt in operations next year. Depends on where you build these data centers. In Japan it could be as expensive as $20 million US dollars per megawatt, and that's only 20% of the total investment that goes in, the rest being servers, chips, racks that that that go into the data centers. And so there's a global, um, uh. I wouldn't say competition, but a global grab for capital as well, right, um, and, and, and I do want to return to that point later, but I do want to go to John. I mean, again, for you, do you see power and water kind of still being the major constraints for data centers, or again, are there, are there new things to keep in mind now? Well, there are, there are further constraints. I totally agree with the war for talent. That's, that's real, um. And er not enough young people coming into the sector is going to become an increasing challenge, but it's up to the sector to to train young people and find ways. The other constraints is really supply chain of long lead equipment is becoming a constraint very much. And some. Long lead items like generators can be over 3 years from from certain um global uh vendors, um. One of the challenges of course is also permitting. I think we've seen a pause in Southeast Asia in Thailand just over the last few days where there was There's a misunderstanding about how these buildings should be classified and uh and that's put a complete pause and there will be a reset there on the planning and permitting and that's after Malaysia's spurting growth that's probably the next fastest growing territory in Southeast Asia. So, getting back to the point earlier about buildings, one of the constraints is also how do we make these buildings, their form and function. Um, capable of still lasting the 15 to 20 years they're designed for, when actually the life cycle of the chip, the latest AI chips coming out, well you know where they're coming from, but say in Nvidia and others, is, is maybe less than 5 years, yeah, so we need to have buildings that have the sufficient floor loading to allow for adaptability, um, and, and well as uh as resilience and the ability to actually change the use of a single floor. In terms of the mix between the data center hall or white space and the plant rooms and all of the other equipment that needs to go into that, so we are having to look at that very, very closely and a firm like Gensler are not engineers, we're purely architects and that's our specialization, a global firm and we work with engineers in very, very close collaboration and together with the clients operations teams as well so that we've got actually sustainable designs with the adapt. Level of adaptability that we envisage. So you've got to take a very long term view of that building structure and its form going forward. I mean, we, we were talking about this kind of before our panel about kind of these sort of misaligned timelines. You have the investment timeline for ROI. You have the construction timeline. You have the timeline for the tech. You have the timeline for AI where it seems to change like once every week what the leading model is. You know, from your perspective, how do you get all these things to align? How do you how do you design your operations to get these timelines to make sense? Yeah, so I think, you know, to level set, I think there's two schools of thought to building data centers, right? There's the enterprise lens, right? These are, you'd be building, um, data centers for multi-tenants, usually smaller, maybe in in more prime locations. There are, um, uh, what we call the hyper scale lens, right? This is where you actually work pretty closely. On a pretty bespoke basis with only a handful of global hyperscalers and you're often building these in sort of, you know, um, at uh near prime locations, right? And these you often build at scale and so what we do is, is primarily hyperscale and so we work really closely with our off-takers, right? So, um, you know, to Piggyback off this point made just now, how we think about futureproofing is, you know, we've got these large campuses we generally sell out portions of it of it at a time to a particular customer. And for the portion that we sell, usually we design a couple of buildings at a time. These buildings are laid out where the corn shell structure is generally transferable across hyperscales, and what goes in are what we call modulized containers or modulized solutions. And so the uh corn shell that we build has the right ceiling, has the right right load, um, uh, support to be able to be so-called future proof, right, um, and what goes in, you know, some of the, you know, modular, um, uh, solutions such as, uh, the power modules, you know, things in there such as UPS and transformers can be swappable, right, and so. On a 5, 1015 year looking basis, um, our customers have the ability to essentially swap out pieces of the Lego brick without, you know, changing the structure, um, and so in a way that's how we align. The the design today with the unknown needs of the future. Um, we need to talk about the stakeholder engagement point because, you know, data centers, they are politically controversial, particularly in the United States, but increasingly here as well. I mean we've already talked about Thailand's decision, Singapore and Malaysia have gone back and forth about whether they want. Them or don't want them, I think Australia made some noise about it and then backtracked so the politics are confusing and there's some controversy around around it even here in Asia. So you know, John, from from your perspective again as you're thinking about data center design and making them part of the community, I mean what are the sorts of things that designers, operators, owners have to do to have communities be more comfortable with this industrial infrastructure now in an urban environment. Well there's an interesting case at the moment which we can't name, not even the country, I'll say it's in Asia, North Asia, um where actually the developer is giving back to the community in the form of a um a retail park. So a large proportion of that site is now being given back to the community um and and. We've, we've done some beautiful renders, we think they're beautiful, the client thinks they're beautiful. We've got to hope that the er that the authorities having jurisdiction, the city mayor and the community also agree. um but um that's, that's an interesting aspect now. So we're seeing much more integration of data centers into um mixed-use developments. Yeah. And not just in, in urban areas but in out of town. Be adjacent to shopping malls and community facilities where we can also be more sustainable in terms of the heat rejection can be used for district heating cooling systems, um there's even a case of swimming pools, would you believe, being heated from er from the rejected heat er coming out of the data center. But we've got to get ready for an evolving er landscape there and how, Campuses are gonna have to be able to er adapt for um. On-site power generation and energy storage is all the more important where the grid is unable to deliver power. And even in places where the grid is robust, Malaysia is a case in point where they've had something like 7 gigawatts of available power in very short order. That is not the case everywhere else, and other parts of Asia Pacific are waiting up to 5 years. I think the average is close to 4 years to get grid connection. Um, Kevin, I, I said I would return to the, to the capital point. And I mean again you read the sums of money that are being raised for AI spending, AI CapEx. I mean was like Alibaba raised $10 billion the other like couple weeks ago on the Hong Kong equity markets. So these are just huge sums of money that are all being put towards data center construction, compute, all of this stuff and you know, to, to ask a slightly sharper question, I mean what do you think about that kind of, you know, big spending huge sums of money approach towards CapEx building these giant compounds. You know, is that approach kind of does that, does that, does that meet the moment like is that more than we need? Is it still not enough both in general but also kind of here in Asia? Yeah, it really depends on what the ultimate use case is, right? And so, you know, there's gonna be segments of demand where, you know, it's what we call traditional workloads and these cloud services that's pretty uh inelastic and it's pretty durable, right? A significant portion what we addressed in Malaysia is for certain clients that, you know, do cloud hosting or video hosting for their social media platforms, right, we think that's pretty robust and pretty durable for the AI. That's a big part of the stack, right? Um, increasingly we're seeing evolving architecture where people are no longer thinking just training or just inference. They wanna deploy the latest generation of chips to do, you know, top shelf training, and when those become more. they start thinking about using that for inference and then later down the line 5-7 years down the line maybe using that for traditional services as well. So I think architecture is moving towards scale towards workload agility, right? And so which is why we're starting to see our customers come to us, say, you know, if you can deliver on a 28, 29 time frame, you can offer a couple 100 megawatts the path with a path to a gigawatt, you know, that's very attractive to them. From the operator lens, right, we're in the business of providing the critical infrastructure or a safe operating environment for our customers. So we where we build right in terms The underlying power commitment and also the fiber connectivity is, you know, it creates a strong moat. And you know from an investor's lens, right, if my initial contract with my off taker more than covers, you know, my build costs or has a good path to covering substantial portion of my build costs, that gives me comfort because I do think, you know, the chance of a second contract and a renewal is pretty high given how critical this infrastructure is for these, you know, large hyperscalers. Um, I don't wanna forget the sustainability energy question and maybe that's a question that maybe is suited for John. So I mean data centers need a lot of power, um, and it already seems like among some governments in Asia that there's a little bit of, let's say, kind of walking back some of the renewable commitments. There's a recognition that like, you know, fossil fuels are pretty reliable and data centers need reliable power. You know, how are you then approaching kind of the sustainability questions when it comes to internet to ensure that there isn't any, that there isn't any kind of real backsliding towards, you know, dirtier forms of energy and other things. Well, in terms of the timeline, we as a firm have made a commitment to um zero carbon by 2030. And that was made at the United Nations by our co-CEO at the time, Diane Hoskins, who now co-chairman. That's looking extremely challenging amongst our 4300 clients. I'm talking beyond data centers now, all 34 market verticals we address. Um, but that's a commitment we're trying to live by. Whether we have to lose some clients along the way, I hope not. But we are trying to ensure that um all of the materials which we are specifying into our buildings have um a. Carbon footprint which is lower than the alternatives, and I can give you an example for a hyperscaler in Europe where phase one we measured the carbon impact of all the materials which were supplied to phase one and for phase two we said, well, actually we need to get all of the vendors to tell us exactly what their data is and we will exclude the offenders in favor of those who have the Lower carbon impact and saved over 25% on the phase two of an almost otherwise identical design but with different material vendors. So that's, and we keep a register of that and those are the kind of ways where through design and tight specification and monitoring of what is installed by general contractors and their subcontractor supply chain. I look forward to seeing a beautiful data center in a neighborhood near me sometime soon. Um, I wanna thank Kevin and John for joining us today and talking through, talking us through this really interesting topic. Please give them a round of applause. Thank you.