Speaker 1: 0:16

Chris, you've been in this game for a while. Yeah, it seems to be the hottest space on the planet right now, like what's changed. Seems to be the hottest space on the planet right now, like what's changed. It's sort of, uh, it's come from I don't know. You started in before dot com. You saw that explode, you saw that come through. Data centers were interesting. Now they're the coolest place to be. Yeah, ai, yeah, oh, it's going on.

Speaker 2: 0:39

All the trends yeah, um, yeah, no, it's been an interesting I mean it's 20 plus years, right, and, and just watching the multiple secular trends come into the market, right where you know. I came from networking, yeah, and you kind of saw the writing on the wall where networks are critical component, just like megaport, right, like you have to interconnect these things and then live the whole. Let's build data centers and watching where the value was being created and a lot of that was in the data center and that was just the early days of, like, content distribution. Yeah, that was a big problem we were all solving for at that time, which was solving for the internet and getting that close to folks.

Speaker 2: 1:12

Yeah, latency you know proximity to the actual consumer than cloud right and I think people forget. There's a long time where data center operators cloud friend or foe and we had to educate the market on the cloud lives.

Speaker 1: 1:25

Yeah, the cloud's going to take everything away from data is irrelevant.

Speaker 2: 1:27

Actually, it turns out the clouds are in the data center yeah, yeah, like there's, believe me, there was some uh storied big, uh big names of investors where it's like it's all over.

Speaker 1: 1:36

The dream is the server it just moved and they're like what?

Speaker 2: 1:39

and so you had to slow that down.

Speaker 2: 1:41

It's the same building, yeah exactly all building is made from this building Exactly. Well, long story short, it wasn't because I was COLA only at that time, okay, oh. But now the shared like we get to build it from both workloads of private and public inside of the cloud. I see you build both, yeah. Yeah, I know Digital, really pervasive PAB, cage, megawatts, hundreds of megawatts yes, and interconnecting all those with our partners yeah, that is the most pervasive platform in the market today.

Speaker 2: 2:13

But to answer your question I just want to close off on when we saw cloud, we learned a lot. Right, like, remember, for many, like almost two years, we were consuming the cloud over transit, over the internet, and people forget those days, yeah, early days. Yeah, how we really got the data center to be relevant to the cloud is providing private access, productivity, yeah, private access with everything. And I think you guys are at the epicenter of that. This enabling hybrid, multi-cloud, multiple destinations in the most simplistic fashion. That's what brought the true colo-type data center environments relevant to the cloud, is allowing that true business class consumption of these public services. Yeah, and so that's what I think was foundational in watching that data center gain in relevance and enabling that hybrid multi-cloud scenario.

Speaker 2: 2:57

But then now ai. It's a whole another thing. Yeah, how do you consume ai today? Yeah, well, most people transit. Yeah, right, and it's just that's true. You're going to have to relive this thing. Yeah, so that's why it's like the next cloud, in effect. Yeah, yeah, so we've been talking about with Megaport and everybody, like private AI exchanges are going to be right.

Speaker 1: 3:27

Well, I think you're in 25 plus countries, and so what we're seeing and I know you see this because a lot of these GPU as a service companies that are appearing out of nowhere and they're different to the cloud providers, of which there was started with one and then became two and then three, but we're seeing sort of a lot of random companies appear. They're actually deploying in a lot of your data centers all around the world and then they land with all these gpus and they think how do I connect to them? Right, forgot that bit. You know. It's like this interesting. That's the last thought inside most people's head. But how are you seeing that play out? Do you think it becomes fragmented? Do they consolidate? It feels like gpu's are moving so fast. Yeah, you know, all of a sudden the old gpu is a new gpu. Jensen's like innovating so quickly, it's super. Yeah, it's. What do you see?

Speaker 2: 4:14

yeah. So I kind of take a step back and and give you a framework on how we think about it and how we talk with our customers. Right, and you know it evolving quickly and so we're always being mindful of what our customers are trying to achieve, and so the most simplistic way to think about it is tokens to Watts, to dollars. Like what is your token production, which is the most simplistic output from an AI kind of algorithm, and how many do you need? Like is there proximity values or latency scenario, and where does that token have to be delivered? That's really where you start to dig into how are these things going to be performing over time. The second thing we look at is watts, not only power capacity, but the density of that power.

Speaker 1: 4:55

And that's about close keeping them close to each other.

Speaker 2: 4:58

Well, no, it's the thermal density required for, like these, nvidia chipsets. Oh yeah, just the sheer, sheer yeah, for, for you know, I've been at digital 10 years but for the 10 years we've been here, we've always been looking at and working with nvidia and these partners, with dgx certification program to. There's so many data centers that can't support, yeah, that power density within their act. And the last one is dollars, not just the dollars to buy it, but then how to operate it and is that how mainly, or is that a mixture?

Speaker 1: 5:25

It's both.

Speaker 2: 5:26

Because of the expense to build out this, particularly AI infrastructure. It's rather expensive, right? Yeah, I mean, a 32 megawatt build can be over a billion dollars, wow, and so you don't want to get stuck in that, but then you also want to watch to your point the operation of that. Yes, so, to answer your question directly, if you're always thinking about that equation and finding the efficiencies in it, you'll have a good outcome. Yes, and so, with your neo clouds, or these gpus of service providers coming to market, we're always watching how this, what do you mean? Neo, that's the name. Yeah, I think gartner or one of the analysts gave this the sector. They're calling them NeoCloud.

Speaker 2: 6:03

So the new cloud bringing these GPUs to market, and we're watching that, and this is why I always orient people back to the lessons learned that we collectively as an industry learned. Is that bare metal as a service? We've all seen a lot of the starts and stops of these businesses. Super capital intense yeah, it's capital intense. You have a window of opportunity to get a return on capital and the option to expand it. Yes, it's really tough. And so you're right. Nvidia, with the rhythm of chips that they continue to bring back and innovate, that's what's so crazy.

Speaker 1: 6:35

They're moving so fast, so fast. That's what's weird. It's like the perfect storms arrived and most of the time and if you rewind back to 2000 was like, let's just roll that fiber cables, but they weren't innovating that much. It's like a piece of glass. Yeah, maybe it's getting faster optics or what have you, but now we'll look at jensen. Every year he seems to be, oh yeah, quadrupling, yeah, the speed at these things, and so you know, you just gotta keep updating them.

Speaker 2: 6:59

I don't know what's the well, you're gonna see a lot of shifts, right, like, so these neo clouds of gps service, you're gonna see everybody start talking a service? Yeah, okay, because they want to obfuscate the hardware underneath. Outcome yeah, like, do you just want to buy tokens and you don't need to meet them in the pricing of these tokens? And I'm happy to go into detail, like some of the newer models yeah no, it's interesting and how we think about it.

Speaker 2: 7:19

But, like that innovation factor on learning from fiber, and we're constantly watching how we, as digital, can be performant and supporting not just the market today but future-proofing where we see our major customers going, because we've grown up supporting the largest hyperscalers in the world and so we've been looking at how this is going to evolve and how we can sit down and be foundational to their success without going up the stack, if you will, and that's where I think these GPUs of service they are popping up all over the world. I think it's interesting that the workloads are pervasive for AI. It's not vertical specific. We're still early days, so it's exciting times, but you really want to watch that efficiency factor. Like I talked about tokens, watts to dollars.

Speaker 1: 8:02

And so do you look at country. When you've got an ability to go anywhere in the world, because you've got all those locations, are you looking for lowcost power options or is it? Is it more the us at the moment? How has it all sort of playing out on a global?

Speaker 2: 8:15

sense. Yeah, we spend a lot of time like our capital, like at the core of what we do for our shareholders is efficient allocation of capital, and so we're looking at the market dynamics. One perspective is price of power, but more times than not now it's availability of power.

Speaker 2: 8:29

Oh my gosh like the interconnects in a lot of these core markets. It's tough but for yeah, the company's been around for 20 years. Again I've been there 10. We master plan with the community, with the utility operators, to have that long runway in large capacity blocks. That's our heritage and our asset class speaks to them. Yes, um, so we're, we're really watching the dynamic within the market, where it's not just how much it costs but it's the availability, and so working that balance in each of these markets is what we do for our customers and it's not easy, right? The supply-demand imbalance is definitely out of whack.

Speaker 1: 9:05

Well, that's the. You know and I find, like publicly everyone. So data centers became cool, so chat speededy appeared. Now everyone's in the data center game, Right, and they're all like. You see, I don't know, there's some really interesting public figures out there now buying land in all these sort of places. Never been in data centers in life, but they're real estate developers, apparently, and they're going to build all these data centers. The point is A it's a long build, it takes time. You've got to know what you're doing. You guys have been doing this a long time. Not only do you have to physically build it and build it in a way that caters for all these challenges that come with crazy amount of power and support and protection and safety and security, and you've got to find the power and do, I think, agreements with the, the states or wherever it is the entitlement is yeah, is it's a culmination of events that have to come together to get a data center turned up and powered up.

Speaker 1: 9:55

So do you find it funny, Like seeing everyone's in the I don't want to be disparaging towards anybody. I've got random builders coming like telling me they're going to be building data centers. I'm like what are?

Speaker 2: 10:04

you doing this will make a lot of people laugh people out. I'll be, I'll take the high road and I'll be on the positive answer of that. Um, I think you know not all data centers are equal. Like I, I really impress that and that's why a lot of our bellwether customers continue to come to us on a global basis is it's not only getting it built, but then it's operationalizing as well. Run and the modularity required inside of these data center builds right because, like when I bring one to market, you want it to be relevant for 20, 20 years, 20 plus years and so being able to bring in the electric distribution, the cooling, and constantly ideate with our customers based upon how the infrastructure evolves.

Speaker 2: 10:39

That takes a balance and a heritage of engineering talent. Yes, you know, we've been perfecting this over 20 years, so there are a lot of upstarts. We don't put our, our capital, into far-off markets where we don't see that durable workload lasting for a longer period of time. Yeah, like in some weird place. We're really proximate to the city center. That's kind of our core focus and differentiation going forward.

Speaker 1: 11:02

And it's a long-term asset. Then if something's slowed you can build it with something else. You can switch it to Colo.

Speaker 2: 11:11

Or, you right, hyperscale, traditional, and that's where it's a little harder than just going, hey, colo, hyperscale, but like, having that versatility and the four walls of the facility is is absolutely critical and that's where we've differentiated ourselves. When I joined and people might not know this um, I joined, like the day we closed on telex, because I wanted to make sure that we would have the culmination of capabilities coming to market, because the reason why I changed my career coming from Equinix to digital really is I kind of saw the fact that customers were outgrowing a colo only model and needed larger and larger capacity blocks. Right. So phase one, phase two was great, but then phase three, like taking a much bigger and that's only grown exponentially with recent AI, and then it just took a different form factor and then working with you, right, like one of the first things I did is we got our partnership set up with Megaport.

Speaker 1: 11:57

I think you're the reason we even landed in the United States, 100% I mean we were forming and thinking through that.

Speaker 2: 12:04

it needs to be an open platform Because, also, people miss like. You got to really be mindful of not only where your data is being created, where you want to store your data. You want to be in an open platform because it's not going to happen in one single data center, in one market or one provider. You need choice and you're going to be able to do that easily.

Speaker 1: 12:23

And that is probably unique. If you look at a lot of traditional companies, they consolidate globally. Take Amazon Eventually it takes down Amazon, down amazon, apple, google, all these things. Data centers are still, even though there's I mean, you're a monster, but there's a lot of other data centers out there, it's like quite disparate in terms of the number of companies, yeah, that are operating and I I think that's probably because it's more of a real estate place that'll be different to, yeah I mean know like I often get joked with that you're a technologist at a real estate company right, Like what are you?

Speaker 2: 12:55

doing Right, and then they kind of start to dig into what I do at Digital. Really, you need both right. You need the acumen to understand where the land is, the entitlement to the land, and understanding how to grow. Like just to pick on one of our largest markets and how it came to fruition. Is Ashburn, virginia, yeah Right, like that took a master, like that took a master stroke from a real estate side. Like we have multiple campuses there over a gig. I mean it's the epicenter of cloud and it will become the epicenter of AI as well, and they're stretched for power. Yeah, I mean yeah, yeah, yeah, and so we're working through that. But, like what's nice is just to your real estate point. It takes a bit of that, but you also have to marry that with the technical acumen of the designs, the services, to enable your customers to be successful and find that runway. Because we have a digital does is another gig coming online right next to the airport.

Speaker 2: 13:49

I mean we are not like. We are going to continue to support our customers in that evolution.

Speaker 1: 13:54

So one thing I always love about I think your role and I think that anyone listening would get insight from you're kind of at the tippy top of it all in terms of engaging with NVIDIA, as an example, or Supermicro or whoever it may be, dell, all these different interesting cats but you're seeing what they're seeing around the corner and I know you can't share a lot of that publicly, but any interesting insights just to what you're seeing. I mean, nvidia is obviously the really interesting one. I know you're close with those guys, so what do?

Speaker 1: 14:26

you see, it's kind of mind-boggling, like looking yeah, I mean, I don't know um it's.

Speaker 2: 14:32

It's. The core of your question is what's kept me in the data center business for so?

Speaker 1: 14:37

long. Well you've, you're very curious guy. Yeah, no, I love to learn.

Speaker 2: 14:41

I love to learn I love to innovate, and innovating in a vacuum is no good, and so having exposure to all of the different types of operators, nvidia being one, a lot of our customers, the amount of work that we get to do with them privately and in different labs, it's what shapes the outcome that a lot of our enterprises get the benefit of. Or like early days, like people may remember, like we worked with Zanga and Zanga wrote kind of an early blog where they were doing Farmville, and like there was a blog and it was phenomenal because he's like I want Right, like I'm an enterprise and I'm paying 30, 40 percent more than Google's getting. Like why? And so similar to technology and how it's going to evolve, not only on a cost basis, but you really want to stay close and we learn in.

Speaker 2: 15:27

A lot of our enterprises benefit from a lot of what we're seeing come from a lot of our partners and NVIDIA is one, particularly in the GPU space just really staying aligned to them, not only on the rapid innovation of chip sets that come to market, but, like she's inside, oh it's absolutely, it's absolutely fast, but I think you know, again, they're the one of the fastest innovators out there to really be performant and you know what they've brought to market. And just to give one highlight that I can talk about because it's public, is up in denmark, right. So we won, won a major deployment with Novo Nordisk called Geffian, and Jensen went and visited it and sat down and it was a great talk because what he saw that Novo Nordisk Foundation was doing in this facility, he called it a center of intelligence, right, because it's not just around the infrastructure and the tech stack, it's what's being done with this Lifter. Yeah, it's absolutely amazing, and so you know it's one of the largest DGX pods in Europe and being able to support that rather quickly with our platform up in Denmark, where they wanted to be close to where their corporate headquarters were, to a bit of data sovereignty. It's phenomenal, and what we were able to do as a collective to support that.

Speaker 2: 16:35

And you know some of these hyperscalers with interconnectivity and a lot of the on-ramps that have been coming to market and evolving our on-ramp strategies and just removing complexity from the enterprise is where I see a lot of the dialogue I have with these hyperscalers, because we've interconnected and we've done a great job for the hybrid multi-cloud. But we really need to start thinking about what's going to be required for enabling AI. Yeah, because private consumption has to happen. Higher capacity throughput has to happen. It's not about lower latency, it's about consistency. Latency Because one of the things I will impress upon everybody if you're not thinking about the internet working within the four walls of the data center, where most of the conversations happen beyond the four walls is even more important Because as new models come to market with reasoning, generating a token with reasoning, it might have to go through three or four models to really truly answer the output of what the original request is. Yes, so those types of things are going to continually put pressure on us.

Speaker 2: 17:32

Because bimodal it's here, right, you're going to get video, like. One of the greatest use cases I got exposed to is NPCs right. You're going to get video. Like one of the greatest use cases I got exposed to is npcs right. Non-player characters and games yeah, super latency, sensitive and, like my son, mason and samantha, they love roblox right and the worlds that are being built, oh my gosh, and like. So I think like that's kind of like what we saw with farmville way back in the day where I joked with people because we were talking like what is what do you do? And like we help you sell pigs faster and they're like what I'm like it's a cross connect and so hopefully, in the not too distant future.

Speaker 2: 18:05

I'm going to tell my kids that character right there. We help that communicate in a more human fashion inside of your game.

Speaker 1: 18:12

Yeah, our npcs are kicking your ass. Yeah, which? The latency is perfect, exactly. I don't want to. I've got another question for you. Lots of data centers build locally, so there'll be a bunch in the US, but they don't expand to different countries. The complexities that come with it. There's plenty of reasons why there's jurisdictions, but something you'll see that I think most others won't is how AI plays out for sovereign and the governments. And look, every government is going to be different. It's moving so quickly. Regulation in each place. Do they want this sovereign data? What's has it playing out on the government side?

Speaker 2: 18:49

of things. Yeah, I think AI again is pervasive right and sovereignty the sovereign clouds or the sovereign AI is being built out. I think you'll see a lot of that and I think those will be kind of purpose lag for their outcome. Oh, it is lagging, it's um, I don't know, maybe it's. Maybe it's important to kind of I always give things frameworks that help everybody. Yeah, about it. So in the world of disruption, which is kind of what we're talking about with, some people are ahead or behind. You know, I always watch three factors.

Speaker 2: 19:17

There's hardware right, there's the hardware how performance is going to come, what's the power requirements associated with it? Like, yeah, like the token production, the watts to flops inside of these chips. I'm always watching that right, because if you have something come out and it can do, you know in 100 kilowatts what we're doing with a megawatt. That changes, and so we're always watching that. Yeah. The second bucket is software right which could also influence the first. Yeah, it does. Yeah, yeah, that's why it's. The second bucket is software right which could also influence the first. Yeah, it does. Yeah, yeah, that's why it's the second kind of critical component. So that software element we're always watching in like open ai, like how is it performing at the apis? When they brought out omni with voice and you just keep watching some of the core technologies in it. And I think most recently and I'd be dismissed not to mention it, but if you want to do some research, it it's most recently come into the press in the past, like 60 days deep seek. You should go read about it, right, because it's 96 cheaper per token than like an open ai and it uses a lot less power, yeah, or compute cycles to produce that, yeah, so there's a lot of impact.

Speaker 2: 20:16

Oh, it changed. Yeah, so your point around some stuff like they're behind. Maybe they're not, because the software is evolving so quickly. Maybe you don't have to do it like some of the others originally did it to get the same outcome? Yes, well, hopefully, because it's a huge capital. Yeah, the capital. But I think it's pervasive again that I don't think the demand is going to shift. But but how you got to that end? Answer is gonna change. Third bucket, and this is near and dear to our heart right, like I've lived, grown my career through internetworking, like having built the first on-ramp and cloud exchange, we're gonna. Yeah, that is what I did at Equinix. And the third bucket is networking.

Speaker 1: 20:53

Right, because again, which is the bit that they always forget. It's like everyone's talking about power land they go. And it's like everyone's talking about power nobody they're going to talk, and then they go.

Speaker 2: 21:00

Oh yeah, oh, hang on yeah, no, we got like right now, 24 was the year of land grab, right, 25 is going to be the year of efficiencies. Because, yeah, we were just building networks everywhere because time to market was critical yes, and they also weren't thinking through like and because the way the models were very arcane and like it's training and then it's inference right. But now you're seeing these models where the inference has micro kind of training and within it got a key sort of up to date and the mixture of experts coming into these models. That takes a different type of networking, which is why we often talk about the private ai exchange. It takes a community like us and others to come together to solve that connect all this.

Speaker 1: 21:37

That's cool. The last one is on that the countries need to think about the power generation inside those countries. We're getting to a point where these data centers are consuming I mean, yeah, you, you give us some stats on some of that yeah, gigawatts of power that's been used. I mean, yeah, a data center is going to consume like almost what a country's yeah is that?

Speaker 2: 21:57

yeah, a little bit. I mean, like, one of the things that I often always try to hit on is, like the value of the data center is becoming more and more understood. Yes, right, but it it does take a good bit of power, but I think the value generated, that power is an important element to it. But like, and that's where countries need to think about, what they're, oh yeah, how they allocate.

Speaker 1: 22:14

Like how they allocate, you think australia, there's enough power to deliver this. So then do you fall behind as a company, country, country? Yeah, you can't. And then it takes you 20 years to build the power.

Speaker 2: 22:24

So you're like this as the same in america. Yeah, like, we're always watching, like, because the grid is constrained in most places and it's not only around production, it's distribution. So, yeah, it's a concerning scenario, but you know, I always watch. Like you know, we run a little under three gigawatts. Today we have three gigawatts coming online. We're really doubling. Yeah, because the power, the supply to me, I mean it is, it is crazy days, which, I mean, has changed this conference right?

Speaker 2: 22:49

yes, yes, yeah but um before, like a lot of folks here trying to work out where to deploy, right now, like I was doing an interview not too like it was a while ago and it was with the BBC and he was like so what else don't I know? And I'm like, because I'm like, well, nuclear is going to be a real deal. And he's like what? And so he wrote about data center operators going to do nuclear and you see all these nuclear deals recently, but not traditional nuclear, but new nuclear, and I'd be remiss not to talk about the generation. Yes, right, and I think we'll be. I've been excited and working with a lot of smr providers, like, and some of these other ones where which is small, yeah, small, medium react, yeah, small modules, small modules. Yeah, it's an elegant solution, right, you can tuck it in the grid, you can like, so it overcomes the grid limitations it's I mean the way they, and then much, and they're much quicker to build.

Speaker 1: 23:39

Yeah, much cheaper.

Speaker 2: 23:41

But there's a lot of. We don't want to rush into this, we want to make sure it's safe, but even the recycled fizzle material and stuff that they're able to use, I mean, there's no way around it. It's a great story for humanity because it really alleviates a lot of that carbon that we're under today, that we're under today, which is, I think, lots of different countries are trying to figure that piece out.

Speaker 1: 24:00

But it's, I always think, the nature of the United States. When they find an opportunity, when companies find opportunities, that they're constrained by something, if it's not the government that will solve for it, the corporations solve for it, and that's what we're looking to Really interesting companies now getting into power gen, which is something you'd never consider.

Speaker 2: 24:22

We were doing an opening in one of our data centers in Tokyo and the PR team inside were like you cannot talk about nuclear, like, do not talk. I got more questions at this opening about nuclear because even in that market they're just thinking through exactly what you're asking. They don't want to be left behind, they want it safe. Yeah, that's a worry. People are embracing this.

Speaker 1: 24:41

Well, it's moving so fast and if the country is left behind and can't offer that, it will go elsewhere and it becomes like the workload moves, yeah, and then you miss out and you can't provide it. And can you catch up? I don't know, it's hey, thanks for taking the time, but thank you for actually investing in Megaport. I don't know what year it was, but you're the reason we landed in the US. It's our largest market. It's still growing the fastest for us. We think we're just scratching the surface. We still think we're still educating the market that network as a service is actually a thing and you can deliver connectivity in 60 seconds. So thank you for that partnership. We'll keep building in your data centers, wherever you land. That makes sense for us to be 100% yeah, and thank you for that.

Speaker 2: 25:28

Thank you for the opportunity and thanks for the partnership as well, and it's a critical thing to continue to focus on. So thank you, pleasure Cheers, awesome Thanks.

Speaker 1: 25:34

Chris Awesome mate.