Ep16: The Power of Interconnection in Digital Infrastructure

(upbeat music) This is UptimeNow, a Norden Roseville-Bite podcast dedicated conversations about data centers from development and construction to cutting-edge technology and operational breakthroughs and the people driving it all. Joining us today is Juan Font, President and CEO of Corsite and Senior Vice President of American Tower, where he leads the strategy, innovation, and growth. He brings more than 25 years of experience in general management, direct sales and business operations and finance in the data center and telecommunications industries. Welcome to this show, Juan, how are you? - I'm doing great, great to be here. - Glad to have you. I'm sure folks would love to know. Can you tell us a little bit about your path here? How did you end up at Corsite and American Tower and as a president and CEO of your company? - Yeah, so I've been in data centers for 20 years and I joined Corsite in 2010. In my initial role was in sales and then I was able to rise to GM, which is a bit of a CIO at the market level, where you look at profitability and growth per market and was doing that for the Eastern C board and eventually took over the entire company for a GM perspective. And eventually ended up with a position of American Tower of Corsite ended up as president and CEO. So it's been the thrill of my career and not honored to be here. - Curious of all the roles that you've had so far, what would you say is the most challenging? Would you say it's this one or something before? - You know, I'm blessed to have a fantastic team. So, dimming my job a little easier. From an institution standpoint, I don't need to worry about the operational data data match. My job is more about capital allocation and strategy and managing the board and American Tower, but the day they performance to the company is really the problem with my team. In my previous job as a GM, I think it was a little more challenging 'cause I had to always ask for money and try to prove that I could deliver the return on the VISTA capital. I'm more a smaller company. Now I think we have the right resources. We have a great team and I'm pretty well capitalized. So it made my job a little easier. - It's great. It's always helpful to have a great team behind you. Can you tell us a little bit about Corsite and American Tower and just how the two companies work together? - Sure. So Corsite is a premier operator builder of data centers nationwide. We have city data centers in other markets about 5 million square feet of data center space. And our focus is to build data center campuses that are highly interconnected. And so American Tower is the largest operator of cell phone towers in the world. In the US alone, they have roughly 45,000 towers. So the idea was to merge a data center company that has a lot of interconnection capability across the US and a portfolio of communication realistic assets on the mobile world. I'm preparing for a future in which thinking about communication more broadly, there will be a convergence between wireline and wireless applications. Think of the integration of CISG services and robotaxies and those glasses and wearable devices. They're all going to require a lot of compute close to the edge. And leveraging both portfolios, I think would provide us an ability to deliver end-to-end solutions that require a lot of capability at the market level, at the edge. And leverage our access to clouds and generate AI solutions in combination with the broadest portfolio of cell phone towers out there. So that was sort of the genesis and the thesis behind the acquisition. - A lot of people outside of the industry, even people inside of the industry, sometimes in discussions tend to couple, every type of different data center together as if they're just like one thing that serves one purpose to everybody and that's obviously not the case. The comparison example that I have is like, if you were to talk to somebody about the difference between a coffee shop and a Costco and a Michelin restaurant, they immediately know the difference. But what are the different restaurants of the data center world? - That is actually an analogy. Another one would be like an airport, where more like a hub where you can go to multiple destinations to different carriers. In the case of restaurants, you know, there are the majority of data centers that are in play today, that are in the public mind, are what we call hyperskill data centers. Those are very large purpose-build buildings that typically host a single user, typically hyperskilled like an AWS or a Google or something like that. And that's, you can think about those of most as a warehouse for servers. And a Costco would fit that analogy. In the case of Corsite, you know, some of the single restaurant, if you think about, for example, an enterprise cafeteria, where the enterprise in the headquarters has a cafeteria for their employees, they control the asset, is there an employees that work in the kitchen, but also is the captive audience. It is really the people that go eat there, no matter how great the food is, it's going to be your employees. And that's sort of like an on-prem data center. And, you know, to this day, most of the data centers are still an enterprise that have built it themselves. And they are like that. And in between model, it's sort of the collocation data center provider, like Corsite, and think of us more like the town square, where you can, we can own that, and you have multiple restaurants. You can have a Michelin star restaurant, you can have a coffee shop, you can have a big old shop, but all of those are accessible. You can park your car within a couple of steps, you can ask is that a broader ecosystem of restaurants. And that is what a collocation provider does. So, I'm a believer that, you know, Nautilus Quinoa was created equal. And when it comes to data centers, it's a huge disparity between the different business models. - And what is the business model that you guys are most focused on? - So our business models centers around interconnection. It is a place where we enable digital communities to interoperate, and those digital communities consist of enterprises, just regular enterprise like a bank or a hospital, digital platforms like Zoom, for example, or things like that, digital platforms like your digital media provider, for example, the company's stream services from cloud providers, where the large, you know, the largest cloud hyperspeeders out there, and networks. And all these companies and organizations interoperate with each other on our sites. So for that to happen, you have to have multiple hybrid networks that are natively present. You have to have cloud-owned ramps that are natively present. And you have to have enough space power and cooling to support mission critical applications so that customers can deploy the infrastructure and access their digital ecosystem or the other deals applied chain, if you will. - And can you talk a little bit about the different types of workloads that are out there, and how they're changing and how that's reshaping the types of data centers that are out there and the people are building? - Sure. So if I go back in history, so the first deal revolution was the internet. I mean, that was roughly in 2000s. Just like that decade, where social media emerge and applications like search and websites and gaming, for example, there was a role, you know, require specific data center buildings. And that was the emergence of dedicated data center and purpose-built data centers, but you're sort of needed all these networks to connect the internet, those companies to the internet and applications to the internet. You first forward to the 2010s and that's the emergence of the cloud. That's the notion that enterprises could access compute on demand, you just pay as they use. And that revolutionized the industry. You know, somewhere the service was invented and it's not a result of that. And that required a, the hybrid industry started building sort of cloud compute clusters of ability zones. And a government like Corsair had to evolve its business model to provide access to those cloud clusters for enterprises. And the workload that emerged from that generation of technology change was what we call hybrid IT. Hybrid IT is the notion of an enterprise that deploys infrastructure and a third party business and like Corsair. And the substance of the workloads are in their own dedicated servers. And the rest of the workloads have performed in the cloud or more recently in multi-cloud environments. So you need to have a designer operator that has space power and cooling at a proper density but also very quick integration to multiple clouds. And most recently, I would say the third sort of generational revolution is intelligence. And that is AI with this portion of AI. Which to me is probably the most relevant because you know, you go from an architecture of CPUs to GPUs. And that shift is exponential in terms of power usage. And that has created this very large scale so AI data centers, hyper-scalers. But also in our own case, even though we multi-tenant and do location, we have to be able to host GPU workloads that require much more intensive power and cooling densities. And also enterprises that are doing hybrid IT now are also doing that not only just for cloud compute applications, but also for generative AI. And oftentimes they need to connect to new clouds to rent those GPUs because those GPUs are very, very scarce. And so most enterprises don't have access to buy sales and video chips and deploy them on their own data centers. They need to go rent them from someone. But when you rent, you have to connect through some sort of fiber interface that has to be scalable, flexible in expensive. And that will, of course, I provide. It's access to not only the current cloud provider, but also the nascent AI frontier labs, if you will, but also the near cloud. And that's a bit of a evolution of the work most as a different, no, digital revolution takes a take shape. It seems like there's so many different considerations when you think about location, from fiber to power to where is the actual compute, to all the interconnection that you just described. How do you think about location and where you want to be? That's a very question. I mean, it is really no different than an enterprise in the company that wants to set out the headquarters. You want to be where people are, where you have resources, you're close to the transportation hub, you have access to schools, and all that. In our case, it's the digital services, right? It's fiber mostly. And the major fiber hubs were set around the biggest meters like here in the Virginia, or Chicago, or Dallas, or Silicon Valley. It's also close proximity to the subsidy cables, for example. And so over the decades, the organization, that networks specifically, both domestic and international, ended up clustering in a few locations. And that's what we set up our shop, basically close to all those fiber metro and regional and international fiber hubs. But in addition to that, you have to have access to land and access to power. But the population centers are also important, because that's where all the major population, the enterprises, and people reside. And our data centers' obligations serve both regional, local, and enterprises. And if you think about Silicon Valley, for example, most of the startups on the digital platform side emanate from there. So they need, when they first start, they want to have a local collocation provider. So that proximity to an enterprise base is very important. But it also, people, a lot of the streaming services collocate with Corsis to access the eyeballs with relatively low latency. So the relations are typically, for us, it's large of metros with large population centers, where you have an established fiber density and fiber hub in place. That's also the main criteria. Yeah, it's funny how quickly expectations change, because a year ago, I don't think I was using AI at all. And then now, virtually everything I do, I'm plugged into AI and some capacity. And you go from not using it all to having this incredible tool to being frustrated when it takes a little bit too long for where to come up with a response to whatever you've asked it to do. I guess, talk a little bit about how latency factors into all this, and what are the overlooked benefits of being physically close to major network exchanges, cloud providers, carriers, it used to be that people thought I think that cloud adoption would make physical locations less important. And now we've got the rise of AI real-time applications and just a massive amount of data that has to move very quickly. Yeah, so it's interesting, 'cause I was a typical notion you ask anybody that would say it's performance, latency, but the biggest benefit is cost savings, because generally AI uses a massive amount of data. Massive. I mean, what we're just looking at charts here, a core site where two little platforms, say two cloud providers, from tier lab, and a mega-hyper-skiller interconnect of our facilities, they have about 77 terabits per second. That is enough to stream 10 million videos simultaneously. That is, people are focused on the computer layer, 'cause it gets all the news right on the computer, these very large sites, you can watch a capacity, but the interconnection layer is being built before our eyes. And if you're not in close proximity to where the compute takes place, 'cause data ultimately is computer-generated data. So adjacency is super important. Otherwise, you're transporting not terabytes of data across the country. That is enormous expensive, very difficult. So your ability to scale, if you're an individual user and your AI takes 30 seconds, it's probably okay. But if you're a financial quantitative trading platform that needs to access market real data and make instant decisions, or you're a navigation aid for a car that it takes accidents and rewards you, that latency is real. And so the more sophisticated the application, the more important is to be co-located with the compute takes place. Well, at least in very close proximity, because you're gonna have, you know, much more bandwidth and a lot less cost. Also, a lot of the cloud providers have established their terrorist structure so that you pay a lot more money for every time you put data out of the cloud. But if you're interconnected with, say, if I were cross-connected, then that share doesn't exist. So ultimately, the benefits are flexibility, scalability, and cost and also performance. But I would say performance is 40 seconds to cost or even feasibility. - Yeah, I guess we'd love for you to talk a little bit more about the interconnection piece because in my role, like what I do, coming in as a legal service provider, helping to get some of these transactions done and funded. We don't think about this piece as often, so I'm quite interested in it. Interconnection is described as a product or service. You talk about it like more of an ecosystem. What's the difference and what does that distinction matter? - I mean, the ecosystem is composed of products. So I don't think they're mutually exclusive, but a product is a cross-connect, for example. Cross-connect is a piece of Dart fiber that interconnects a customer A with customer B. You have a vertical D mark on one side, a vertical D mark on the other side, and you made that connection. And that is a product. And the core side has about 46,000 cross-connects right now. So those are 46,000 connections that customers are in and operating with each other. That creates the ecosystem. We have one measurement that we call ecosystem density that is the number of customers in our portfolio that are connecting to at least five different users. And that number is about 80% of our monthly revenue is customer like that. Let's just tell you that companies that come to core side become because they have to interconnect with their digital supply chain. And that's where the ecosystem is. So it is access to networks, it's access to different digital platforms, it's access to frontier labs, to NeoCloud, to cloud service providers, and all of your other fingertips can go right within 24 hours. That's what our ecosystem is. It's that ability to get to whoever you need to very rapidly and very quickly, I mean, any especially. - And I guess is it the way that you think about this and sort of put these pieces together or like what would you say is it that makes it difficult for your competitors to recreate these ecosystems into compete against you? - It's a great question. And we get asked that a lot by investors and such. There is a true mode of our campuses because it takes decades to build that integration density. When networks decide to put their points of presence in the market, they made a decision and deal with it for decades. You're not gonna have a certain network or a certain cloud put a non-ramp and then replicate that across 10 sites because it is super expensive. So when it's happening is the network effect where when you already have one clone ramp from player A, then player B collocates with you because you really have all the customers that are using player A. So this network effect is a flywheel 'cause it enables, you have a network that brings in the cloud provider, then a cloud provider in center prices which in turn brings networks and it goes on and on and on. And now if you look at the latest cycle of the cloud run ramps and the largest players out there, they really have no choice of where to collocate. They really have to collocate in places like us because that's where they can access their own ecosystem. And actually the cloud providers have massive speeds 'cause all the network edging fresh thrusters where they built these facilities. So it's like a self-fulfilling virtual cycle that takes place, but it is incredibly hard to replicate. - Yeah, what is currently your biggest challenge? I mean, there's a lot that's happening out there in the industry right now, both in public sphere, the political sphere, supply constraints, labor constraints. What is it that you guys are dealing with that you would say is your biggest challenge? - Yeah, the biggest challenge is and that should come as a surprise, our power. Simply because with the advent of AI, power power, you know, used by data centers have grown tremendously and we all fight for the same resources. Even course side is the last molar player. Our increments are order of magnitude smaller. Still, it impacts everybody. So it just takes longer. So we have to partner with utilities and just be much more thoughtful, put more capital at risk earlier on in the development cycle and that's fine. And the other element is public sentiment, is turn against data centers. And I think the challenge for me is we need to educate people. It's not their fault, they just don't know better. They think most of the public would say, like to your point earlier, all data centers have the same. The role of these gigantic buildings that are sucking only power and the reality is, course side is critical. Digital infrastructure supporting everyday lights. You cannot use your smartphone any given day in the US, course side. Everything resets the data center and we're enabling that integration to happen. But because of the public sentiment, you know, local politicians have to react and sometimes they're making it more difficult just to get approvals for development and things of that nature. I think everything is just a matter of education. And yeah, and again, it just takes another more planning, bye-bye. operating with local communities in which we operate I think we're being a pretty good neighbor generally speaking and we're continuing to do so and I think you'll be fine. If you had to pick one factor that will create a competitive advantage for companies like yours going into the next say five years would you say it is power or power ability access power location connectivity ecosystem density what would you say is going to be the most critical in terms of creating competitive advantage yes I mean facetious obviously being sustainable competitive advantage for core side is certainly the ecosystem like I mentioned before that just keeps building on itself and there will be new iterations new players you have the Chinese AI models and all these other technologies that will that will emerge and they'll collocate with us because they want to access our customers and then our enterprise customers want to access those models as well so that will keep you know the virtual cycle will continue the network effect is very strong but for that to exist for that to continue to grow you have to have power you have to keep buying land you have to have access to capital you know and one of the things that our partnership with American Tower brings is a very strong balance sheet because this is a very capital intensive business so the combination of access to capital just becoming a more mature developer generally and I think we're evolving towards from a more bespoke model with more initial markets to being able to go more into the suburbs with bigger campuses all of the things are are interplay with each other you have to have power but eventually power will be table stakes you know a like mission critical 100% up damage table stakes you want to be a reputable operator there's a number of them you have to deliver on those on those things what is really unique is to provide access are your fingertips with a very very wider array of of little platforms and networks and and cloud providers that are that are very hard to secure and to grow with so I think eventually that is the will provide differential returns on a longer sustained competitive advantage yeah I like that power is or should be table stakes it's it's it's become you know when we're in crunch yeah there'll be more investment in solar in nuclear I mean you know most utilities have not invested in in decades they haven't had to and a lot of sudden you have a conference of factors grabbing huge demand including electric cars vehicles and that sort of thing and the decommissioning of fossil fuel power plants so you have a little less supply you know more than man and that imbalance will get resolved I'm sure yeah I think there's just a timing lag because you know there are all sorts of regulatory processes that the utilities have to go through they aren't set up to move as quickly as what's currently being demanded of them but we are working with a number of power developers and utilities and data center companies and everyone seems to be moving as quickly as they can towards that goal of getting power the business model is that they're not known for the agility of the very very conservative companies generally and there's sort of been forced in a way to be uncomfortable and I move faster and but again I'm confident that maybe no in two three years but in a decade that will be it will be solved for sure well Juan I want to thank you so much for for being on the show today and for sharing your perspective and and telling us about your companies it's been great time you want and very educational for me well thank thank you for the for the invitation it be my pleasure thank you for listening to uptime now if this episode sparked an idea pass it along to someone new network and leave a comment I'd love to hear your thoughts ideas and recommendations for guests until next time thank you and remember to subscribe