Artificial Intelligence Optimization in the Data Center World With Brian Eichman

Full Transcript

Scott Kinka:

Hi, welcome to The Bridge, another episode of The Bridge. My guest on The Bridge today is Brian Eichmann. Brian is the VP of Business Development for CoreSite. Brian, it’s great to have you on the show.

Brian Eichman:

Yeah, thanks for having me, Scott.

Scott Kinka:

Looking forward to this conversation. And Brian, we had Miley Kaiser on the show last year. I know that you work with her. She’s quite a personality.

Brian Eichman:

She has a big personality. It’s gonna be tough to follow in her shoes.

Scott Kinka:

I appreciate that. It was a well traveled episode. Let’s say she must have a lot of fans out there on the interwebs. But for those who didn’t catch the last episode and just sort of set the bar from an expertise perspective, give us a quick read out on CoreSite. Who you are, what you do. Just give us a minute or so.

Brian Eichman:

Yeah, great. Yeah, so CoreSight, North American data center operator. We’ve been doing this for a little over 20 years now, publicly traded REIT. CoreSight, really, our specialty is delivering a purpose-built and highly scalable data center platform that really accelerates customers’ digital evolution. And really, what we focus on is delivering out in a robust ecosystem that makes it easy and cost efficient for customers to come in and integrate to their digital supply chain.

Scott Kinka:

Got it. So I’m going to break that down and put you on the spot. What’s the superpower? I mean, there’s a lot of that data center operators out there. Give us your superpower.

Brian Eichman:

Yeah, there’s two things that we bring to the table that’s very disruptive. One, we bring scale and density to our campuses. So CoreSight really likes to build ground up, purpose-built data centers that can support modern workloads and the power densities that are associated with them. And then we marry that scale and density with interconnection and make it very easy for customers to integrate over low latency connections, either to their networks and ISPs or the native cloud on ramps to providers like Amazon, Direct Connect, Microsoft Express Route, and things like that. We provide that on net low latency interconnection.

Scott Kinka:

Got it. Both of those things I had done some pre-reading, and we did some talking on the pre-show, are things I’m going to want to dig into a little bit. So one would be, we’ll talk a little bit about your ecosystem because it’s interesting, how the ecosystem is both Competition and cooperation. Or as we say, Co-op-etition, in a lot of ways. So we want to talk about the ecosystem. The other one I want to talk about is, I love the conversation about power density. Which modern applications have changed even maybe, even since we talked to Miley. In terms of who your big power consumers are. You recently wrote something super cool about AI. I want to engage in a little bit, but before we do that why don’t you just share with us. First off a little bit about your background, just who you are, where you are, how you got here, and then we’ll kick into sort of what does it mean to be the VP of business dev. But, just answer the first part of that question. Tell us who Brian Eichmann is.

Brian Eichman:

Oh, the existential question, man, that’s deep. I don’t know if we can satisfy that one in 30 minutes here, but guys, I’ve actually had a really fun journey. So career engineer, 20 years in IT engineering and strategy, everything from help desk all the way up through systems and network engineering. And I was, I don’t know, blessed or cursed to be one of those IT guys that didn’t mind speaking in front of people and being able to, in layman terms, explain technology and its application and the business value that you can extract from it. So I had an opportunity to pivot about seven years ago out of engineering and did a stint in product and then have moved over to sales, really just as IT has evolved over the last decade, distributed applications, cloud, newer technologies. It opened up an opportunity for folks like myself to get more on the business side and help customers along their own journey. And so in my role now, I manage a group of solutions architects. These are customers, consultants to make sure that they’re doing things as effectively and efficiently as possible within CoreSite, data centers and leveraging our ecosystems and products. But then as well on the business development side, it’s working with partners like Bridgepoint to evangelize CoreSite services and make sure your technologist knows how to apply our technologies for your customer challenges. And then on the other side of the business development coin is really working with our ecosystem partners like Amazon and Microsoft and Google and Oracle and others like that on how to come in, integrate into our ecosystem and ultimately cultivate new business relationships from the enterprises that are co-located with CoreSight.

Scott Kinka:

So we’re going to explore that, the latter part of that conversation in a minute. Cause I think the ecosystem conversation is really interesting, but I have one that maybe you weren’t prepared for. So I love that you were in product.

Brian Eichman:

Ooh, this’ll be fun.

Scott Kinka:

I have a passion for the product role, and I think that end users in particular, I think all these great technologies get built by engineers and then sold by salespeople. Talk about being in product and sort of where that sits sort of in the org in your world. And sort of it’s the indispensable nature of it at a tech company. I’m interested. I’m a reformed CTO myself, founded a cloud company myself, so I was in the middle of all that. Tell me about product development in your mind.

Brian Eichman:

Yeah, so I came in at a really interesting time. So data centers, I’ve learned over my almost 12 years with CoreSight, are very real estate-centric minded organizations. And they view everything through kind of the rent, occupancy kind of prism. But what we found is that the customer profile changed in my tenure with CoreSight. I’m sure we’re not alone in seeing this shift in customer profile. Where initially, the types of customers that leverage co-location were networks and ISPs and then sophisticated service providers, and then a few enterprises. But in that time, it has really evolved quite rapidly where now you had the advent of public cloud, and so you had these new cloud providers moving into colocation in a big way. Then as well, you had this last wave, which was the enterprise really shifting their data center strategy from legacy on-prem and customer-owned data centers into more co-location environments. So our product journey had to go along that kind of same flight path, if you will, where we had to shift our mindset and our product set from pure space and power solutions to customers looking for more advanced networking services. They were looking for other kinds of value-added services from their data center provider to just make it easier for them to deploy their own products and services. And so one of those product sets that I took over very quickly was our open cloud exchange. And this is Core Sites network automation platform that delivers software-defined networking capabilities. And CoreSight brought that to market in 2013 at the behest of one of our big cloud providers, Amazon. Amazon said, hey, we don’t want to just manage fiber cross-connects from our customers, we want to do more automated provisioning. CoreSight, can you develop something for us? And I just think Scott, as an organization, CoreSight didn’t really know the value of that platform, how do you evangelize it, how to sell it, how customers would integrate with it. And that’s really where I stepped into product with my IT background was really kind of bridging that gap from the real estate mindset of CoreSight to what our enterprise customers needed.

Scott Kinka:

Got it. Better yet creating the, what do they say? Sometimes there’s solutions looking for problems. I think the product’s job is to basically fit it to the problem in a lot of ways. Just because you can do something doesn’t mean you should, at the end of the day.

Brian Eichman:

You’ve got to be disciplined with product development. You can get backed into a lot of development initiatives, but you really have to be judicious about how you do it and making sure you’re placing the right bets.

The Surge in AI: Challenges and Opportunities in Data Center Capacity

Scott Kinka:

Agreed. So speaking of those bets, obviously one of those bets was the open cloud exchange, which was super interesting to hear that you were on that product team. Miley and I spent quite a time talking about it, but I mean, at its core, it’s a logical fabric for interconnection, from my understanding and knowledge of the product. You know, and at first it was focused on the hyperscalers, since it also includes access to the big SaaS players and some of the other network providers and things of that nature. But today you own that ecosystem. Let’s talk a little bit about that ecosystem. For people who don’t live in the data center world, they would probably think that Amazon and CoreSight are direct competitors. You know what I mean? And so tell us a little bit about what lands there, what lands with you, how you work together and give us the ecosystem story for CoreSight as it relates to the hyperscale providers.

Brian Eichman:

Oh, if I could put a number on how many PowerPoint decks or SLT meetings that I had to present at, is cloud friend or foe? It would be more than I could show on two hands. But I think that has been a huge boon for CoreSight, the cloud, and that works in two different ways. So… The cloud providers many times do not build their own data centers. In some cases they do. But many times they’re leveraging providers like CoreSight to deploy their products and services. Things that I can talk about with the big cloud providers is they like to use CoreSight for their network edge. And what you will know mostly for that is it’s where you peer your network with those cloud provider networks for internet peering. But it’s also where enterprises are privately connecting their applications with their cloud applications. We know this is Amazon Direct Connect or Google Cloud Interconnect, those cloud on-ramps. And so I’ve worked with those providers to deploy those services. And we’re really helping the cloud provider distribute their applications and their products, but we’re also making it easier for enterprises on the other side to integrate in a hybrid fashion to those cloud environments.

Scott Kinka:

So let’s assume somebody listening to this podcast is somebody who’s either wholly in hyperscale or wholly running on their own hardware, either in a data center like yours or perhaps even their own. What’s the story about them coexisting? Like in your mind yes, I get the connectivity, but why? You know, what fits where? And of course, every workload is different, but as a general rule of thumb, What’s the kind of work that you see in your data centers versus what’s living in the hyperscale providers when you have a shared customer.

Brian Eichman:

Yeah, no, I think that’s a great question. So going back to just my engineering experience, for me to be successful and what I’ve seen from my peers be successful, you have to be careful not to get overly reliant on a single technology or a single vendor. And it’s always about best fit for purpose. Those that get the opportunity to leverage multiple technologies to solve a variety of challenges, in my experience, are more successful at what they do. And just once again, just making sure you’re using the right vendor, the right technology, the right capability for whatever requirement or challenge you’re looking to address. That is what hybrid cloud or hybrid IT is all about. It is about leveraging the best vendor or technology for whatever your requirement or business outcome demands. And what we are seeing from experience is customers that maybe have, I think this is a good cycle. So all the SaaS companies that came out a decade or so ago, and they were new entrants. They didn’t have a great understanding of what their product absorption was going to be like. They didn’t have a predictable growth pattern. For their SaaS application, Cloud was awesome. It was a great place to get going. Many of these SaaS and startup types, kind of digital natives, if you will, didn’t have operations teams. They didn’t have a clear understanding of what infrastructure they were going to need to support their application. and they needed to be able to scale rapidly. The Clouds were perfect. But then they learned as they went along their journey, as they got a more predictable growth pattern, OK, here’s what I need to support my application and steady growth. And then they started pulling certain workloads and applications out and deploying that infrastructure inside of a data center. It gave them greater control of the underlying assets. They had more of a CapEx model, so they were able to reduce OpEx. And then they stood up their own operations teams that ultimately gave them a much better ability to respond to changing customer demands and things like that. So that was kind of the initial wave that we really saw.

Scott Kinka:

Got it. There, you hear this a lot depending on who you read and what you read. I mean, is there a scarcity of asset challenges in the data center industry right now that’s being caused by the hyperscalers or perhaps other new entrants?

Brian Eichman:

I would say that probably what you’re maybe speaking of right now is really a surge in AI. There has definitely been rapid growth in the AI industry, a lot of spending going on for just large chunks of data center capacity. And then that really comes from three different profiles of customers. One, yeah, you do have the big hyperscalers buying up large chunks of capacity for machine learning and AI applications. I would say not all of that is just around AI and ML for the cloud providers. They tend to buy in cycles every few years, so they buy up a lot of capacity, and then they can grow into that over a two or three-year time span just for their general cloud customers. So we see that kind of twofold from the hyperscalers. But then you have these new MSPs that are coming into the market. And these new MSPs are really specializing in AI and machine learning applications just because it takes a new skill set from a service provider perspective to address AI and machine learning challenges. It’s not just about having a big engine full of GPUs, there’s a lot of intelligence at the software level and just kind of at the engineering level to make AI and machine learning work the way it’s intended to. So you kind of need a specialized MSP that knows how to optimize software and hardware stacks for AI to hum the way we want it to. We’ve seen these from applied digital and core weaves of the world that are getting massive amounts of funding, taking down large chunks of capacity ultimately to address enterprise demand. But then the new one that we’re seeing, Scott, sorry to be a little long-winded here, is from the digital natives.

Scott Kinka:

No, go.

Brian Eichman:

These are the folks that are AI driven organizations. Maybe they have an AI product or capability that were born in clouds that are starting to say, hey, I’m collecting customers here, I’m growing. And I better start looking at more scalable, cost-efficient ways for me to deploy my AI capabilities above and beyond the public cloud. And we’re starting to have conversations with them to kind of shift into more of a hybrid kind of architecture, if you will.

Lessons from the Evolution of Cloud Computing

Scott Kinka:

I think there’s two things I want to read out with you on that in those categories. One is, I’m with you on the GPU. Can you just explain sort of the difference, why GPU is an important thing in machine learning models and then just sort of how that adjusts in a lot of ways your sort of physical asset to power ratios, like how that sort of offsets a little bit about what’s going on inside the data center. For those who just aren’t aren’t familiar with what tthe learning models themselves require. If you could speak to that for a moment, that’d be helpful.

Brian Eichman:

Well, I think as Scott and I kind of talked a little bit before the podcast I’m not an expert yet on AI. I am definitely working as hard as I can. I’m taking the challenge to learn as much about it as I can. But here’s what I can tell you succinctly. And I think in layman terms, this is where I’m at the end of the day, GPUs are more efficient at processing certain types of workloads than CPUs. Just kind of going back to using technologies that are purpose built and the best fit for the challenge, a GPU many times can just outperform a CPU for a variety of reasons as it relates to data analysis and analytics. And so with that, you have to design more high density types of power solutions to support these new processors. And to kind of put this into perspective, we used to see like the average density kilowatt usage in a cabinet used to be around five to six kilowatts a cabinet. It could be a little north of that, it could be a little south of that, but that was kind of the good average, if you will. And now with GPUs, high performance compute, AI, machine learning types of applications. We’re really seeing that tick up above 10 kilowatts a cabinet. And many data centers are just not designed to support those densities.

Scott Kinka:

Makes perfect sense. Thank you. That was the expert answer that I was looking for. Don’t sell yourself short. The reality of it is that these applications there’s the digital natives, which is the next conversation that we’re going to have. Those are kind of traditional front end workloads that are accessing the models. But the models themselves are eating GPUs and power to calculate up that database that’s being used by that digital native. So the bottom line is that you’re kind of the legacy construction ratio of power to cabinet has been set on its ear. I mean, it’s one thing to have the hyperscalers pre-buy traditional workloads because they’re just heavy buyers. It’s a completely different one to have all our set of heavy buyers come in and eat power and cabinets at a way different ratio than buildings were constructed for. You mentioned earlier that was a big focus of you guys having that, that level of density. I think it probably goes without saying, but I’ll clarify. Your core site is largely AWS data centers built for those customers, those AI customers, particularly those who are doing the calculating work with GPUs at this point, those are customers of yours.

Brian Eichman:

They are. And so that’s a good difference. There’s a lot of data centers out there and just because of how popular, some of the dynamics we talked about at the start of the call today is that there’s been a lot of new vendors coming in and starting selling data centers. They’re not all created equal. And CoreSight, just kind of in our journey. We really wanted to build purpose-built data centers. We like to buy vacant land and build from the ground up, so we can design to our specification to what customers are demanding for densities. It’s a lot easier to do it that way, to design it from the ground up without having to buy and refurbish a legacy office complex or a high rise building, because many times they just can’t support it, or it’s very costly to retrofit those buildings to support those kinds of modern applications. So CoreSight’s really been a leader in this space. We certainly have carrier hotel assets. Within our portfolio, those are necessary to provide the interconnection components of our business. But I think what we’ve done better than most, if not all, in our space is tether those interconnection assets like the carrier hotels to purpose-built modern data centers that can support super high power densities.

Scott Kinka:

Yeah. I think not having been in the business for a lot of IT people or partners or who may be consultants, who may be listening to this pod. I think it’s the cumulative effect starting from the workload on out that I think people don’t have a good feel for. So your point about building them up is AI requires calculation requires GPU, GPU requires more power. more power requires either additional power generation. It certainly requires additional cooling particularly as those are in smaller physical densities. I mean, there’s just, it’s this cumulative effect of doing it right. So speaking of doing it right, let me ask a different question. In the preamble you mentioned, Hey we messed up a lot of things in cloud, those of us who’ve been in this for a long time, and there’s sort of a next gen of new companies popping up in AI. So, before we get to the fun, which fun questions are coming up, what would be the advice that you would have for them having been in this business as long as you’ve been?

Brian Eichman:

Yeah, no, I think that’s great. This one means a lot to me because of my engineering background. I always like to look out for my brethren on that side of the house. But learn from your predecessors. And AI is very new, but there are some really great and valuable lessons that we have learned over the past decade with this shift to cloud that we can repurpose for today’s world with AI. And it is being flexible with technologies. And vendor management is extremely important. And not just looking out at the business that you have to support today, but being mindful of tomorrow, and always questioning and looking for how to do things better, faster, cheaper, if you will, in the future. And I think I was just recently at the AI4 conference in Las Vegas in early August. And I met a lot of folks, got an opportunity to talk to a lot of these digital natives. And I came away from that going, I think that there’s going to be the folks in that kind of profile that are thinking about tomorrow as much as today. They’re going to be around the next AI for that’s right. There’s a lot of new names, new companies popping up overnight. I kind of bring it back to the dot com days. There were companies born overnight, but there were a lot of companies that didn’t survive overnight either. It was those companies that were thoughtful about their scale and were always looking for better ways to do things. It stuck around and really made a name for themselves. I think some of those same lessons can be applied to the new logo showing up in AI.

Insights on Business and Technology Trends

Scott Kinka:

Sure. Or said a slightly different way, just because you can, doesn’t necessarily mean that you should. I think in a lot of ways, look, at some point, the business on paper fiscally has to make sense. I know that at the beginning of any industry like this, a lot of money flows in with no real intention of making money, just innovating and selling off. You know what I mean? But you’re not one of those fortunate ones that there comes a day when you’ve got to operate. So it’s super interesting. I love that article. What you put in, what you had on your blog. We’ll make reference in the show notes. I’m nodding off to Gene here. That was a great blog entry. Appreciate that. Can we just shift to some fun before we wrap? Is that okay?

Brian Eichman:

Yeah, let’s party.

Scott Kinka:

All right. Three questions. First one, and it doesn’t matter what venue this is. I’m just looking for some kind of shameless prediction. Tech politics, sports entertainment. Doesn’t matter. For the next 18 to 24 months.

Brian Eichman:

I think the Chiefs will win their next football game.

Scott Kinka:

Next football game?

Brian Eichman:

Yes, the next football game.

Scott Kinka:

All right. You didn’t go crazy with that. I thought you were going for their next Super Bowl, but I’ll take it. Care to conjecture on the Super Bowl then? Is it going to be number two in a row?

Brian Eichman:

Oh, I think that’s gonna be a tough one. We’ve won two out of the last four. So I don’t wanna get selfish here. I’m trying to stay humble.

Scott Kinka:

All right. I like that. You definitely have a quarterback who I think focuses a lot on staying humble, so we’ll take that. All right. Second one, somewhere in that office of yours is some kind of business reading that’s going on right now. Do you have any recommendations for reading? Maybe you’re on AI or something else that’s interesting to you that you would recommend to our viewers and listeners.

Brian Eichman:

Yeah, so from a tech perspective, not as much. So here’s just kind of how I spend my day. I try to find 20 to 30 minutes every day to read, different sources, if you will. That could be the register, that could be data center knowledge. There are a variety of industry sources. And I find myself spending 20 to 30 minutes every day going through those, finding out trends. I love looking at product announcements from the cloud providers many times, because as a key partner to those cloud providers, it’s many times what they’re doing on the product development front that translates to new opportunities, not just for CoreSite, but for your technologist as well. So I like to spend some time on that. And then on the reading side, oh, heck, I kind of did this backwards. There was a new series that came up on Apple called Silo. And I’ve just found it very intriguing, not to tell you too much about it, but human civilization now lives underground in a series of silos. And we start becoming aware and questioning why and how we got underground to begin with. And it was actually a book series before it came out. So I’ve watched the whole series, but now I’m going back and reading the books and they’re fascinating.

Scott Kinka:

So sort of like an earthbound version of the Matrix. Like you’re as opposed to being a tech version, you’re like, we’re literally in the dirt here. We just are not sure why we’re here. There’s another world out there. Interesting. I’ve seen some ads for that. I will definitely have to check that out. All right, last one. Let’s put it in that the machines take over. There’s some dystopian event and there’s one application that’s still working on your phone. You get to choose which one it is, which one and why.

Brian Eichman:

Oh, that is a fun question. And probably not all that hypothetical for that matter. I’m actually kind of thinking they’ve already taken over at this point.

Scott Kinka:

Could be.

Brian Eichman:

So on my phone right now where I’m at, I love coaching youth sports. I’ve got three kids, a couple of whom are in soccer, and I coach them. And I find strategizing a lot of fun on how to beat the team that next week. So right now I am using an app called Play Metrics, which teaches me different coaching strategies and tactical formations on how to win the next soccer game on the weekends.

Scott Kinka:

Brian, that tells us a lot about you. You’re playing moneyball with youth sports here. I appreciate the approach.

Brian Eichman:

We play to win, that’s for sure.

Scott Kinka:

I get it. So sports and tech all wrapped up. You’re a man after my own heart. I appreciate that. Brian, this has been a great conversation. I really appreciate the time today. And it’s been super interesting getting to know you. We will include some of your blog entries in the show notes that I mentioned. But before we go, if anybody wants to find out a little bit more, outside of engaging with their friendly neighborhood Bridgepoint strategist, how would they do that?

Brian Eichman:

Now there’s a few different ways that you can engage us. So one, just approach me directly. I love to talk about this stuff and would be glad to give you any more information I possibly can over a phone call or coffee. But coresite.com is a wealth of knowledge. You can have use cases, customer testimonials, see the different products and challenges that we’re helping solve. We run social channels on YouTube and LinkedIn. And then lastly we’re out and about in the field. You know, we’ve got a good one coming up. I’m always excited to do it with Bridgepoint. This year, it’s at a new venue down in Palm Springs. I’m gonna miss Ohio, but you know, Palm Springs isn’t a terrible place to be either. So I think I’m gonna be on a panel or so at that event and you know, we’ll certainly be there shaking hands and meeting new technologists and things like that during the show. So you know, there’s a lot of different ways that you can engage with us.

Scott Kinka:

Perfect. And I appreciate you saying that we’re going to miss Ohio also, we’re just too big for that resort now, which is what it boils down to. So we’re excited to have you there, at a significantly expanded agenda. So I’m really excited about that as well. I will see you there in a couple of weeks, Brian, and we’ll have a cocktail and extend this conversation at that point. But thank you so much for being on the bridge today. Thank you so much.

Brian Eichman:

Yeah, thank you, Scott.