A Conversation with an iMason
Dean Nelson talks about Infrastructure Masons and all things data storage in a follow-up conversation with Raymond Hawkins.
Raymond Hawkins: Okay, well welcome again, to another edition of Not Your Father’s Data Center. We are joined again, back by popular demand, Dean Nelson, friend and colleague, and also… What’s the right way to say… He and I come from the same generation. I think we’re within a year of age. So life experience is the same and Dean carries lots of titles, wears many hats, CEO of Virtual Power Systems, as well as the founder of iMason. And Dean, happy to have you, and thank you so much for coming back.
Dean Nelson: Yeah, thanks for having me back. I always loved our conversation, so we get to do it again.
Raymond Hawkins: Yeah, I appreciate you getting to chat with me. It makes my job really easy. I ask two questions and then everybody gets to listen to how smart you are, and it makes it go really easily. So, not every podcast is that simple.
Dean Nelson: Usually they say I talk too much, so thank you for the positive nature.
Raymond Hawkins: Yeah. Not a problem when you’re recording a podcast. We need you talking. So that’s excellent. Well, man, I know you and I have chatted at several times. Last time you were on, you did a lot of VPS and talking about how managing virtual power. If you’re willing, I’d love to do a little bit of a revisit of that. How much we want to talk about virtual power systems, how much you want to talk about iMason, or any other subject you like, I just love listening to how you think and how you see the world. But if you’d review a little for folks that don’t know what virtual power systems is, if we could start there, that’d be great.
Dean Nelson: Yeah, Virtual Power Systems is a software company. And what we’re focusing on is unlocking stranded power in data centers. So that’s the basics of it, but I’m going to broaden this out a little bit for your audience because there’s been some really cool movements, I think, in trends that have been happening since we last talked. And the way this works is that, if you think about our industry, today we have 7 million data center locations. So think of that as unique addresses. These are places that have some type of compute in them. And there are about 104 gigawatts of capacity built. So 104,000 megawatts of capacity built globally in those 7 million data center locations. That’s capacity that could be used. So then, from a consumption standpoint, there’s 594 terawatt-hours of consumption. And if you now equate that, that’s 2.4% of the total energy draw globally.
Raymond Hawkins: 2.4%, [inaudible 00:02:53]
Dean Nelson: This has been a contentious discussion from so many different people, right? Because how do we actually define what it is that actually is included in that number? And so, through iMasons, we created this baseline on purpose. And what it includes in this is basically three categories. So digital infrastructure in general is providers networks and crypto.
Raymond Hawkins: Okay.
Dean Nelson: All right. And we created that on purpose because providers is basically any colo, enterprise, cloud, edge, hyperscaler… Just anybody that is providing a service either to others or to themselves. So we lump them all together.
Dean Nelson: Then we’ve got networks, and this was not really considered before. When we talk about “data centers“, they excluded the network. So think of fixed networks, broadband, telcos, all of those aspects. And then the third one was we haven’t been considering crypto and all the blockchain elements that are driving crypto. And so right now, 80% of crypto consumption is Bitcoin. It’s the math, the hash rates that are going in there to actually discover the actual coins. And so, but if you look at and break that down, 216 terawatt-hours is the providers. That’s cloud, that’s everybody. 216 terawatt-hours. 266 terawatt-hours is network. That was a big surprise. People didn’t expect it to be that big.
Raymond Hawkins: That’s a big number.
Dean Nelson: Yep.
Raymond Hawkins: Because you tend to… You think of two post racks and little bitty locations and not a lot of power and any sort of your vision. I am shocked to realize that one’s bigger and bigger by 25%. It’s not just a little bigger. Wow.
Dean Nelson: And the thing is, that network was here first. We didn’t have these big core data centers. We had [inaudible 00:000:41:48] Offices and all this stuff all over the place. It still is there, and more and more of it. So think of also the wiring closets and the IT aspects and buildings. And there’s tons of stuff out there, but network was a big chunk. And this is controversial, right? The way people are pushing back on the numbers. Which is great, because in the end of it is the category, right? Yeah, providers, network, and crypto. Okay, those are the three big consumers. If you want to move things around, doesn’t matter. It’s still the total amount of consumption.
Raymond Hawkins: So I’m doing the math quickly in my head. I’m assuming this leaves us with what? So that’s 482. So that leaves us with 124. Is that what it is?
Dean Nelson: I think it’s 114, 116, something like that. So 216, 266 and 114 if I remember correctly, but that’s crypto. Now the important thing about crypto is that just four months ago, it was 140 plus. China made a decision to no longer enable it. Well, what happened is it shut down in China, but it all is moving.
Raymond Hawkins: It’s coming online somewhere else.
Dean Nelson: It’s going to Europe. It’s going to the Americas. It’s going to [inaudible 00:06:01]. It’s going all over the place, so that number’s going to continue to increase. But think about it. 114 terawatt-hours is crypto. It’s not going away. The reason I bring this back up is if you look at that’s 594 terawatt-hours. Okay, great. But we built 104 gigawatts of capacity. How much capacity in any given second is consumed in that 594 terrawatts? Those terawatt-hours translate to 67 gigawatts of energy consumed of the 104. That means we’ve got at least 37 gigawatts unused globally.
Raymond Hawkins: Yeah. Built infrastructure, not being used.
Dean Nelson: Right. And why is that? That’s because of buffers and buffers and everything else that happens from enterprise applications, to cloud, to data centers. And everybody’s saving to ensure that they don’t have an issue. The data center buffers it. The hardware team buffers it. The actual shared platform, buffers it, the applications buffer it. Everybody buffers.
Raymond Hawkins: The power company buffers it. Everybody, yeah.
Dean Nelson: Right. Because they can’t have the state where they don’t have what they need. And so the example here that I use is that, by the way, that’s conservative. 37 gigawatts is assuming that we have 60% utilization of built capacity for actual providers and for network. We don’t.
Raymond Hawkins: In your three buckets, which I really think that’s a great categorization, I like the categories, providers network, crypto. How do you count for when a crypto is in a provider? I’m assuming you’ve done the math and not double counted because there are cryptos that sit in providers buildings. How did you account for that?
Dean Nelson: Right. Not much, by the way. And the reason is they don’t need that. And the majority of them don’t want ups, generators. They don’t want any of that stuff.
Raymond Hawkins: They don’t want cooling, they have a very unique-
Dean Nelson: And by the way, this is the debate that was happening. Well, what if it’s a carrier hotel that has a colo in it? Okay, great. It’s a location. That’s going to have a segmentation of each.
Raymond Hawkins: It’s going to have some network and it’s going to have some provider. Yeah, exactly.
Dean Nelson: But the difference here is that it’s a location. It’s a address that’s unique. That unique address has a certain amount of capacity built and it has a certain amount of source energy. So you should be able to now add up all those things and represent it the right way. But the thing that’s important is we need to have unique addresses. Just like we have unique IP addresses on the network. There’s no duplicate networks. NAT hides it all, but in the end of it, on the internet, everybody’s unique.
Raymond Hawkins: Yeah you get a [inaudible 00:08:27] and a fire that goes with it.
Dean Nelson: You know exactly where it is. So data centers need to be the same way. If not, we can’t have a baseline measurement to show progress. So this has been an iMasons effort. And I’ve been working with Rob Aldrich. He’s the chair of the sustainability committee for iMasons. And we’re finishing up this paper to publish about this. And I’m just sharing the numbers with you.
Dean Nelson: And Eric Massinette is with UC Santa Cruz as well. He has worked with DOE and a whole bunch of companies and other academics on getting these baseline numbers. And this is how we’ve now rolled them all together. So we’re starting from there. And what I care about is we have a starting point. We got a baseline. Now, the reason I bring this back up is that if we’ve got this much stranded capacity out there, what are we doing with that stranded capacity?
Dean Nelson: And so I look at data centers around the world and we sell one product today. We sell a five, nine SLA with a certain term, with a certain efficiency at a certain price. And the majority of them are the REIT structures. So they care about the asset. They are not driven by the utilization. I build a data center, I sell it. It’s used 60%. It’s a hundred percent sold. I build another one.
Raymond Hawkins: That’s right. It is a hundred percent sold from an iGenerate NOI on my asset from a dollars perspective. There’s not really a technology understanding of it. What did I build? What did I utilize? What’s the utilization? It’s rather, “Hey, I generated the rent out of that asset I needed. And I’m good.”
Dean Nelson: Right? Exactly. And again, we understand why the rates are driven by assets and the rollover of the money that has to go in because of their structure. So there’s the reason this happens is because they have to report in a certain way and utilization isn’t something that’s going to move the needle on the other side. So tons and tons and tons of money is coming in. But to me, from a sustainability standpoint, that’s not a sustainable model. We’re basically perpetuating low utilization globally. And that’s going to happen in these emerging markets too.
Dean Nelson: So what we’re we care about is can we help unlock that stranded capacity? Can we still keep the models that work for REITs, but also bring them additional revenue. But I gave you this baseline so you get an idea. There’s 37 gigawatts capacity that is translates to about 222 billion worth of investment. That is not yielding returns. There’s money left on the table. From a sales standpoint, I should be able to sell that capacity.
Raymond Hawkins: Yeah, well, so hold on. Let’s talk through it, Dean. I get what you’re saying in the sense it’s not generating return because it’s not being utilized, but in the calculus of the REIT, they built the building. They generated the rent they wanted on the building. And in that calculus is some stranded capacity, is some overhead, is some protection. So I get your point of it’s not generating revenue, but from a REITs perspective, “Hey, I got the return on the building. I wanted to get.” What I hear you saying is, “Hey, that’s nice, but let’s look at it. Let’s look at the same picture from the other direction and go, wouldn’t it be nice if you could utilize all of that asset.”
Dean Nelson: So, basically they’ve achieved their objectives. They have an asset that has value that’s predictable over 10 years. That’s what the REITs really want. That’s why they invest in airports and power generation and all that types of things. They’re infrastructure assets that are predictable, but if you flip this back around and you think about the people that are investing in these companies, they are ESG focused the investment in sustainability. And so we’ve got this disconnect between that. And so when you start to expose this to the actual investors to say, “there’s money left on the table, but there’s also a large embedded carbon footprint. That’s exacerbated because we’re not utilizing what we built.” That’s going to be counter to what’s happening because Blackstone, BlackRock, Macore, like all of these really big investment firms.
Raymond Hawkins: Yeah, they’re all ESG focused. No question. Yeah, “How do, how do we, how do we make sure our dollars are being spent in a planet wise fashion?”
Dean Nelson: Right, and for us, if you think about it, okay. We’re 2.4% of the total energy draw for the world, but that’s digital infrastructure enabling everything.
Raymond Hawkins: Yeah, exactly right.
Dean Nelson: It’s great. But we’re a technology. We should be able to know exactly what the real time carbon impact is of digital infrastructure. So by having this, we now can be able to see unique addresses, the actual built, the type and the energy sources of it. We should be able to calculate carbon footprints.
Raymond Hawkins: Right. So hold on, Dean. I want to make sure… I want to talk something out and make sure I’m tracking with you. So what I hear you saying is, and the 65% or 37 gigawatts of capacity that we’re not using, what I think I hear you’re saying is two things.
Raymond Hawkins: One is, there’s a bunch of capital investment in that equipment, in that plant and equipment that is prepared to provide electrons to digital equipment. We spent a bunch of money on it. We delivered a bunch of it, and it’s not delivering any use utility to the digital infrastructure world. What you’re not saying is, “Hey, we’re wasting the energy because, if there’s not drawing, we’re not spending energy.”
Raymond Hawkins: What you are saying is Raymond, what about all of the energy and all of the effort and all of the plants and equipment and all of the raw materials and all the product that have gone to build a third of the global infrastructure, that’s sitting there doing nothing. That’s what you’re talking about.
Raymond Hawkins: Imagine if we, I don’t know how many cars are in the world, but imagine if a third of the cars in the world never drove anywhere, that’s really what you’re saying. Did we need that? I don’t know how many cars there are on the planet. It’s a bad analogy, but did we need that other 1.2 billion cars that just sit in people’s driveways and go anywhere? And that’s what you’re saying, right?
Raymond Hawkins: Yeah.
Dean Nelson: There’s a disconnect here. So, what I, and the reason I’m bringing this back up is I want people to understand that there is a large pool of capacity that is not utilized, and then we’ve got a model that we do in data centers right now that we sell the RFP process, the contracts and everything else is that it’s driving a certain behavior. And I’ll give you an example. When I was at Uber, we had a full region fault. We lost the West Coast done, out. And so we had the East and the West serving the world. And when we lost the West, all of those things had to fail to the East.
Dean Nelson: So think about this, when Uber was basically trips, eats, freight, any of those things, they served cities out of a region. So Paris and New York and LA are served out of West. So when we fail, that actually goes and shifts to be served out of the East. When that failure happened, the total increase in draw in the East of half, the world’s capacity, migrating to the other region was less than 10%. It was single digit increase in capacity consumption.
Raymond Hawkins: So you’re pointing to an overbuild environment? Again, buffer against buffer, against buffer.
Dean Nelson: Think of the server. In the server, they put power supplies, the power supplies are oversized because all these components could be in there. I could have more CPU-
Raymond Hawkins: That’s right. And they also could be running 100%. That’s right.
Dean Nelson: But they don’t, and so now you’ve got this buffer of the power supplies. Then you have the buffer of the data center. Then you have the buffer of the utility. But then, you take the next layer and say, that compute is shared in this shared compute layer. They buffer it by sizing it appropriately to say certain virtualization, which means the buffers go up that direction. Then you have the applications to say, “well, I have to have this much.” So they buffer that. Then you have resiliency of the application. They say, I must be in three zones in a region and have multiple regions. So you can imagine how much waste is in that stack. It’s like the electrical distribution system generation, and how much is lost by the time it gets to where those electrons are consumed? Huge amounts, transformation, distance, all that stuff.
Dean Nelson: It’s the same stuff here. We build in all those buffers, because “we have to have the resiliency.” And what I’m telling you is that every company in the world has the same architecture. I haven’t found anybody that has something different. It just comes down to how efficiently they operate it. So there’s a thing called replication factor. And that basically means that if I have two regions, I need to have 220% of my capacity. Because I have 100% one region, 100% in the other. Then I have 20% just in case in one and 20% in the other, 220. So that’s a replication factor of 2.2. If I now do three regions, my replication factor goes down to 1.8. If I do four regions, it goes to 1.7. It keeps going down to where, if you’ve got enough distributed capacity, you could have only 1.2% of the capacity. You need to have the same resiliency.
Raymond Hawkins: Yeah, because it’s divided over multiple ways to essentially fail to replicate.
Dean Nelson: But this is a systems architecture challenge of how people go in and do this. And it’s complex. It’s hard to get down to those things, but people have been spending a lot of time in that. My point is that there is at least half of the industry, half of the companies in the world have this legacy approach where they’ve got so much excess. So all this translates back into it, manifests in the data center of capacity. This is not used.
Raymond Hawkins: Right, and Dean, you’re talking about changing a global mindset. I mean, that’s really what we’re talking about here, right? You’re saying, “Hey, we got here in a legacy way. I understand how we got here, but is it how we want to go forward?” That that’s really what I hear you saying. And how do we begin to change the mindset that I don’t need redundancy on redundancy on redundancy, even though that’s how I started this business. I meaning the industry. That’s how the technology space started.
Raymond Hawkins: When we first started running computers and wow, that’s really important. And something happens that’s a problem. Great, we’ll do two of them. And that was the response, right? And oh, if we’re going to do two of them, well heck then let’s put them in a raid array and let’s have a whole group of them. And that’s what we’re talking about.
Raymond Hawkins: And essentially it’s a crude analogy and I’d have to have my data storage friends from days gone by. But essentially what I hear you saying is, “Hey, let’s raid R-A-I-D, the digital infrastructure, not just the storage subsystem.” That’s what I hear you saying, because you can manage your replication factor dramatically better in a RAID array than you can in the way the data center industry is built today.
Dean Nelson: Number one, the only way that works is with software. And if you think about it, we hit a wall in compute about 10 years ago, and that was dedicated servers, dedicated storage, dedicated network. And what did they do? They came up with virtualization and then virtualization became containerization. Kubernetes came out, so orchestration started happening and they were able to use software to now get more out of it.
Raymond Hawkins: There’s, there’s no question. That’s exactly where I’m going way back in technology to use the RAID example, but that’s it. And it used to be, if you needed a 100 megawatts and you wanted it fall tolerance, you got a second 100 megawatts and we started going, “oh no, that’s not the best way to do it. We don’t need two of the same.” And it started at the storage level and it’s cascaded up the stack of compute. And I hear you saying, let’s keep going, let’s take it into the data center side of the compute world as well.
Dean Nelson: This is nothing new. If you think about it, they virtualized the compute, the storage and the network. We have not virtualized the power. That power plane is still dedicated.
Raymond Hawkins: You’re saying compute, let’s just go through the computer. Because I have an answer I give folks when they say, “Hey, Raymond, do you think that the tech, the data center industry’s going to keep growing as computers get smaller?” And I’m like, “well, hold on a minute, we had three major events that shrank the computer.” And what were those three major events?
Raymond Hawkins: The first was… Four major events, excuse me, not three. First was, and not in order, we virtualized the operating system layer. We took utilization from 12% in servers to 90% in servers So, we got to use a lot more of the server by virtualizing the operating system and slicing all the assets in the system. How are we able to do that?
Raymond Hawkins: Because we’ve got to put more cores on a single processor. We used to be one core, one processor. We got multi-core processors. How do we manage that? And how did we handle the intensity of multi core processors? Well, then we virtualize the memory. We used to, excuse me, virtualized the storage, all our storage used to be spinning platters, which have all kinds of limitations on their speed because a platter is spinning, electricity’s being used. And we virtualized that by putting them on flash drives. So we moved, we made the huge shift from physical spinning platters to flash drives.
Raymond Hawkins: And then we are now doing virtual networks. So those four components all fundamentally changed the physical footprint, the energy used and the utilization level. And what I hear you saying is “Raymond, all those things let’s do it outside the server, outside the rack, let’s do it at the digital infrastructure level.” And it makes complete sense because you’re right. We’ve already done all of that for lack of a better word inside the computer. I mean, that’s a terrible description. But yeah, that’s what I hear you say.
Dean Nelson: You’re you’re right on. We just haven’t applied this software methodology, software defined power, to the power plane. We haven’t virtualized the power plane. And this is, I think people have a hard time getting their head around it because like, “well you can’t virtualize power.” You can virtualize power. And here’s what happened. Think about a server. You just outlined it’s still physically bound.
Raymond Hawkins: It’s no different than virtualizing the processor. It’s no different. You just think about it differently. I like the way you said it. There’s a virtualization layer and then orchestration layer on top of a physical asset. Well, if you can have that on a microprocessor, why can’t you have it on electrons pumping through the wall? It’s the same, it’s a physical asset that has its own physical constraints, but you lay a virtualization layer and then orchestration layer on top of it so that you can utilize that asset more efficiently. That’s all you’re saying. That’s what you’re saying. And that’s what you’re saying, right? I mean, let’s lay a virtualization, orchestration layer on top of the physical asset of power.
Dean Nelson: So we have an intelligent control of energy that basically is a virtualization layer. And what our software does is we start with measuring. We’re looking at what power is being drawn where. Then we’re assessing within that one. Is there anything has changed? And then we’re taking action.
Dean Nelson: So the difference here is that today when we sell power, you sell the power and it’s just dedicated to somebody. But that power itself is not all used. So if you’re now saying I’m going to stack rank my virtualization of my power, this manifests in SLAs. Today, we sell a five, nine SLA. That five nine SLA is for that 10 megawatts and that’s it. And however you utilize, whatever it is. But assume that they’re using 50% of it. Okay, well, I also have that 10 megawatts in a five makes four design has another two megawatts of redundancy on top of it.
Dean Nelson: So I have 12 megawatts and I’m using five. You have seven megawatts never used. So you offer up additional SLAs, imagine a two, nine SLA. A two nine SLA means 1% of the month I could have an outage, a two nine SLA is seven hours a month, if you have an issue. So I sell another five megawatts of two, nine SLAs. I’ve already sold out 10 megawatts of five nine. They use 50% of it. I sell another five megawatts of two nine SLAs. Let’s assume they use 50% of it. That’s seven and a half megawatts of 10. You still have another two and half megawatts headroom. So you can oversell this capacity. And then what you have to have is the relief valve. And this is what happens in cloud, and I’ll explain that in a minute. But if you’ve got a two nine SLA, that means that if the other ones are starting to consume, you shed the load of the two nine SLAs. It’s part of the contract.
Raymond Hawkins: Because he’s agreed to it. That’s right.
Dean Nelson: Because today, we basically say you can have any color you want, as long as it’s black. [inaudible 00:25:20].
Raymond Hawkins: As long as it’s five, nines, you can have whatever one you want. Thank you, Mr. Ford.
Dean Nelson: There you go. Got model A or model T, yeah. So, when we now start saying that is not the way that applications work, let’s go into cloud for a second. Cloud today has a flexible model. They call it elastic compute, elastic storage. Why? Because it can actually grow or shrink.
Dean Nelson: So what you do is you say I’m going to spend X amount of dollars and then I’m going to use the cloud. So that manifests into three blocks for infrastructure capacity in cloud is basically reserved instances, on demand instances, and spot.
Dean Nelson: So those three blocks, reserved instances are just like reserved power in a data center. I have them. I pay for them. If I use it or not, it doesn’t matter. You’re paying for it. So that reserved capacity, an instance, by the way, for people that may not know, an instance, is a virtual machine. It’s virtual CPUs and memory and storage and a network thing. You’re carved up this virtual system, this instance for people to use.
Dean Nelson: So I’m going to get reserved instances. They’re mine, they’re dedicated to me. Then they’ve got on demand instances. If you suddenly need more, you can burst into them. You pay more for it, but it’s there when you need it. Then they have spot. So spot is basically the capacity is not used sold on an open market, but they have a two minute grace period.
Dean Nelson: Basically, if they need it back, they’re going to tell you. And they shut down in two minutes. So people architect to this already. They have applications aligned, services aligned to the different types of capacity. So take this concept, reserved, spot, and on demand instances, translate that back down to data center, reserved power, on demand power, and spot power.
Dean Nelson: So the way that manifests is I’ve got a five, nine SLA rack that rack is reserved power with on demand, power that they can burst into. It’s still five, nines. So you sell that, you basically have this capacity, that’s going to be there all the time. They’re consuming and they can burst in as they need it. You’re basically giving them the insurance policy, the buffer that they would normally do. You provided this burst. So I’m talking about the colo.
Dean Nelson: Then you have spot capacity, spot power. You land the racks and you allow them to consume what they want when they need it. So reserved is based on kilowatts rent and then consumption of power and PUE. On demand is kilowatt-hours. It’s a loaded cost when you consume it, you just pay for it. It’s metery right on spot. Same thing. I’ve got the ability now to put in compute. And when I use it, I pay for it Well, what that works out to be is all of a sudden, you’ve got these things. You can roll five nines and two nine SLAs. Spot is two nine. Spot is the stuff you’ll shed to ensure the five nines are never shut off.
Raymond Hawkins: This is back to your 10 Meg 12 Meg example. What you’re saying is “instead of me trying to recover all of the costs of those other seven megs in the rent per kilowatt of the five megs, why don’t I offer that at a one price and then offer the other two spot and on demand at another price and being able to say to the guy, “Hey, you can have this capacity, but you have to agree to this level of SLA. Hey, the reason you get it for a 40% discount is because it comes to the radically different SLA. And, and you understand your…” Frankly, it’s a little bit, this is really tortured now, it’s a little bit like subordinated debt. You understand that when that business has a problem, all right, the bank goes first and this creditor goes second, and you go third. You’re just agreeing on the front end, “hey, for that huge discount, I’ll take third spot.”
Dean Nelson: I’m third. Yep. And by the way, the beauty of this is that it’s not like they have to rethink everything. They do this in cloud every day.
Raymond Hawkins: That’s right. It’s already part of the cloud infrastructure. That’s right. It’s how they think about cloud infrastructure.
Dean Nelson: So with the way that these companies go in, they architect and orchestrate services to leverage, those different “SLAs and cloud”.
Raymond Hawkins: So their cash register application is running in five nine. Their HR application is running in two nines because if it’s out for seven hours this month, it’s okay.
Dean Nelson: Right.
Raymond Hawkins: It’s not costing us any money.
Dean Nelson: The RTO that returned operations is-
Raymond Hawkins: My apology to HR professionals. I’m just saying, it’s not the cash register.
Dean Nelson: You can wait for a bit for your vacation to be registered, you can’t wait for a payment.
Raymond Hawkins: Exactly right.
Dean Nelson: Every company has these different service levels. There’s different services, their services need to run. And so what we’ve missed in the data center industry is mapping that. So what we do as VPS is a long way to get around to this thing is we unlock that strand of capacity by enabling cloud like flexibility on Preem through, reserve power capacity, which is really right size contracts, on demand power, and spot power.
Raymond Hawkins: So, Dean, I got to believe that my friends in the co-location business who are trying to manage multiple customers, multiple SLAs, multiple environments with a single set of infrastructure that this has to be viewed as revolutionary and just an incredible opportunity. You don’t have to name customers. We’re not asking for competitive or proprietary information, but how is this getting received in the marketplace?
Dean Nelson: What’s great about this is that we pivoted the company earlier this year. We were building hardware for a long time, and the hardware we had was where we protect the grace period, because again, a spot rack that goes in would have two minutes, two to five minutes of grace period, which means you need to hold it up for that period of time. What we found is that again, because of the utilization, the majority of them don’t need additional energy. When you’re at lower utilization, you’re using the capacity that exists. So you can give the grace period and shed it down to get rid of it. Because you got that headroom. But we had built hardware that allowed us to do energy injection. We could inject current to provide that hold up time. It’s still needed, but it’s when you start to get to higher utilizations. So let’s say-
Raymond Hawkins: That’s later in the cycle.
Dean Nelson: But so we’ve got partners now CE + T and Schneiders and other, they provide the capability of the hardware. We don’t need to do that. There are partners that do this really well, we just need software control of those components and those are our actuators.
Dean Nelson: So there’s really three big things. What we built before was energy injection, to add energy in parallel to protect the upstream breaker. The second was phase balancing because things get out of phase based on workloads and all the other things. So that is in parallel as well. And then intelligence switching. The only way that this is going to work is that you have the ability to enforce the two nine SLAs.
Dean Nelson: So think about this. We’ve got smart RPPs that allow us to be able to do switching of actual breakers. We’ve got smart cans that allow the same thing and we have smart rack PDUs that do the same thing. So we can literally shed the two nine SLA to protect the five, nine SLA. From our standpoint, what we care about is software will orchestrate those elements. If software’s not there, it doesn’t work. But we also need the hardware components to actuate to enforce the SLA. Does that make sense?
Raymond Hawkins: Yes, absolutely it does. Yeah, and I’m making a tortured analogy, but this is the same thing that happened when we started to virtualize machines, I’ve got to be able to manage those assets. I’ve got to be able to decide who gets what, and then I’ve got to be able to orchestrate it when it happens.
Dean Nelson: Right. Exactly. And so this is a transformation or I think a disruption in our industry and it’s not as scary as people think it is. Because think about it. Most of the colos right now still oversubscribe at risk.
Raymond Hawkins: That’s right with no controls, no orchestration layer, no controls. They’re just doing some, “Hey, I think I can do this.” And what I think, and I don’t want to use the word getaway because that implies a negative, but you know, “Hey, I think we can manage to this.” And what you’re saying is don’t think let’s actually manage to it.
Dean Nelson: Right. And then you can ensure those SLAs. Because then the end of it, if you have to shed the… Because okay, let’s say we take that 10 megawatt example again. And I sell 10 megawatts of five nine and they use five and then I sell another five megawatts that’s going to be two nine. Let’s say that 10 megawatts starts to increase and they go to seven or eight megawatts I don’t need to shed the two nines unless I have a fault because I’d lose that other two megawatts of redundancy. But I also don’t need to shed all the two nines if I have that issue. I can shed portions of it to get down under that level. But what workload besides gaming and machine learning actually has spiky consumption? Almost none.
Raymond Hawkins: Yeah. Not many.
Dean Nelson: It’s predictable.
Raymond Hawkins: Yeah, you can model for it.
Dean Nelson: Data centers are utilities, best friends because they’re constant load, right?
Raymond Hawkins: That’s right, steady state load.
Dean Nelson: Not like air conditioners and whatever else is going on.
Dean Nelson: So that’s where I’m saying that our algorithms go in there and calculate those elements and we can show that, okay, you could do 200% of a subscription in this. So it just comes down to the company’s willingness to go in and say, “well how much do we want to do?” The other one is what demand? So what we’ve been working on is I went and spent hundreds of hours doing dozens and dozens of interviews with end users, my peers from the past and said, “if this was available, would you use it?”
Dean Nelson: Every one of them has an orchestration layer, software play right in cloud that allows them to be able to do this. You just point it at the infrastructure and data centers now. It’s not like rocket science that you have to go back and reinvent the wheel. They have these abstraction layers that allow it to happen. So we just don’t offer the product.
Raymond Hawkins: We think about it from a cloud bursting, cloud performance arena. We don’t think about it from the power utilization. And that’s what you’re saying is, “Hey guys, I’m not asking you to think about this a totally different way. I’m asking you to think about something you already do and put it on top of power. That’s really what you’re saying. Not a new unique way of thinking. Just point at a unique problem. A problem we haven’t looked at from a power utilization perspective. Well Dean, I did not expect us to spend 40 minutes talking about VPS.
Dean Nelson: It’s been 40 minutes? Goodness.
Raymond Hawkins: Yeah, it’s so good. This is just so helpful and I enjoy it and it’s awesome. And I’m amazed that it’s a sign of an interesting subject and a good conversation that it slips by us as quickly as it does.
Raymond Hawkins: I’m going to back up a little bit and take a little bit of a ESG sort of stewardship comment about all of this. At the end of the day, you use the statistic or quote the statistic, we’re 2.4% of the global digital or electronic infrastructure. We’re eating up 2.4% of the draw. And I think that our industry gets a little bit knocked on the head of, Hey, you guys are really big power users.” Individual municipalities, individual counties go, “man, I don’t know if I want those guys here.” Well, you made the point, wait a minute, we’re 2.4% of the draw, but we are enabling 70% of what goes on in the planet. So, in reality, not that bad. Now that being said, I’m not trying to discount that we need to be responsible stewards of the energy that we consume.
Raymond Hawkins: But also, I think it would help, if our industry could continue to talk about, “Hey, what do we actually enable with that 2.4%.” And a bad analogy, I think a little bit about America, America. Oh, America consumes 25% of the global goods. Yeah but we also produce X percentage of the global engine to provide for everything. So a little bit of that, “Hey, yes, we are eating up 2.4% of the draw, but we are enabling”, meaning our industry, “enabling an awful lot of things to go on on that 2.4%.” So I just think it’s good to keep in perspective, the digitization of the world is being fueled by that 2.4%, which in that context is not out of whack.
Dean Nelson: So if you think about digital infrastructure today, it runs the world. It’s the internet of everything. If we didn’t have digital infrastructure, the world, as we know it would stop. Think about it. If you lose your phone, what happens?
Raymond Hawkins: Yeah.
Dean Nelson: You don’t know what’s going on.
Raymond Hawkins: Yeah, You can’t watch the news. You can’t get in a car. You can’t book a room, you can’t travel. You can’t… Yeah. I mean, it’s, 100%. That’s the point is the digital infrastructure has become the backbone of virtually all levels of commerce and transportation.
Raymond Hawkins: And it’s funny, I said to somebody the other, but where we’re talking about redundancy, you’re going to laugh. I actually suggested somebody. I think I need to buy two phones because I don’t think we realize how valuable that asset is. And I think it’s silly that if I lose one, there should be a backup, sleeve in my backpack. That here’s my second phone Because it’s such an incredibly… I mean, when I travel, my boarding passes, my IDs, my credit cards, my actual car. I don’t get a car, Uber comes and gets me. I can’t get in an Uber without my phone. I physically can’t get to my meetings. I can’t move. I think I underestimate how vital that phone is to how I function every day.
Dean Nelson: Yeah. Two weeks ago I was in Virginia and my phone stopped working. So I didn’t lose it. But all of a sudden it wouldn’t boot. And I plugged in the wall for an hour and it kept giving me this error message. I’m like, oh my God, I don’t even.
Raymond Hawkins: What am I going to do?
Dean Nelson: How do I…. So suddenly that panic about what it was, I’m like, “do I need to go buy a phone and I’ll reinstall?” But then I thought my phone is my two factor authentication.
Raymond Hawkins: How are you going to log in?
Dean Nelson: How do I-
Raymond Hawkins: How you log in to replicate to your new device? Because it’s getting to… That’s right. That’s what I mean, that’s… I literally, I mean, as we talk about infrastructure, I’m not suggesting everyone go by two phones, but I do for guys that do what you and I do and travel as much. I think a second phone is almost… And I also think… So, I see you have a bunch of hard drives back over your right shoulder, and I back up my laptop to those because when I travel somewhere, I want my data resting somewhere that’s not at risk. I think the same thing with my phone. I think that, it’d be great if I had two copies everywhere.
Dean Nelson: And the thing that… Okay, we’re going off on a different topic here, but…
Raymond Hawkins: It’s a whole different subject now, but yeah.
Dean Nelson: I literally just ran out of storage on iCloud. They only offer two terabytes. But if you think, I literally have a 64 terabyte array sitting right there. I can’t get it anywhere else except here, which is a risk. What happens if there’s a fire in my house?
Raymond Hawkins: Well’s a fire in your building. God forbid, but yeah what if there’s a fire or what if there’s a lightning strike and it fries the device?
Dean Nelson: So, I think there’s… To tie it back to what you were saying, digital infrastructure enables the world and it’s only going to grow, but it’s only 2.4% of the actual energy draw. If you think of a comparison, Chris Crosby gave me this example, 13% of energy consumed is for water. Water purification, 13%. So of course we need water. We need shelter. We need power. We need food. So all those things, digital infrastructure’s one of those core elements that we have to have. It’s going to grow. It’s definitely going to grow, but we’re a lot smaller than I think people believe and what we enable. So there is a messaging thing here. There’s a communication element to say, people need to understand. I don’t know, if you saw the actual protest in Ireland about data centers?
Raymond Hawkins: No, I didn’t see them.
Dean Nelson: They’re saying you’re using all of our energy and they want to shut down data centers in Ireland. And so that is propagating to a lot of different places. It’s people that are not aware that everything they do every day goes through the data centers that they’re actually protesting in front of.
Raymond Hawkins: Yeah. Dean, it’s funny. People ask me what I do for a living and then the easy answer is I say, “Hey, I’m a real estate developer.” And then that usually stops the conversation. Thanks. And they get it when people really start to dig into… That’s all I say. When they really start to dig in, I go, “look, here’s the easiest thing I can say to you. Everything you do on your cell phone runs in our buildings. That’s the easiest way I can say it. If you do it on your phone, that means it’s happening in a building we build. That’s what we do.” And yeah, that’s exactly what you’re describing.
Raymond Hawkins: Well, Dean, thank you so much. I always enjoyed getting a chat with you. I’m so grateful that you’re willing to do the podcast. So my listeners can benefit from your experience and wisdom. And man love talking to you. Love seeing you out there in the marketplace and love what you guys are doing at VPS. I think you’re changing the way we need to think about the way power gets utilized and that’s something important as stewards of a significant chunk of our planet’s energy utilization. So thank you, Dean.
Dean Nelson: Awesome. Thanks for having me on.