AI Proving Ground Podcast
AI deployment and adoption is complex — this podcast makes it actionable. Join top experts, IT leaders and innovators as we explore AI’s toughest challenges, uncover real-world case studies, and reveal practical insights that drive AI ROI. From strategy to execution, we break down what works (and what doesn’t) in enterprise AI. New episodes every week.
AI Proving Ground Podcast
Assessing AI Workloads: How to Choose the Right Environment for Enterprise AI
Where should your AI workloads run? It's one of the most overlooked questions in AI strategy. From surprising constraints around power, cooling and floor space, to the growing demand for GPU-as-a-Service models, this episode delivers a field-level view of the challenges enterprises face when moving from AI proof of concept to production. You’ll hear why infrastructure readiness assessments are essential, how AI workloads differ from traditional IT, and what to consider before buying that next GPU cluster.
Support for this episode provided by: Equinix
Learn more about this weeks guests:
Chris Campbell is Sr. Director of AI Solutions at WWT, overseeing AIaaS/GPUaaS, Facilities & Infrastructure strategy and delivery. He's held leadership roles in executive engagement, engineering, and architecture at WWT, and led practices at Forsythe, Red Hat, BEA Systems, and AT&T. Chris holds a BA from Columbia University and an MBA from the University of Maryland, where he was a Dingman Entrepreneur Scholar.
Don Molaro is a tech professional focused on making data centers carbon neutral. Based in Cupertino, he holds a master's in Computer Science and 33 U.S. patents. He specializes in systems programming, high-performance storage, and software-defined systems, with experience across industries from finance to HPC. Previously, he held senior roles at Citibank, DataDirect Networks, and Hitachi Data Systems.
Chris and Don's top pick: What is GPU-as-a-Service (GPUaaS) or GPU Cloud?
The AI Proving Ground Podcast leverages the deep AI technical and business expertise from within World Wide Technology's one-of-a-kind AI Proving Ground, which provides unrivaled access to the world's leading AI technologies. This unique lab environment accelerates your ability to learn about, test, train and implement AI solutions.
Learn more about WWT's AI Proving Ground.
The AI Proving Ground is a composable lab environment that features the latest high-performance infrastructure and reference architectures from the world's leading AI companies, such as NVIDIA, Cisco, Dell, F5, AMD, Intel and others.
Developed within our Advanced Technology Center (ATC), this one-of-a-kind lab environment empowers IT teams to evaluate and test AI infrastructure, software and solutions for efficacy, scalability and flexibility — all under one roof. The AI Proving Ground provides visibility into data flows across the entire development pipeline, enabling more informed decision-making while safeguarding production environments.
AI's insatiable hunger for power is reshaping the global energy map, and fast. Just last month, the International Energy Agency warned that data center electricity use will double by the year 2030, with AI responsible for most of that surge. Reuters put an even sharper point on it by the end of the decade, one in eight US electrons could be flowing into server halls instead of homes or factories. One in eight US electrons could be flowing into server halls instead of homes or factories. Now, that's the landscape. But our two guests today, Chris Campbell, a senior director of AI solutions, and Don Malero, a chief technical advisor, both of which are colleagues of mine here at WWT, are here to talk about the reality that today's AI infrastructure decisions aren't just technical, they're operational and they're financial out their core.
Speaker 1:Now you might be asking why does all of this matter? Well, the competitive edge in 2025 isn't just having AI, it's running it where it performs best, at cost and under a risk profile you're comfortable with. Get that calculus wrong and the same model that promised to cut fraud or discover a new drug now becomes a drag. So stick around. In a few minutes, don will share a quietly powerful rule of thumb he uses before any deployment for whether a workload belongs on-prem, in colo or in the cloud. It's simple, it's actionable and it might save you more money than the GPU's cost in the first place. This is the AI Proving Ground podcast from Worldwide Technology. Let's get to it, chris. I get the sense from speaking to either clients or hearing from experts that we have here at Worldwide Technology, that consideration of AI workloads and how they're deployed can sometimes maybe even often be an afterthought. Is that an accurate assessment, and what are you seeing from that perspective?
Speaker 2:Yeah, there's kind of a few things we're seeing is there was a lot of folks that were doing what I'll call stroke of the pen type purchases of equipment, and so they would buy this equipment without really thinking about how they were going to support it from an infrastructure standpoint, power standpoint, data center standpoint. So now they're looking at it after the fact and saying, well, where should this stuff actually go? And in some cases they may not be able to support it from their data center. We've seen that in quite a few different people and really from our standpoint we say, well, let's go ahead and try to plan ahead. So we're really encouraging our clients now to think ahead. We do a facilities assessment and this will land in one of a few places. But with that planning, I think that'll help them to really be able to deploy this in a meaningful way without any slowdown or anything that's going to happen with them to be able to delay these projects.
Speaker 1:Yeah, don you know, for those that maybe did treat it a little bit as an afterthought or thinking that it would just kind of take care of itself, what are the stumbling blocks for them if they don't take this into consideration from the get-go?
Speaker 3:Well, I mean in the facilities business, there's the iron triangle, which is a triangle between power, space and cooling. The first stumbling block that we see organizations run into is power. They realize that, you know, with an IT rack that, on the high end, is drawing in the 17 to 20 kilowatt range, and they're talking about deploying an AI rack that starts at 50 kilowatts and can conceivably go up to well over, well over 100, into like the 130, 150 range. Right now, power becomes the first thing that they have to deal with Very quickly, and it's a different kind of power. We're going from 208 volt power in a traditional data center to most of the ai systems are requiring 415 volt power. So there's not only is it a lot more power, it's a different kind of power.
Speaker 3:Um, the second piece to that that comes along is cooling. Obviously, uh, the second law of thermodynamics you know it still applies. Uh, the power you put into the rack has to be extracted somehow. So cooling becomes the next issue that organizations and customers start to deal with. There's lots of different ways to do cooling, and we can talk about that at some length, but another piece of it that comes along very quickly on top of that is the space itself.
Speaker 3:Traditional raised floor data center raised floor environments simply can't support the weight of modern AI systems about things that are well over 3,000 pounds going into a data center that might be built for 1,200 pound racks. So we're dealing with things at the maximum. So we're dealing with challenges on power, space and cooling. The final challenge that gets put onto this is the cabling and the amount of interconnect that these systems require, and we've got lots of issues around structured cabling that goes into these systems. That is orders of magnitude larger than you would get in an IT workload. So it's power space, cooling cabling, and that's just to get to day one operation. So there's lots of challenges all over the map for this.
Speaker 1:Yeah, I like that iron triangle. Did you come up with that on your own? Is that an overpriced area?
Speaker 3:That's a traditional thing in the facilities business that certainly my colleagues Bruce and Mike have been putting up in front of customers for a long time. I'm relatively new to the hardcore facilities end of the business. I come at it from a compute and an IT operational point of view and it's been a real education working with the facilities infrastructure team and the depth of knowledge that they have. But yeah, I can't take credit for that. I'm going to give full credit to Bruce Gray and Mike Parham on that, for teaching me about facilities, Sure, sure.
Speaker 1:Well, chris, maybe take a step back for us here and articulate, just for those that may not be familiar, what exactly AI workload means for our enterprise customers. What exactly AI workload means for our enterprise customers? Is it different from traditional IT workloads that we've been talking about over the past, however many decades?
Speaker 2:And if it is different, why is it different? Yeah, you know this is not your father's AI workload or infrastructure workload that they have. It requires a fully different, as Don alluded to different setup in the data center, different amounts of power, but the workload itself really is you kind of consider what folks are trying to accomplish. So they start with a use case, they have some data that they use to point to, and then they use a training model within AI to be able to go and train it, and that training model requires a lot of compute power. That's why NVIDIA is so front and center in this is that they've got these chips and allows them to really process this in a really fast way. That's why, when you use a GPT-4 or some other kind of chatbot, it reacts so quickly and this is why they've got such a dominant place in the industry.
Speaker 2:But from a workload standpoint, that workload is how are they processing that information rapidly to get the answer that they need to?
Speaker 2:And so ultimately, when it trains, it requires a much stronger and more powerful infrastructure to run on. But once they can move to inferencing, which means that they're actually processing on a more long level, and it's actually what they're doing, from the time they train it through the rest of its lifespan, is inferencing. That requires a slightly less powerful infrastructure, and you could do it in a variety of ways, maybe even using different types of infrastructure for it that allows it to be able to run and not require that power. So it is very different. This is why we're seeing this rapid shift in the data center space to try to find spaces for these super pods that NVIDIA is putting out there and other places, because they really can't fit in any other type of traditional data center and in many cases, the legacy data centers that they have can't even be retrofitted with the requirements needed to supply it. So it's vastly different than what it has been in the past.
Speaker 1:Yeah, and we'll get into talking about where these workloads can and should run, but Don real quick as well. Are all AI workloads created equal? I'm assuming that there's a level of nuance to all of this, and what are some of those considerations that we should be thinking about, that our clients should be thinking about as they're assessing their AI workloads and then as we start to think about where they should run?
Speaker 3:In some way AI workloads are created equal and in many other ways they're not. They're all pushing the limit. They're equal in the sense that they are pushing the limits of what is capable in the industry at the moment, whether that's leaning on what the compute engine is and quite often NVIDIA GPUs, the networking infrastructure that is going with that. There are a lot of very similar considerations between organizations the latency, the love, you know, the latency, the bandwidth that happens on the network side and on the storage side, you know, in making sure that the right data is in the right place at the right time. Those are all equally important considerations that organizations have to look at and that and there are common or reference solutions to to all of those problems and that's sort of what we do in the AIPG is is we have those reference designs, we have them built out, we have good examples of what, of, of what proper reference architecture looks like for those solutions.
Speaker 1:Yeah, aipg being the AI proving ground that we have here at WWT to help our clients go through their AI journey. Chris Don mentioned at least several of them computing storage. What about data sensitivity, security, latency, things like that? What else do we need to be considering?
Speaker 2:Yeah, I would say that. You know, depending upon what they're trying to accomplish and the sensitivity of that data, certainly you have to consider that and you know that's truly something that, as we're looking and working with some of the early adopters, they do not want this data to be any place other than an on-prem environment or a private environment, and so I think that as we continue to progress and more use cases are released and they have less sensitivity about that data, you're going to see other areas, maybe the cloud or other places that'll go, but that data sensitivity is certainly an issue and then in a location and I think what we're seeing now is you could train this in one area and then ultimately, as you go out to do the inferencing, you may want to move it with more proximity to a cloud source or other sources so it can run more effectively and efficiency and not have that latency.
Speaker 1:And where does considerations of the use case come in, if at all?
Speaker 2:Well, the use case will depend on how quickly they want that information in many cases. So you know what we're seeing in, for example, finance or we're seeing in health care. Those are really the early adopters that we're seeing. Life sciences are doing that as well as some of the retail, and so that proximity and speed to that information, depending upon how quickly they want to get it let's just say it's fraud protection that they might be using within a financial institution They'll want that response time to be very, very fast. And Don's certainly an expert in that area and can jump on and add to whatever I might be missing, but I think that's something they consider as well.
Speaker 1:Yeah, don anything to add there.
Speaker 3:Sure, chris is exactly right. The underlying infrastructure is important, but the use case and the business value that goes on top of it is what the customers are ultimately investing in. They're looking at that business outcome and that's where all of the other layers of an AI solution whether that's learning or whether that's inferencing or even traditional AI ML that's what matters to the customer. The infrastructure piece of it is simply a means to an end. So we have to be we're very careful about how we build out those environments whether that's a retrofit on premise, because sometimes it might be for very modest footprints, but we've seen that some one of the financial institutions I'm working with they're doing. They're doing their inferencing-prem because they can't eke out enough power and cooling in their existing data center to do it.
Speaker 3:For other organizations, they're building and retrofitting dedicated environment relatively small compared to their overall footprint, but still dedicated and reasonably significant environments in their own data centers. Or we're working with customers who are placing workloads in co-locators, and some of our key co-locator partners there are adapting their environments for very high scale, very high load AI type deployments. And finally, we're looking in, as well as the cloud, the, the the old cloud service providers, we're looking at new ones that are specifically dedicated to providing GPU and high performance computing in a secure environment for organizations that have that have those requirements. Chris is exactly right. These workloads are. They are core to the businesses that we're dealing with. They are quite often some of the most important workflows and applications that they're running. You can, fraud detection is a critical workload at a bank, and so deploying that and doing it correctly and doing it responsibly is absolutely critical for these organizations, and they're having a really, really hard time doing it on premise with modern computing equipment.
Speaker 1:Really really hard time doing it on premise with modern computing equipment. Well, chris, we mentioned those options there Colo, cloud, on-prem. What are you seeing in the market right now in terms of the most appropriate way to divvy all that up?
Speaker 2:So the answer is it depends. I think if you, right now, there's three options that people are really looking at, and if you look at what some of the analysts say, and they're talking about a third of them being on-prem, a third of them being in the cloud and a third of them being in some kind of private hosted environment like an AI as a service or a GPU as a service, which is a leased GPU environment, and ideally, I think that middle area, that last area I talked about, which is the GPU as a service, private hosting is going to be far more prominent. The challenges that they face are if you want to go on-prem, your data center may not be ready. And so, to Don's point, some of these folks might be eking out inferencing that they have, and they might be doing that to do the best they can. In other cases, they're saying, well, maybe we can try to retrofit it, but pushing it to the cloud or pushing it to a hosting provider like an Equinix or a digital realty, I think, makes a little bit more sense in some of these cases. Now, you're not going to go put in the cloud if you've got some kind of sensitivity to the data.
Speaker 2:So that idea that this private hosting environment is going to be something that's prominent for some of these workloads, I think is very true.
Speaker 2:But as these early adopters come on, some of the easy button is well, we're going to just go and do this in the cloud We've already got an AWS account or they're going to go ahead and buy something and put it in a private hosted environment because they're going to make a longer term commitment.
Speaker 2:But the GPU as a service providers that I mentioned is something where, if you want a short term testing and we're seeing this with our partners like CoreWeave and Scott Data Center, applied Digital they all are getting a lot of business right now and people just using short term testing so they can prove out some type of use case that they're working on or be able to go get fundings, much like they're doing in our AIPG, but with their own data. So it's a really interesting time right now to see this. But as this business continues to build over time, I think you're going to see these decisions being made and while folks will invest in HPA, on-prem, high-performance architecture on prem, I do think that some of the easy buttons for them, instead of building a data center, buying $20 million worth of gear is to figure out a place to go and host it or to lease it from someplace. It's kind of that build versus buy discussion that people have.
Speaker 1:Yeah, Don. What do you make of that GPU as a service angle, and doesn't that also just run the risk of potentially running up the cost? If people are looking to you know, take advantage of that scarcity.
Speaker 3:So. So there's there's as with all these solutions, there's a I mean, there's, there's a lot of nuance in the actual solution. So, even even when people talk about GPU as a service, you'll have environments. You'll be able to go to a GPU as a service provider, and there's lots of them that we're partnering with Scott Data Center, coreweave, lambda Labs, Applied Digital there are lots of them in this space. They all have slightly different product offerings and dealing with them and getting your workload working there is non-trivial. For example, they may have a debit.
Speaker 3:You may be in a situation where you require a cage, dedicated environment that even the GPU as a service provider doesn't have access to, because you want to run confidential workloads in there. That is some of the offerings from this. There's much more traditional, even all the way down to virtualized environments where you're multi-tenanting on hardware with other corporations or other clients of those organizations. So even the offerings that you'll see from a GPU as a service provider, there's a lot of nuance there. There's also a lot of nuance about the level of service you're going to get from them, which is very different than what you would get in a cloud service provider. Cloud service provider is going to provide you a complete end-to-end solution where they're going to manage all of the infrastructure, all of the services and everything that goes with that In a GPU.
Speaker 3:As a service provider, they might just be providing you the bare metal, or they might be providing you up to the operating system level, or they might be providing you up to the operating system level, or they might be providing services up to the Kubernetes and Slurm level, or even beyond that. And this is why I think there's a lot of value in a customer coming to WWT, partnering with both WWT and the GPU as a service provider to provide an end-to-end solution that includes access to the hardware, access to a high-performance computing environment that has the right services all the way up through the application. And this is where it's a real case of one plus one equals a lot more than two for the customer. Because that's where the value of the partnership is, between WWT and our GPU as a service provider partners to the customers. I think it's not only is it just a matter of having a place to run your workloads, it's actually, you know, we really believe that you can execute a better solution for the customer and those kinds of environments.
Speaker 2:Yeah, brian, I'm going to plus one what Don was talking about here, but really talk about.
Speaker 2:You know the investment people have to make.
Speaker 2:So if you're asking, if you're a CIO or you're running a center of excellence at a client and you're trying to go ask for $5, $10, $15 million to go to a use case, they are sometimes wanting to prove that out before they get approval from their board on that cost.
Speaker 2:So, using a GPU as a service provider that already has that equipment in place we see this again in our AIPG as well, in our AI proving ground, where folks will come in and just want to run it on one of the 15 reference architectures that we have. Well, that's kind of a first step. And then, when you really want to use your own data rather than some kind of synthetic data just we use in the AIPG, that's where they could go into a GPU as a service provider, really prove this out and determine whether or not it works, and I think that's something that saves them a lot of time. It probably saves them some money in the initial stages and so you know, we see a really high level of use coming for this GPU as a service as folks continue to put their budgets toward AI projects.
Speaker 3:This episode is supported by Equinix. Equinix connects businesses globally through its extensive network of data centers, enabling seamless digital interactions. Expand your reach with Equinix's interconnected infrastructure.
Speaker 1:What's the market look like? I heard a story the other day talking about how there's people out there looking to build a network of folks that just have excess capacity of their GPUs, thinking like kids in their parents' basement playing on a GPU card in a video game. Obviously that's kind of an off angle of that area, but is that market going to continue to expand and grow or is it going to be tight?
Speaker 2:area. But is that market going to continue to expand and grow or is it going to be tight? Yeah, the biggest thing that we're seeing for GPU as a service providers is not necessarily the kid in their basement, but it's people that have access to power. So former crypto companies or crypto companies that have access to power just somebody that has raw access to power they are getting funding. There's a lot of private equity money that's out there looking to go and invest in this space. So build a data center, buy a bunch of NVIDIA gear, set it up to go run GPU as a service. We actually at Worldwide are working with them to be able to develop those services. So we have a whole group of folks that work with those different types of companies as they want to get these up and running, to help build them. But I think that's going to be a real interesting thing over the next three years because this access to power is so critical and if you look at what the power companies are saying, they can't generate enough, cannot meet the demands of what some of the hyperscalers are asking them for, even in an RFP instance. So you have these smaller places that have access to this power and they're going to go develop it. So I think we see anywhere from 50 to 100 of these that are sort of out there in early stages in some way. I think they'll probably like most things that are kind of a little bit like the Wild West. You'll see a handful of them actually survive and really get up and running beyond what they do. But we're very pleased with the ones we've been working with to help them.
Speaker 2:I think that as we see them develop and we see the demand for GPU as a service, it is a little bit like a build it and they will come. People will end up coming and using it. I think the dependencies are going to be on how sensitive that data is, what type of market they're in. I would see that much like a car, where folks will buy based on what their requirements are. You might have folks that go for a more fully managed, more luxury experience, where they have somebody with a lot of enterprise experience to get that going. Or you might see more mid-market companies that can't afford that, using some of these other providers, or even verticalization of those providers, as maybe somebody focuses on utilities or somebody who's focusing on retail and you'll see them break into those types of segments, but we're in the very early stages of this. There's a handful of providers that we partner with that Don and I have a portfolio that we have and that we believe are good partners for our clients, but there's a lot more to come.
Speaker 3:Yeah so you go ahead, don.
Speaker 3:So I'd like to comment on that a little bit is these are the kinds of workloads that we're talking about are very, very different than a traditional it workload.
Speaker 3:Um, uh, supercomputing, which is really what we're talking about here, and it computing are very different specialties. Uh, from an operational point of view and this is where I think there's a lot of the organizations that are either looking at doing it themselves, or some of our customers doing it themselves, or many of the small new entrants who have access to power or trying to stand up environments, don't quite realize just how hard it is to run high-performance computing systems on day two and day 100 and day 200 type operations. So these things are big, they're complex, they consume a lot of power, they have requirements that have never been delivered or extremely rarely delivered inside corporate data centers. Partners is evaluate them and see whether you, whether they're even in a position to run these kinds of architectures and run these systems, um, in a responsible way to the business. Um, I think that's an important thing to talk about, especially dealing with some of the large, highly experienced GPU as a service providers. It's just how hard it is to run these environments.
Speaker 1:Expand on that a little bit. What do you mean by in a responsible way to the business?
Speaker 3:Failure rates. When you, when you pump a 130 kilowatts into a very small space you are going to, you are going to generate more failures in the electronics. It's just. I mean, that's just, that's just physics. And so you have to do things like how do you manage spares, these things? How do you manage extremely expensive spares for things like GPUs? How do you, how do you build resiliency into the system that is right for the business?
Speaker 3:High performance, traditional high performance computing environments, kinds of things you see in national laboratories and three-letter agencies and places like that. They don't have the same level of redundancy built into them. They get their resiliency through different methods. They go through a method of checkpointing, which is not what you do in a traditional IT space. You have to understand how those workflows go through the system and understand sort of them from an operational level, how to deal with failure, how to deal with, you know, making sure that the that your very, very expensive resources, um, are properly utilized, um and how to account for that Um again, different kinds of things than you would get in a traditional it space. Uh, having those kinds of skills sort of built into the organization is not something you can just turn on overnight. So these are some of the more nuanced and day two and day 100 kind of conversations that we're having with customers around sort of how, how, how to make sure you're getting the value to the business that that you're paying for.
Speaker 1:Yeah and Don, one of the funny things that you know I enjoy hearing I've heard you say on multiple occasions is you know how you might find yourself in a meeting and you're just delivering only bad news. So if you're talking about that day to day 100, talk to me a little bit more about what you mean by that bearer of bad news. And is it just that the reality of the matter is that not many are ready for this?
Speaker 3:Well, what, what are, what are I mean I? I mean I say that in jest, you know around that, but really what our customers want is that trustworthy, informed opinions about how to get to the business outcome they're trying to get to. And quite often part of our job is to look at that customer and look at their environment and say your environment is not going to work or your operational model is not going to work, and we're going to explain why. As opposed to, I mean, giving bad news is easy, giving an informed opinion is what our customers are asking us for Now. Again, we have that experience right.
Speaker 3:I have plenty of time doing supercomputing in previous roles. I understand sort of what the requirements are at the highest levels. I have plenty of experience in dealing with traditional IT environments and the kinds of things that you have to do in order to make those operate from an IT point of view. Plenty of experience and understanding sort of the capabilities of data centers. Again, now we're looking at the workloads that we're getting from our OEMs like NVIDIA and others that are going into those environments. And again, because WWT has that experience in high-performance computing in IT, you know we can look at that and we can honestly go to a customer and say you're not ready for this, and here's why Customers love that kind of advice. Telling them everything is going to be OK is quite often not what they need to hear. Okay is quite often not what they need to hear. And so long as we can tell them, you know that. You know there's a. We've recognized what your problem is and here's a path forward.
Speaker 1:That, I think, is the real value that WWT brings to this. It brings to our customers in this equation. Yeah, chris, from what you're seeing, do you find organizations try to pigeonhole themselves in any one of those particular areas, either on-prem or in the cloud, just based on how they've been doing things from a traditional standpoint, or are they open to those conversations?
Speaker 2:Yeah, I think they have to be open to them, just because of the cost involved and you're not talking about small investments that they're making and so they might have some type of use and been burned by cloud as an example. I think everybody moving over to cloud had unanticipated costs that they ended up building up, and so they're trying to determine, well, with these two workloads that are different, is that cloud cost going to end up being something that they don't want to repeat again either? But then you know, ultimately, as we're trying to look at and talk to what they do with these workloads, we want to help them evaluate what's best for them as well. So if they have the ability in their data center to go do it on-prem and they have money and the workload requires it, I think that's something that they would go down and do, that. We have some existing folks that have that capability today and that's where they're comfortable doing it and they prefer to keep it on-prem.
Speaker 2:I think what we're seeing also is and I talked about this a minute ago is this is still very early stage, so where's the long-term investment going to sit? And I think the answer is it's going to be a hybrid approach. So they might keep some of it on-prem, they might run some of it in the cloud, they might run some of it in a GPU as a service or a private hosting or AI as a service type model, and I think that's, you know, from a cost standpoint they're going to evaluate. We are working on those types of cost models ourselves so we can help them better understand how to make those decisions and where they fit, so they can be a little more predictable in the cost.
Speaker 2:I think, if you talk about talk to Gartner or anybody else, I think that some of their reporting says that in the next four or five years that 50% of these projects are going to end up being more expensive than they anticipated and go over budget. Then they also talk about operationally how about in the next three or four years? You know 30 to 40% of these are going to fail because they aren't architected the right way or the model they chose wasn't good enough. So I think there's a level of predictability that they're looking at that we're trying to help them with. But you know we're looking at where these want to fit.
Speaker 1:Cost becomes kind of the king when it comes to this as well, as does it actually support the workload in the way they want it to work? Yeah, don, what about? I mean it's interesting that you talk about cost, chris. I mean certainly everything kind of boils down to that. But, don, what does events or inflection points like deep seek? How does that change the discourse of this conversation, thinking that there's probably going to be more deep seek, deep seek type examples moving forward. What does that do to that kind of spreading of AI workloads?
Speaker 3:So deep seek. I mean the industry over rotated a little bit on that, but there is a message there that matters, which is this is an incredibly complicated and complex computing problem and in computer science the way you solve incredibly complicated, complex problems to start with is brute force, where you use brute force algorithms and you and you do, and you start with that. What we're seeing with algorithms and improvements like deep seek and there there will be others is we will see good computer science come in and we will see improvements in those algorithms and quite often you can get very large speed ups and very large efficiency gains by just simply using better algorithms. And this is for me as an in in my background as a, as a computer scientist. That's where I think there's going to be a lot of interest in this space.
Speaker 3:Now, invariably in those things, there's going to be trade-offs in those algorithms, there's going to be improvements, there's going to be blind alleys that the industry looks at that. Some really, really smart people are starting to get involved and looking at it from an algorithm design point of view, which I think is again very interesting from a computer science point of view. So I'm actually really encouraged by it. I think it's actually a good. I think it's a sign that the industry is maturing a little bit, that we're starting to see these kinds of incremental advancements in algorithm design.
Speaker 2:Yeah, and I'm going to also add onto that as well. You know the idea. When it first came out, I think all of our phones kind of blew up that you know China's got something and it's going to be really bad for everybody, and very quickly. If you started looking at it, you know open source is very positive and NVIDIA was very positive about it, and as people really looked at it, they said, well, hey, this is just a way, a far more efficient way, to run some of these workloads. And I think everybody's looking for some type of increased efficiency.
Speaker 2:And you know there's a term that you know I've seen batted around a lot, which is Javon's paradox. And if you look at Javon's paradox, they kind of talk about increased efficiency can lead to increased consumption rather than decreased consumption, and so this started with coal. When they did it, it was semiconductors, it was cloud, I mean. You see all these examples of this. So I think there's nothing to fear here and I think, in fact, it'll end up being more efficient for us to run in the long run, as we're all learning how to do this, and I think you can only look for people that continue to try to innovate and make this better and I think it'll just ultimately lead to more increased consumption, as we talked about. So I think it's an interesting time. I think we're only at the beginning of how people will innovate, trying to make this either more cost effective, more efficient. You know you can't just continue to have runaway costs in this space. I think it's got to be brought out of reality fairly quickly in this as well.
Speaker 1:Well, Chris, another thing that you mentioned that I wanted to touch on, but I will. I'll flip the question over to Don. You mentioned how you know lessons learned from the cloud period. You could talk about cloud, you could talk about virtualization, you could talk about edge. What types of lessons learned from those eras, so to speak, should we be applying to ai workloads, recognizing the fact down that early on you said these are just different beasts. Ai workloads are there lessons learned there?
Speaker 3:I think there are. I I think one of the I think you didn't mention to me, which is the, which is the um. What's the analogy that I give people? 10 years ago, everybody was putting everything onto Hadoop. I mean, everything was going onto Hadoop, every single workload was going to be an analytics workload. And, again to me, very similar kinds of decisions were being made. We were organizations were setting up massive, huge, very expensive, um, um analytics environments, uh, and with the intention that they'd be moving tons and tons of the, if not the vast majority of their workloads whatever that happened to be into analytics environments. There's an immense amount of value in doing analytics, the way that that's done in those environments.
Speaker 3:But that mania died out. Sort of cooler heads ended up prevailing and there's incredible amounts of value coming out of the. You know, coming out of that work. We're in my mind, we're we're sort of in a similar sort of stage right now is everything. You know, everything needs to be tagged and has a tagline around AI and how it's innate. Everything is an AI workload and there will be a lot of workloads that do benefit from inferencing and from learning models and RAG models and all the things that we're talking about in AI. But at some point Chris is right there's going to have to be real value generated from this and the people who are paying for it are going to have to be able to see that return on investment. And we need, as an industry, we need to get to that place as quickly as possible, because that's when this particular industry will really start to mature.
Speaker 1:That's when this particular industry will really start to mature. No-transcript.
Speaker 2:You know, I think that some of the things that we everybody's in reaction mode and so there's this idea that this is all new.
Speaker 2:We have to figure out how to react to it, and I was just actually talking with a vendor group and presenting to them as part of a panel this week, and one of the things I talked about was trying to think even more forward.
Speaker 2:You have to look at what the next iteration is going to be and then also trying to make sense of the chaos that this presents to some of the clients, because clients are still trying to figure it out. We have done a lot of work within Worldwide to educate our clients and help them to better understand it and, I think, for the partners that are out there really helping them understand as well. The clients are getting hit with a lot of noise right now and a lot of FUD, and so what can we do and what can these partners do to be able to help them? To to really, you know, help the customers understand the landscape, make decisions in the right way and, you know, really help them in some ways to to maybe not not overreact to what's happening right now so they can make smart decisions about the future.
Speaker 1:Yeah, Don any questions that you're not seeing. Asked a lot, but do deserve answers at some point.
Speaker 3:So I mean I'd pile on to what Chris just said, which is the questions I would like to see our customers coming to us is around strategic planning around this is around strategic planning around this. Chris is exactly right. We're seeing an incredible amount of reactive stuff. I have a project. I need to go get it solved. I need to get it solved in the next few months. Those are the vast majority of the requests that are in the pipeline right now.
Speaker 3:What I would like to see our customers and the industry asking a little bit more about is what does this look like as a strategic plan? How are we going to use this technology to transform the business, given that we have to use this technology at a certain level? How are we going to change our organization, whether that's we retrofit our existing environments, we look for the right GPU as a service partners, we look for the right co-locator partners, we look at the right kind of cloud services. I want to see our customers sort of level up the level of strategic thinking that they're coming that then the industry overall is dealing with. Again, we're in sort of this mania stage right now and I would personally be much more comfortable if things were if a few more cooler heads prevailed at the moment, but that's the world we're in right now.
Speaker 1:Yeah, a good final note. Well, to the two of you, thanks for taking time out of your busy schedules. I know you're always diving into the ATC going on meetings that may take you out of your hometown, so thanks again for joining.
Speaker 2:Yeah, thanks for having us. Brian, Appreciate it.
Speaker 1:Okay, as we wrap today's conversation, a few lessons that stand out clearly, lessons that anyone serious about scaling AI needs to take to heart. Anyone serious about scaling AI needs to take to heart. First, ai is not just another IT project, because you have to design for it intentionally or risk discovering that your data center simply can't support your ambition. Second, location matters, but so does sensitivity and scale. The choice between on-prem colo, gpu as a service or public cloud isn't just about convenience. It's about understanding your workloads, the sensitivity of your data and the intensity of your compute needs, and how quickly you expect to grow. And third, infrastructure decisions are business decisions.
Speaker 1:Investing in ai infrastructure isn't about building shiny new tech for its own sake. It's about delivering trustworthy, reliable business outcomes. Bottom Bottom line. If you're serious about AI, you can't treat infrastructure as an afterthought. The success, or lack thereof, of your AI strategy depends on asking the right questions before deployment. If you liked this episode of the AI Proving Ground podcast, please consider sharing with friends and colleagues, and don't forget to subscribe on your favorite podcast platform or on WWTcom. This episode of the AI Proving Ground podcast was co-produced by Naz Baker, cara Kuhn, mallory Schaffran and Stephanie Hammond. Our audio and video engineer is John Knobloch, my name is Brian Felt and we'll see you next time.
