Why Sovereign AI Is So Hard to Sell (And How to Fix It)

0:00 Stephan Fabel: We talk to researchers all the time. I talk to somebody who is responsible for hypersonic flight missile research, and that's of obvious importance to the national security. And they don't have access to the GPU resources that they need. They have to wait for for six months or nine months, actually, as what he said. He wait he waits for nine months to get six minutes on a on a national lab, and makes you know incremental progress. And so the the notion that that that team would get access to large scale infrastructure on an you know pro long time in a dedicated manner that really makes a difference and really pushes that research forward is it's more than mind blowing. It is, I mean, there were tears in eyes,

0:46 right? I mean, they they are really motivated by that mission, and so we're tying to that. Hi, this is Stefan Fabel, CEO of This Way Global, and you are watching the Data Center Go-to-Market Podcast It's Joshua Joshua Feinberg: Feinberg, host of the Data Center Go-to-Market Podcast and today I'm welcoming a very special guest, Stefan Fabel, who is the CEO of This Way Global. Thank you for Stephan Fabel: having me. Joshua Feinberg: Stefan is focused on high performance computing and AI, specifically for neoclouds and cloud service providers at universities and other research institutes. Stephan brings more than 30 years in technology, robotics, computer engineering, and cloud AI infrastructure. Most recently, period of about six years, leading the NVIDIA DGX

1:41 infrastructure initiative, and he brings extensive deep tech and open source cloud computing experience. Stephan, welcome to the podcast. Stephan Fabel: Thank you so much, Joshua, for having me. It's a pleasure. Joshua Feinberg: It's a pleasure to have you joining me as well. I think the first place I'd like to start is being in a place where you've worked for some very large companies in the past, and some smaller companies throughout the data center ecosystem, and now are leading a startup. I'm super curious to get your thoughts on what you see as the biggest go-to-market struggles that companies are dealing with currently within high-performance computing and AI data centers. What do you see as the single biggest hurdle, the single biggest friction

2:22 point that GTM teams, leadership teams are working through right now. Stephan Fabel: Well, I mean, in a way, Joshua, it hasn't really changed over over the time. I mean, at the at the beginning of our cloud transition, when we moved from virtualization over to private cloud or to public cloud, and we moved from you know more you know vertically integrated application stacks into cloud native stacks, the question has always been the same, and it's the same thing here as as well. Is the the how do I explain to the customers the benefits and the advantages and the value of introducing change to the way that they're doing things, so you know whether that is talking about the acceleration that that switching to a cloud native

3:10 application infrastructure brings with it, or whether it is figuring out how a a true sovereign AI strategy looks like that we're looking to to implement. Customers are asking always the same thing, you know, what is sovereign AI? Why do I need to add AI in a sovereign manner? You know, what is going to be the benefit once I've done that, and why shouldn't I just use some of the offerings that are out there? So our ability to explain new technology that's at the forefront and at the edge of of what's out there is really the the core friction point of of go to market teams today. Joshua Feinberg: How do you think the challenges of explaining that compared to other seismic changes transformations that have happened in IT over the last few

4:00 decades, when you were explaining that, it brought me back to thinking about when the world was changing from premise based and client server to cloud in the early 2010 s and trying to get people a little more comfortable with generative AI. How do you see this being similar, being different than convincing someone to abandon client server and move into cloud, move into SaaS? Stephan Fabel: I think that the biggest thing is is that technologists tend to abstract it into a technology argument. So there is a technological argument for why you'd want to use GPUs for certain problem, for example. There is a technological argument for why you'd virtualize something or move to a private cloud infrastructure, and so on and so

4:42 forth, but I think what technologists frequently underestimate is the cultural impact that has on an organization, and all the other sort of soft quote unquote things that have to happen in addition to this technological change. So, for example. When we talked about moving into a cloud native infrastructure, it's one thing to be able to stand up a a cloud like an OpenStack or Kubernetes or VMware native cloud, make that available internally and expose let's say a software development pace layer or something like that. That's yet another thing to get all of the multitude of developers successful successfully using that, and so what we found in the past is is that a lot of the adoption of private cloud, for example, was you know prone to failure. I

5:34 mean, there were not a lot of private cloud initiatives that had a lot of success, at least in the enterprise space, and similarly, recent reports from the MIT sale suggests that we're at a similar point with you know AI POCs, where not a lot of AI POCs are actually making it into production or are being fully adopted, and you know the reasons are largely the same. We don't really know exactly how to, you know, let's say hold it correctly. We don't really know how to implement it concretely into our current workflow. You know, we may not really have an idea how that impacts future or actually plan changes on on future outcomes. How do we actually bring this to bear, and how do we make this useful and turn it into something that's

6:25 productive and a net positive and seen as something that is a success rather than a risk to your career when you introduce it? Joshua Feinberg: I imagine different stakeholders have different ways of looking at this. I'm always surprised when I see in recent months when CFOs are freaking out over the size of their token bills, when the exact same crew was freaking out over public cloud costs and running up this whole repatriation argument over the last five or six years. Stephan Fabel: That's right. I mean, in that sense, it's it's ironic because it should be fairly obvious that renting a car is you know for the long term, is going to be more expensive than buying one, and so it's really it comes down to

7:06 economics, and I think we're we're sort of leaving that off off the side, and that's yet another aspect that continues to be ignored for a lot of these initiatives, where you know people are overly rotating on the technology, overly excited about the benefits that it may bring, but don't really take into account what it would take to move the organization along with them, or the impact to the bottom line and what that means in terms of opex versus capex spending. Joshua Feinberg: Because of your previous role with NVIDIA leading DTX Cloud, you likely have a different perspective on the shift to high density AI infrastructure than a lot of the other people in the data center industry who just have been dealing with in the past two three years. You it's been

7:52 dealing with it in one way, shape, or form for well over five years. And what I hear from most founders throughout data center operators and neo clouds and tech and facilities companies is the shift to this really high density infrastructure has changed how they approach go to market. How when you have these conversations now with people that are still warming up to this idea of 600 kilowatt racks, like how do you have these customer conversations? How do you think about your GTM strategy now? Stephan Fabel: Well, I mean, there is a certain amount of inspiration that needs to go alongside with that, right? I mean, you do want to be seen right now as a leader in this space. There's only two types of motivation here. The either is

8:39 you're either inspired by what's going on, you want to participate, and you're excited about it, or you're you're driven by by fear, essentially trying to avoid a competitor overtaking you and you know reclaiming their their leadership in a space that that you might might not be ahead in. As a as a data center provider moving into high density cloud infrastructure is going to be incredibly challenging. Is going to require a connective tissue between somebody who's you know typically not interacting with you know some of the parties that are typically not interacting with each other. So, I mean, look at some of the more successful rollouts of of high high density chips they typically required a tight integration between you know parties that typically are are

9:32 much more at arm's length to each other. You'd have a you know an energy provider, for example, having to work with a developer, having to work with a construction crew, having to work with a cloud infrastructure provider, a data center operator, then a you know potential application that you rent this out with. That is a whole host of of partners that have to come together in order to make something successful. Prior lives would would not have had to. Negotiate across so many boundaries, and certainly not the amount of capex and at each of those transactions. So, it is a new world. Joshua Feinberg: What's interesting about those integration challenges is about a year, year and a half ago, I started hearing from some of the largest technical IT

10:18 infrastructure companies that they were getting a lot of requests to integrate and acquire pieces that used to be delivered by a lot of channel partners and integrators in the past. One because the capital requirements were so much larger, the delivery timeframes were so much more complex, and just customers were demanding that. Are you seeing that in the field as you talk with clients, where a lot of separate categories are now just simply getting bigger by necessity. Stephan Fabel: Oh, absolutely! It is definitely a trend that things are coming kind of there's a consolidation, you know, across the the stack vertically, if you will. It happens sometimes. I mean, typically, actually, it happens between folks that are typically considered developers and energy

11:02 providers, so those two tend to kind of come together. But it also happens between cloud providers and energy providers. We see this in in neo clouds like Crusoe, called Cruso Energy, right? I mean, they are they are actually delivering much more than a typical cloud provider would would deliver, and we'll see different a different push in other spaces as well, where you know you might even see them collaborating very deeply with the manufacturers of some of the equipment necessary to deliver the power or the cooling in order to start standardizing these components and being able to accelerate the the rollout of them. So it's it's just a consolidation of all of that in order to make it more efficient, more streamlined, and more predictable in its delivery.

11:48 Joshua Feinberg: With all of this change in the industry, we know that buyer education is still in a really, really important part of laying the foundation and driving alignment and getting everyone on the same page. But technical buyers, IT professionals, engineers, even a lot of people on the construction, ESG, compliance side have notoriously been not big fans of marketing and sales. They they don't really like being sold to. In many ways, they have marketing allergies, sales allergies, generative AI, and the immediate gratification with their favorite tools has allowed them to circumvent and bypass marketing sales like they've never been able to in the past. How do you effectively educate and build trust with

12:31 modern technical buyers long before they ever agree to an initial meeting with someone from your sales team? Stephan Fabel: I think that you know it's a difficult question to to definitively answer because I think it really depends on on who you're talking to. In our case, we are delivering sovereign data centers and sovereign infrastructure, and so what that means is is that we're talking to states and state entities and state agencies and public universities and folks that are motivated by you know by a mission and by a responsibility to the public, and actually have funding to do so, and so we're tying into that mission by by expressing our desire to serve them and to serve that mission, and so that usually helps in the alignment. It does require a lot

13:19 of marketing if you you know if you can express your your motivation for why you want to do this, and I think we you know we frequently show to our buyers and our partners or customers what we're all about, and so what what we are trying to bring to the table here is a differentiated view that doesn't just dictate that GPU infrastructure is meant to serve the few; it is actually meant to be made available and broadly accessible in a sovereign manner that that allows you know initiatives of national interest and of of the public interest, frankly, to be driven forward in an accelerated manner that otherwise would not be achievable, and so, but I mean, if I wanted to generalize this, that's really the core of it: is

14:11 getting access to infrastructure and capabilities that are otherwise unattainable is the core promise of a high Density chip delivering data center, right? I mean, it is really about. I don't have to spend all this additional money in order to stand up my own data center, in order to bring my own capability to bear, which I may not have or may have a skills skill gap in house. This is a place where I can actually attach to both the AI and the accelerated infrastructure industry in general, and and deliver that value and accelerate my my business accordingly. So, I do think that alignment with the mission is of of. Great advantage here, and we certainly not we're not having trouble explaining to our sovereign partners why we're doing what we're doing and why

15:08 that's interesting. But I do understand the the necessity of having to relate to to the motivation of your buyer. Joshua Feinberg: So just to clarify, when you talk about sovereign buyers in this way global. This way global actually is global. When we talk about selling to states, you're talking about other countries outside of the U.S. as well. Stephan Fabel: I am. I'm also talking, but mostly I'm talking about U.S. states actually. So we're we're operating very tightly with with the U.S. states, many of them, and you know, are working with them on sovereign initiatives that provide GPU capacity to them through usually a a you know r1 research university in their state and their associated industry partners, and so these are typically partners that are

15:58 you know associated with the mission of that of the research that's being done and that institution, and then offer the capability to to create a space for all the agencies of the state to also take advantage of access to GPUs. And so that's a broader mission than just a purely commercial one, which is why it's not really always a go-to-market question, right? Sometimes it's the motivation is simply funding, not revenue, for these research institutions, and still they they need access to this large-scale compute because otherwise we're not really making progress on some of the missions that they're tasked with, and so that's where we're at. Joshua Feinberg: Yeah, it's really interesting that you explained that, clarified that, because as you were talking

16:47 through that, I was thinking about some of the guests that I've had in this podcast from throughout Europe. This podcast launched a little over two years ago, and in that time, there's been a very different geopolitical environment with how people view each other's in trades and tariffs and geopolitical differences, and it was very common a year, 1218, months ago to see data center operators in different parts of Europe say, if you have assets that are sitting in colocation sites in the U.S. and and so sovereign for sure on a national level, but what you're talking about is very different. It's talking about states where, like a state university system, the UC system in California, or the UT Cal UT Texas system, or the

17:29 the Florida system, or New York SUNY system decides that they want to have their own sovereign AI just to be shared among their state academic institutions. Stephan Fabel: I mean, it's not not a new idea in general, right? I mean, we had national labs for a while now. A lot of these public research institutions have had their own HPC data centers, and they've been operating them for a while. It's just that the the necessity, I think, to create the you know a data center and access that is in this environment is sort of it's counterproductive to trying to build that on your own, and I think that you know, much like maybe the industry at large has moved towards rental more. Is I think that the idea here is as well that public research

18:18 institutions, in order to keep up, they want to get access to the latest and greatest, but you know the latest and greatest is going to be not the latest and greatest anymore in like nine months. So how do you how do you keep up and how do you attach to that motion from a from a public perspective or public sector perspective is incredibly challenging because we all know the time it takes for the government really to do anything, and so you know, getting them access and letting them sort of bypass a lot of the the processes that would normally hold up or slow down the acquisition and then the rollout and the the making available the launching of of such an infrastructure is I think, the biggest value that we

19:04 bring to the table for them. And so, they're, I mean, look, they're they're very very motivated to to make this happen for the researchers. And it's not just a matter of attracting the right talent. It's not just a matter of attracting the right funding. Although those two things are incredibly important, it is also ultimately advancing their research and letting them leapfrog in a field that that you know otherwise may just there be might be left behind or make not might not make any progress at all.

19:37 Joshua Feinberg: Yeah, it's fascinating because a lot of times this research investment flies under the radar screen. Definitely not lost on me. Actually, my first job in the IT industry, going back the early '90s, was working for IBM Academic Information Systems Higher Education, and the professors had big budgets for hardware and software at the time, and then I ran with. Had a system builder business for a few years after that, and at the time, the software vendors were willing to really in the hardware, but Apple and IBM were willing to really, really subsidize early adoption because they know that if someone working on their master's degree, PhD, even undergraduate, got indoctrinated on that particular platform, once they go out into their real job and

20:20 their purchase authority, it makes a huge difference in preferences. Stephan Fabel: Absolutely, and it's and it's an obvious thing. I mean, we talk to researchers all the time. I talk to somebody who is responsible for hypersonic flight mis research, and that's of obvious importance to the the national security. And they don't have access to the GPU resources that they need, they have to wait for for six months or nine months. Actually, is what he said. He wait he waits for nine months to get six minutes on a on a national lab, and makes you know incremental progress. And so the the notion that that that team would get access to large scale infrastructure on a you know pro long time in a dedicated manner that really makes a difference

21:06 and really pushes that research forward is it's more than mind blowing it is I mean there were tears in eyes right I mean they they are really motivated by that mission and so we're tying to that Joshua Feinberg: that almost sounds like a product manager at a Series B, Series C SaaS startup may actually have access to more GPU resources because of their their board members than than someone in a university is working on issues of national important national security importance. Stephan Fabel: It's really sad. It really is. It's a I when I was at NVIDIA, one of the things I was involved in was a an initiative where we worked with the NSF to provide access as part of a grant program, and so even very you know small amounts

21:54 of compute made a huge amount of difference to these researchers, and they would tell us all the time how much of a difference even you know 20 GPUs would make to their to their advances, and we you know we we saw just great success come out of that program. So that was a big portion of why I decided to focus on this.

22:18 Joshua Feinberg: Given your vast experience in Silicon Valley and other parts of the tech ecosystem. Do you have a favorite go-to-market playbook that you found to be effective across multiple stops in your career journey?

22:33 Stephan Fabel: I think that if you're a part of cutting-edge technology and you are implementing this for the first time. I think it is very. I mean, it's always a matter of trust, but it is a very important aspect to have a technological leadership. You don't have to be better than anybody else. You just have to be trustworthy and show integrity, and you have to be factual, and you have to be competent, and those things. If you show up and and deliver those things, the it's it's not really a go to market strategy. It's much more of a one on one sort of personal level between you and the decision maker. You can actually make something happen. You can actually move the needle, and you can actually implement something that hasn't been done before, and that

23:27 requires a certain psychology on on the sides of both the the vendor as well as the customer. So, I work best with builders. I work best with people who want to move the needle. This is a big portion of why I'm doing what I'm doing right now. But it is also almost a self selection with you know of who I work with. It you know there's there's so much to do right now, and I think we have so much to catch up on in terms of our ability to to translate all the advances that we are making in the accelerated computing infrastructure, and in fact, computing in general, that we need to catch up, and we need to we need to deliver that as fast as possible to the industry and to the people who need it most. And so, I think a go-to-market

24:19 strategy for me is, you know lining up the mission, filling it with really good people, establishing a a joint trust with the customer, and then moving the needle together. And I think that is something most can get behind. It's not for everyone, but it's what works for me.

24:39 Joshua Feinberg: Really interesting that you brought up the idea of trust and educating. And if you look at some of our other parts of the data center industry right now, especially on the developer side, especially on the hyperscaler side, they're facing some very existential threats trying to win over community stakeholders for. Fortunately, you're selling more B2B and less B2C, so it's a little bit different. With all of that in mind, most data center GTM teams are flying completely blind. 83 of your buyer's journey is happening before they even speak with someone from your sales team. The Data center go-to-market podcast is powered by DCSMI. We've studied over 1,900 industry leaders to build a diagnostic framework that identifies exactly why deals

25:36 fall apart and revenue stalls. Stop guessing and start benchmarking. Subscribe now to our weekly data center GTM briefing at www.dcsmi.com/briefing. Again, that's dcsmi.com forward slash briefing. B D R I E F I N G. When you look at other parts of the data center ecosystem, what do you see as a go-to-market strategy right now that a lot of companies are still using full steam ahead, but that you feel is rapidly losing momentum and a lot of people are in denial about? Stephan Fabel: So, I I got first active in the internet in the early 90s, and back then there were news groups. And when you were a Usenet on news groups, you you know you'd have things like little signatures that were kind of funny, or you know

26:35 snarky or something like that, just kind of sign off things.

26:40 And one of them, I remember, had, you know, had the saying. It said, "Of course, it doesn't work, but look how fast it is. And I think that if you are enamored with your own solution and you're putting forward how fast it is, you forget whether it works or not, and you're actually not very useful, so it is a shiny object. But you know, as as I would say, a large percentage of usage of current AI tools teaches us there are some rewards in how you do that, and so not being transparent about what it takes in order to use this responsibly or correctly, even not adding that educational piece lends itself to misinterpretation. There are great people working at all of these companies, and none of this is, I think, intentional. I

27:35 think we're just blind to it because we're technologists, and we're not really relating to somebody who is not, frankly,

27:45 Joshua Feinberg: one of the observations that I've made of NVIDIA in recent years is they're running about as good of a training and certification playbook as I've seen anyone run in the history of the IT, and that goes back from the standpoint of following and participating in Microsoft's certification ecosystem and a couple different SaaS companies training in certification ecosystems. However, not all technical content works with all different kinds of stakeholders. Generative AI has gotten the finance person used to getting very specific answers to what they care about, and the compliance person getting very specific thoughts on what they care about, and a lot of companies in the data center space and cloud produce mountains of technical content,

28:26 but a lot of chief commercial and chief revenue officers frequently complain and are frustrated because that technical content is not driving actual pipeline, not driving sales. Why do you find that heavy technical content doesn't seem to be as effective as sales teams would like it to be.

28:47 Stephan Fabel: Well, because it's un, I think because it's unfocused. I think it it focuses too much on being you know a a technical documentation and less being a relatable document that teaches somebody how to improve the day job? If if I'm, you know, if I'm being honest, and I think that that that probably goes for a lot of us, you know, if I look across the vast majority of of the folks who are using AI, they're using probably one of the tools that's made available to from one of the big companies that are out there, right? They may be using Chat GPT, they may be using Copilot, Perplexity, what have you. So they they're using that, right? But they're they don't use it as you know someone in accounting. They're not using it as someone

29:36 in sales. They're not using it as someone in etc. And so they're using it as a generic user, and so while these tools offer ways to to customize the answer and to you know be more appreciative of of the user's situation, that has to be taught. That's not just happening out of the box, and I think so. The expectation from a a layperson on. Thing is that it is AI. It already knows everything that it needs to know, and I just ask it, and it does it, you know, as if I did it. And nothing could be further from the truth, right? You have to learn how to how to use it correctly. You have to teach it, personalize it, and ideally, you have to make it very very concrete in terms of the data set and the memory that it

30:22 carries to your day job, and so I think a lot of the the problems that we have is that number one, that content is only focused on actually using the interface, but not an outcome focused delivery that helps you actually in your job. Number two, I think that the product itself is generic by nature and not necessarily tied to your specific use case, and so you know the the ability to to kind of twist these things into the particular direction and make them work for you is fairly limited compared to something that might actually have a model in the back that is specifically trained or tuned to your domain-specific purpose. So,

31:08 Joshua Feinberg: do you see that across different stakeholders, especially on a decision committee? Do you think their expectations on relevance have changed quite a bit in the last few years? What they might have tolerated or accepted as close enough four or five years ago is not nearly close enough now because they can get a lot more precision. Stephan Fabel: I think that I think that there's absolutely more sensitivity to it having to be a slam dunk, and I think it has to do with mainly the amount of money that is involved. I think that past initiatives were a lot less capital intensive. Right now, an investment is you know you add two or three zeros to everything, and you know that just changes everything. It

31:52 changes underwriting. It changes investment decisions. It changes valuations quite dramatically. And so, you know, being able to afford a mistake here is a much more high stakes poker game than than before. I would imagine. Joshua Feinberg: When I think of even your previous employer, Nvidia, I remember five or six years ago, three or four years ago, family members and friends asking me if I'd ever heard of Nvidia. I'm like, oh yeah, they make the graphics card. That's the upgrade when you buy a Dell desktop if you're a gamer, Stephan Fabel: that's right. That's what people knew Joshua Feinberg: them for. Yeah, the stakes were much lower. It was probably $100 upgrade at the time of purchase. Stephan Fabel: That's exactly right. And so, you know, we've

32:31 all seen what happened, right? I mean, with all the short. And in fact, this is how DJ Exclude was born, right? I mean, we we originally just thought to to alleviate supply chain issues caused by COVID, and was simply you know shortages of of parts, and so we needed to help customers that had ordered gear and had ordered GPUs. We needed to help them you know get productive earlier, and so we decided to to host some of this stuff, and you know it ultimately evolved. But at you know but at the end of the day, what we introduce back into the equation for our customers with us, and this is fundamentally what is being provided today by other neoclouds as well, is this notion that there is a predictable timeline to get

33:15 access and expert help and guidance alongside that access, and I think those two things, if they're done correctly, and they are done thoughtfully, and they're done alongside the customer, I think those are necessary to drive the success.

33:34 Joshua Feinberg: Stephan, when you think about the overall economic buyers on these major sovereign AI deals. You can win over engineers. You can win over architects. Sometimes easier, sometimes more challenges. But a lot of times, these deals are stalling out at the top. How do you bridge the gap between the technical champions and the ultimate economic buyer? Whether it's a CEO, whether it's a CFO, whether it's a president of a university.

34:03 Stephan Fabel: Well, I mean, the I think the first thing that that we have to acknowledge and that that we have to help everybody else understand is that, and I think it's intuitively clear, but you know, it helps to spell it out. IT is traditionally a cost center, and so traditionally you would have a very very different conversation about you know a data center and the compute infrastructure that goes inside of it. However, these days a data center is actually attached to the revenue, right? It's attached to the funding, it's attached to revenue, it's attached to the business, and so making that investment decision is actually, it's actually really a business plan. It's not, it's not really a cost justification. So going in correctly with with that as a

34:53 business plan and making sure that we are, you know, getting that we're we're providing the output that they're that they're seeking. And the outcome that they're seeking is what we're usually doing in order to get that over the hump.

35:07 Joshua Feinberg: Can you share a story of something that you've tried in the last couple of years, or ideally something that you've tried so far at This Way Global that you thought would work, but just didn't pan out as you thought, especially when it relates to the commercial side of the business and go to market, and what ultimately you learned from that experience. Stephan Fabel: Yeah, it was. It was. You know, and I think it is timing, really. In in in all cases, I tried lots of things that didn't work. By the way, I mean, the good news is, you know, I was able to do those with what I call the four green walls of comfort. You know, so I I I was able to try stuff out and you know and experiment and learn, and I think that was a a

35:47 big big benefit to us and to the team, but certainly we learned that at least up to a couple of years ago, having a pure you know abstracted away software platform that that basically hides the the GPU underneath. The market wasn't ready for that. The market wanted to know exactly you know what what GPU their their workload runs on. You know how long they would need for a certain workload to complete a certain job to complete. How many of those GPUs they would need and kind of they wanted to in other words they wanted to turn all the the dials and and twist all the knobs and and all those things so they they really looked at kind of how can I get under the hood and understand exactly what this you know this infrastructure is

36:38 doing for me because I need to peek out every last inch of performance. What we then learn, though, that it is really just about the buyer, and you have certain buyers that are just not ready for this conversation, and you have other buyers that are simply looking to add AI to every single application that they're looking to roll out internally, and so they have a very different way of thinking about the problem. They're looking at this as you know a developer, a cloud native developer problem, and so you know for them consuming APIs is very natural, and that's what they're expecting a a cloud to provide, and even a GPU cloud to provide. So you know we certainly had our mismatches in trying to figure out who would,

37:23 you know, buy what and for what reason, and it was a while until we figured out that, you know, large scale training typically requires a lot of white glove service alongside with it, regardless of how mature and advanced the customer is. There's always going to be a tight integration between operations, the cloud provider, the infrastructure provider, as well as the the consumer, and then you know there is another way to consume a cloud, and that's much more API driven and much more hands off, I guess, from a from a cloud provider perspective. Joshua Feinberg: When you were originally trying to get traction, and they wanted a different level of transparency. Did you find a lot of the people that you were speaking with were

38:05 trying to make comparisons to what they understood about at that point? One mature public cloud. Oh, Stephan Fabel: 1,000% I mean, they Joshua Feinberg: they were looking at you like saying, "Okay, I know how AWS works. I know how Azure works. I know how GCP works. Tell, is this another version of that? Stephan Fabel: That's right. I mean, like we we had a lot of bake-offs. I mean, in the beginning, really a lot of people were saying, "Okay, I'm going to reserve an instance or several of them on AWS and Google and DGX Cloud, for example, and then we're going to run the same benchmark and see you know whether you guys match up to that. And so a lot of that was in the beginning because we had to establish a presence as a

38:47 provider and gain the trust of the folks who are looking to consume. And so, you know, we had to certainly pass that gate.

38:56 Joshua Feinberg: Stefan, another area I want to ask you about today is in the rush to scale, especially with startups, a lot of times companies look to outsource everything from sales development to content to various parts of events. What do you think is one core go-to-market or operational capability that you would advise leaders to never outsource and to make sure that they keep core to their internal team? Stephan Fabel: Well, I mean, besides leadership, of course, I think the most important thing is the operations teams themselves. I think I strongly believe in in-house operations, especially if you are a service provider. I think relying on a you know outsourced approach here is frankly going to be very

39:44 very hard to to do. This has to do with the SLA fulfillments. I mean, being able to go after somebody who's you know messed up and who's their partner, they're likely not going to be able to live up to your expectation. It's just not gonna end well. I think if you are a cloud provider, then being able to run a cloud and operate a cloud is a core capability that should be in house.

40:10 Joshua Feinberg: You find also that there would be a loss in fidelity between potential product roadmap lessons learned and insights from end clients and feeding them back into yeah Stephan Fabel: yeah certainly. I mean, you you don't want to transfer your learnings to potentially other clients of that company you're working with.

40:30 Joshua Feinberg: Final area I want to get your take on is when you look into your crystal ball and you think ahead with a 1218, 24 month time horizon from now. Where do you see data center go to market evolving. What do you think is going to be the same? What do you think is going to be a little different about what it takes to grow a company like This We Global? Stephan Fabel: Well, I think that a lot of it is has to do actually with the you know with what we just talked about, right? The idea that there's a lack of transparency and a lack of education that goes along with the sale and with the, you know, with the with the go to market that these companies bring to the table. If I if I would put my if I if I put my

41:18 crystal ball on this, I would say that we'll see much more closely regulated data center rollouts. I think that there is going to be something coming down the pike from from the federal government, and I think surely the states about rules and regulations of how to you know build and how under what circumstances are you getting the permitting done for your data center? I think that is closely tied to our you know the question of what are we doing with this data center. So I think that that will make a huge difference, right? So if your if your data center is is only there to serve the few or is only there to serve Silicon Valley companies that that you know may not have your best interest at heart. It may not

42:06 also be very obvious to you why the data center has to be built in your backyard. But if the data center is actually attached to research and is actually attached with a workforce development program that you would like to send your your child to because knowing that if they go through this workforce development program at the data center, for example, they are able to secure 30% higher income on average. Then you know that's quite transformative in the perception of what a data center is or should be, and I think that that's that's a big portion that's going to be leaned on more heavily, I think, is this idea that that workforce development, data center rollouts, so this job creation aspect of it is probably going

42:53 to be heavier. I also think that the impact to the environment will just will establish federal standards, and then as long as those are followed, etc. I mean, we learned how to build airports, right? I mean, there's no reason why we shouldn't learn how to build data centers. So, 100%

43:10 Joshua Feinberg: agree on the workforce development part as well. I actually feel that a lot of the community backlash could have been avoided if the developers did a better job of partnering with the trusted K to 12 and higher education institutions in the community. Not that they're perfect either. Higher education, for sure, because of the student loan crisis, is facing its own issues. But when an anonymous or

43:38 despise publicly traded or mass developers coming in and at the last minute using most heavy-handed lobbying legal tactic compared to the president or a dean of a college, like it's no comparison who they're going to trust. What week? I look even in my own home area. I'm in South Florida in Palm Beach County, about an hour north of Miami. The two public schools here have been out hitting the pavement hard in the last six months to a year talking about their quantum computing academic programs. I don't think they actually have very many professors who are ready for this yet, let alone a curriculum. But they for sure are doing a marvelous job of building awareness of that. I think they've even gotten some

44:18 federal grant money. Stephan Fabel: I mean, build quantum. Yeah, that's that's what it is. I mean, 20 years ago, it was robotics, right? I mean, everybody had a robotics program, whether they had one or not. That was just that was just a thing, right? And so, we'll see the quantum program right now. And by the way, quantum research needs a lot of accelerated computing to make happen. So, in order to establish a quantum center of excellence, if you will, there is an associated accelerated computing, you know, GPU-based accelerated computing infrastructure project alongside with that, and so if if your your schools or your your colleges are intending to do that, then almost certainly there is a data center rollout,

44:59 or at least. It's a question of an access to data center, a part of that rollout. Joshua Feinberg: Yeah, I agree that the construction, real estate, developer side for sure is going to force some really interesting conversations nationally and at the state level about how data centers are provisioned, site selection, resource utilization. I'm curious in your very specialized segment of this industry, sovereign AI. What are you thinking about with like a 12 to 2412 to 24 month time horizon? Do you anticipate that there's going to be significant regulation of the models and who can have sovereign AI, how it's utilized? I Stephan Fabel: would imagine that there is going to be a substantive conversation around the definition of sovereign AI

45:45 and what is a sovereign data center and what isn't, and then it's just a matter of compliance to that standard. I'm pretty sure that we're going to have, you know, if we're not doing this right, we're going to have the opportunity to see a lot of non-sovereign sovereign data centers, you know, data centers that fulfill the compliance, but that don't actually do anything for the for the good of the public. We hope to continue to differentiate here, but in you know in about 24 months, what I would expect is is that we have definitely spoken with and started initiatives with all the 50 states about that plan, how do you let your local economy participate in the global AI boom? How do you bring your workforce, your local workforce,

46:35 into the age of AI, and what's the right path forward? And that isn't just a policy question; that is also a technological and a planning question. It is a it is a cloud offering question to me. It's because it is shared access to compute, and so that in itself, at the core of it, is if that's not done right, then we'll end up right back here again in in you know in two years where most of the access is done, or is given to you know the precious few companies that you know where you find like you know half of them are the customers of the other half, and you you essentially you know have the circular money going around. So Joshua Feinberg: I think there's a lot of workforce development conversations that need to happen just on the pushback of

47:24 oh we've heard there's only going to be 20 full time jobs in 2 million square feet is this really going to create jobs no no no wait there's a whole ecosystem there's dozens of different kinds of businesses that build and maintain Stephan Fabel: yeah and so hyperscale Joshua Feinberg: facilities Stephan Fabel: there is a lot of lobbying involved I will not lie about that I mean, this is, but it is essentially education to say there is a plan. There's a pragmatic plan out there that actually does do the right thing. It it is transcending sort of this binary, maybe mid term motivated characterization of what a data center is or isn't. But I mean, at the end of the day, I don't think that that we should necessarily suffer

48:04 from an insufficient definition of what a sovereign datacenter is, what it brings to the table, and how we could all benefit from it. And so that's what we're here to solve. So that's what we're looking to engage in, and certainly in two years' time, I would expect to have had that conversation at all levels of the government, Joshua Feinberg: I think to the idea of customizing and personalizing the education is going to need to become a lot more granular. I had someone on the podcast a little over two years ago as a marketing leader at Cyrus One, a veteran promoting person for the colo space, and she explained at the time they were thinking of educating elected officials, regulators, utility companies, and community activists. And I

48:51 think the one area that everyone underestimated is that fourth area. You can have a rational conversation with full-time people in planning and zoning, and your mayors and councilmen and state assembly and Congress, and you can certainly have a great conversation with people that work for the the local utility company. When it comes to the B to C side, it's way, way, way more challenging. It's so much more fragmented. Yeah, Stephan Fabel: it is. It is. But it's also, you know, it's a matter of what do I read in the news about it, right? I mean, if you imagine right today, I said this before, so I'm going to say it again. I mean, it's a little bit, you know, provocative, I guess. But the way that I would frame it is, is today we're

49:30 we're using AI basically for trivial stuff, right? Cat pictures, have it do our homework, you know, things like that. And but the serious side is we're using AI to justify the dismissal of a lot of jobs, right? Of a lot of people losing their jobs, and so if those are the three things that are in the news and that are being shown and that people read about, then they're right to ask why should we? Support this. Why should we continue to do this, right? If we were to use it for matters of national defense, if we were to use it for cancer research, dementia research, and so on and so forth, I think we'd have a much better case and much better basis for the public education about what it is that we're all about. But I suggest humbly that

50:22 you know everybody who's involved in this, you know, in this industry, whether that's the developer, whether that's a data center owner, whether that's somebody who's a data center operator. Certainly, cloud. I think we all have our work cut out to explain how we benefit everybody else around us, rather than just you know show profits. Joshua Feinberg: And a lot of the developers are going to the playbook of your taxes will be lower, and we'll build beautiful schools and beautiful parks and beautiful hospitals. And like, I feel like a lot of this is product management, and it's trying to figure out along the way who cares about what, what to build, and the different segments and how to explain it, as opposed to thinking that you're going to walk into a room

51:09 with 400 angry people and they're all going to care about the same two or three issues. Stephan Fabel: I mean, from my perspective, if I'm building the right thing in the first place, I don't have to bribe you to look away when I do the wrong thing. So, I I shouldn't have to build a a new school in order to be able to get a data center somewhere, you know, like extract exactly like that's not my that's not what I do, but I I think that's not what what anybody should should be doing. I mean, they should instead be doing the right thing to begin with. And so, you know, I think you know I'm excited to be part of that conversation. I feel like there's a lot to be said here, and I feel like there's a lot of contribution that we can do and bring to the table as

51:49 this way global. And certainly, I'm excited to see where that goes. Joshua Feinberg: Stefan, for someone that's watching, that's a peer of yours in a leadership role at an early stage startup or scale up. What's one strategy that they can take this week that will put them on a much better path for success with scaling their company from a go-to-market perspective?

52:15 Stephan Fabel: Wow, that is a very hard question to answer on a general basis. I would say, I mean, beyond the the generic, you know, measure everything that you do and act on the data. I think what we lack often as an industry is the ability to read the room. I think we need to start listening a little bit more of what the customer is telling us. I think we need to we need to not lean out the window too far in in what we're willing to say and what we're willing to commit to. And I think we need to reintroduce a little bit more of a you know handshake level trust in in this entire Industry, there's almost feel like a weaponization that's been going on, you know, between the sales teams, especially, you know, as as large hyperscalers

53:12 got introduced. I think that just accelerated a lot. The tone's gotten harsher. I think Main Street is not ready for something like that, and they're not ready to deal with somebody like that. I think they they really want somebody they can shake hands with and and you know establish trust with. So what can they do this week? They can call you know their best customer right now and they can tell them they appreciate their business and they can establish contact without having you know to ask for another order or without having to do anything else, just saying thank you for your business and looking forward to continue to work with you. I believe civility will go a long way here. Joshua Feinberg: I thought something that should be a

53:52 close, yeah, a skill. That's right, or an MCP. Stephan Fabel: There you go,

54:01 Joshua Feinberg: Stefan. It's been fascinating. I really enjoyed our conversation today. Your thoughts on how you think about the marketplace and and building trust and different kinds of interests and winning over different kinds of stakeholders. If someone wants to follow what you're working on, reach out to you. Is LinkedIn a good place to start? Stephan Fabel: Absolutely, please. Joshua Feinberg: Yeah, we'll make sure to include links to your profile and your page and the website for This Way Global as well. But really, really appreciate you joining me today. I've been speaking with Stefan Fabel, who is the CEO of This Way Global. Stefan is based in Silicon Valley area, and company is headquartered in Austin, Texas. Thanks again, Stefan.

54:45 Really appreciate it. Stephan Fabel: Thanks for having me, Joshua. Joshua Feinberg: Most data center GTM teams are flying completely blind. 83% of your buyer's journey is happening before they even speak with. Someone from your sales team. The data center go-to-market podcast is powered by DCSMI. We've studied over 1,900 industry leaders to build a diagnostic framework that identifies exactly why deals fall apart and revenue stalls. Stop guessing and start benchmarking. Subscribe now to our weekly data center GTM briefing at www.dcsmi.com forward slash briefing. Again, that's dcsmi.com forward slash briefing. B R I E F I N G.