Broadcom, VMware & the Future of IT: Cloud, Edge & AI with George Crump

Bruce: Welcome to PodMagic. Real conversations about solving real IT Challenges. I’m Bruce Kornfeld, your host. I am the Chief Product Officer at StorMagic. And we’re always exploring how simple, reliable technology can help you and the people that you serve, whether that’s retail branches, manufacturing sites, hospitals, whatever you’re doing for a living, where you’re serving people on the front lines. That’s what we’re here to talk about. My goal is to always bring interesting guests, deliver some value and have fun along the way. So, let’s dig in.

We have George Crump here today. I’m very excited. I’ll do a quick intro of George. He is currently the Chief Marketing Officer at VergeIO And prior to that, he was the founder and lead analyst at his own company called Storage Switzerland, where he spent years breaking down complex IT trends  for CIOs, IT pros and vendors across storage, virtualization and data protection industries. Welcome, George.

George: Hey, Bruce, thanks for having me. Looking forward to the conversation.

Bruce: And for full disclosure, George and I have known each other for more years  than we want to admit. Mostly when I was at a storage company back in the day, and George was doing his analyst thing at Storage Swiss. We got to know each other back in those days. Looking forward to a lively  conversation today. All right, so, George, let me dive right in with a I don’t know if you want to call it a softball question, but an easy one. But let’s just start with talking about IT teams and, I think we’re seeing the same thing, which is a lot of people are reevaluating virtualization platforms at this time. What are you seeing? And why do you think it’s happening?

George: Well, I think the big elephant really is Broadcom bought VMware and has proceeded to fulfil all of our expectations  on worst case scenarios. They’re sending cease and desist letters  to their own customers. I personally like to thank Broadcom for doing that. That was awesome. But even before that we saw a trend where, in the area we play, which is very large enterprise, we saw an issue of scale. That you just couldn’t build systems big enough. You couldn’t do things like multi-tenancy and things like that. So, there were issues there. And then, I think obviously the Broadcom acquisition just put fuel on the fire. And it’s been off to the races over since November 2023.

Bruce: I just thought of this, what if Broadcom didn’t do what they did. What if VMware was still its own entity doing its thing. Do you think the industry would be any different right now. Clearly there’d be less interest probably in switching, but I just wonder what it would be like then. Any thoughts there?

George: Like I said, we were seeing a good uptick as a result. But we were very focused on  the high end enterprise cloud service providers. And those companies pain was real. But obviously you had a well-funded, entrenched competitor, that made that process go certainly a little slower. I think from our standpoint, I think both companies would be fine. We just probably wouldn’t be seeing the accelerant that we do right now.

Bruce: So those of you that haven’t picked up on this yet, VergeIO and StorMagic can be thought of as competitors. The nice thing is, we’re going after a massive market. Verge seems to target the high end data center, StorMagic’s targeting small and ROBO and edge. So, it’s a nice little cooperation we have here. Just an opportunity for us to talk.

So, I can give my thoughts on StorMagic, but since you’re the guest, I’ll ask you first. Which is, what are you guys actually seeing in terms of are customers migrating today? Are they tired? Kicking? Are they thinking, are they pontificating or are you seeing actual movement? Like we’re done with VMware, we’re making the move. We’re going now. What do you see?

George: Well, if you take it from November 2023, the initial reaction was very, very steep. We saw massive adoption within that first quarter. These were people that were already on the bubble, had experienced a Broadcom acquisition in the past, and didn’t want to go through it again.

And then what we saw, and this makes more sense to me, we saw a lot of people, either run in an unsupported state, and just run the string out a little bit. Or more common, we saw a lot of people renew their VMware license for one year. So, there was a little bit of a, I wouldn’t call it a slowdown because it was still faster pace than pre-November 2023. It wasn’t the hockey stick that we saw in Q1 2024. But now we’re starting to see that elevate back up.

Bruce: We’re seeing some of the same things here. It just seems like in the last, maybe the last few months, maybe it’s the start of the calendar year. It’s just been, the checking out, the POCs, the tire kicking is now starting to turn into reality. Which is good for everybody. I will say, on a related note, there’s a lot of customers, I don’t know if my producer will let me, am I allowed to say pissed off? I think I’m allowed to say pissed off. There’s a lot of pissed off customers out there at Broadcom. But you got to give them credit though. I mean they’re brilliant. What they’re doing is brilliant. They locked everyone in. They’re raising prices and they’re like, come along if you want. But if not, we’ll see ya. But it’s working. It is working. They’re very smart people.

George: Well, they took their most profitable customers and kept them. And they got rid of the companies that can’t justify that expense. They’re getting rid of them essentially.

Bruce: So, I think, in general, and I don’t know the Verge product line well, but in general, you guys play in a similar role that we are, which is on prem. On Prem is basically, you’re selling software that customers use in their data centers. And we’re selling software that they use at their edge. So, what do you see these days about cloud? And do you see a movement from cloud to on prem? Do you see new applications? What are your thoughts on the whole cloud computing, where that’s going in the industry?

George: Yeah, it’s an interesting question because when I joined Verge,  I knew right away, and you would have done this too, you got to pick a thing and go do it, or a small company, or certainly smaller than VMware, we can’t do seven different initiatives. Right. And the two initiatives we came down to were either cloud repatriation or VMware alternative.

Fortunately, we chose VMware alternative. But there was a trend, especially back then, that if you think of that era, that’s when, Hey slash Basecamp, did its big move off of the cloud and they were making some publicity about that. I’m glad we didn’t, because what I’m really seeing now is what I would describe as more of a, cloud rebalancing, where there are some workloads and some organization where it does make sense, to put stuff in the cloud.

But I think people are also coming to the reality that there’s a lot of workloads that don’t make sense in the cloud. And if you have good infrastructure software, self-promoting, but like yours and ours, the complexity that you’re trying to get away from by going to the cloud doesn’t exist. And so, the reason you’re going to the cloud might be you have an infrastructure software problem more so than infrastructure on prem is any harder than in the cloud. And so, I think what we’re seeing anyways is a rebalancing the workloads. If it makes sense to put on prem, put on prem, the workloads that should go in the cloud, put them in the cloud.

Bruce: One thing I was thinking is, if you think about this like, the cloud has been, whether you’re an entity that’s cloud first, cloud only, but most enterprises have been heavily invested in cloud for all the reasons we know. But now, technology is so amazing for going back to an on prem.

Servers are so much more powerful and less expensive and all those reasons. Building your own infrastructure isn’t that expensive anymore. Managing it isn’t as hard anymore. But now I’m wondering. I just thought about this the other day. Maybe it was today. This morning. If you’re an IT person and you’ve spent the last 10-15 years cloud, cloud, cloud, do they even know how to build their own? Is there a talent gap now? Because people are like, I don’t know how to build my own stuff. I don’t know, I just wonder.

George: I think there’s a broad talent gap, but I think part of it is, the skills required to run in the cloud are very similar to the skills required to run on prem, right? I mean, at the end of the day, infrastructure is infrastructure. Yes, there’s some nuances between the two. But, the basics of networking, the basics of storage management, the basics of CPU utilization, those all fundamentally remain unchanged.

And I think, frankly, that’s one of the reasons we do see this cloud rebalancing. Because it isn’t that much more difficult to put it on prem. The promise of the cloud isn’t all that great. And one of the things I always tell people is, it’s a business model too. You’re renting something versus owning it. And, like you, I travel quite a bit. And when I travel, I used to rent cars. Now I usually just do Uber or Lyft or whatever, but I don’t do that when I’m around my house. I don’t Uber everywhere I need to, because I live here. It would be incredibly expensive to do it. But for short term bursty things, it makes sense.

And I think that’s where the cloud comes in. If I was trying to develop a new application

or testing out a workload, cloud’s perfect because I can scale it and get rid of it and do all the things I need to do. But maybe when you bring it into production, that’s when you bring it back on prem and hold it. And then there is still the whole data privacy thing. There is the ability to make sure that you can, and I think in your guys world, the ability to operate locally, independently in case something does go wrong anywhere else in the infrastructure has some incredible value.

Bruce: And that’s a big one for us, which is, we make it really easy to put just two physical servers and be able to be an island, if you need to. If the internet goes down, which it does, it has outages. So, customers want to know they can keep running that retail store. Or the factory doesn’t go down or, God forbid, you’re running IT in a hospital. You can’t have outages.

So, there’s a lot of that edge computing, being able to run without a connection to the cloud temporarily is a really important part of our business, for sure. What about hyperconverged? I assume you guys have a similar packaging of you don’t need to buy 15 servers and big external arrays and networks, which is, you’ve got this combined thing, this all in one hyperconverged.

Do you think it’s working well for customers? Do you see any push to go back to three tier. How do you guys see that world, in terms of this, all in one versus maybe a three tier architecture? Any thoughts there?

George: Yeah, it’s an interesting situation because, if you say we’ve got an HCI camp and a three tier architecture camp, the leaders in those camps, neither have done a spectacular job explaining why A is better than B. And so, as a result of that, A, which is the three tier architecture, tends to continue to win because, if it’s not broke, don’t break it type of mentality. And so, I think the onus is on hyperconverged or let’s call it converged vendors to deliver a more, resilient, architecture, make sure that there’s no compromise in going to this type of architecture.

So, our unique property is that all of our code, our hypervisor, our storage, our networking, and now our AI, is all integrated into a single piece of code. And so, that’s one thing that gives us our triple digit scaling and things like that. And so, it allows us to do things and keep things simple, that you haven’t seen anywhere else before. I mean, if you look at the VMware product line, it’s 17, 18 pages worth of stuff? To their credit, Broadcom realized that that was part of the problem and made everyone buy at all. I don’t know if that was necessarily the right way to get there. But it was a strategy. And so, I think that concept is really what we need to examine. I think convergence makes sense. I think that doing things inside a pure storage array. Let’s just pick on them, or, you name your vendor. It’s still essentially an Intel box. Probably made by either Dell or Supermicro, that they put their software on. They buy somebody else’s drives, put it in there, and then somehow the drives, once they magically go inside a dedicated storage array vendor’s piece of hardware, somehow become worth ten x more than they used to. One of the things for me is to let people know that, high end NVMe flash is about 120 bucks a terabyte now. They’re not paying that, they’re paying ten times that from a vendor. And so, if you got a guy that needs 30 terabytes of capacity, you’re talking like a 12, it’s just not that expensive anymore. And so, once they start doing the math on that it starts to make a lot of sense.

Bruce: So, I had another thought here. What about, what do you guys see in terms of getting a customer to make that jump? Because there’s brand loyalty. Maybe you can say Broadcom’s, ruined some of that, but it’s still hard. It’s still hard to give up the way that you do things. They have all their tools.

Do you have any advice or any thoughts to customers out there on making that decision? Like, it’s time to move off  there’s better solutions out there. They’re going to save you money, that are easier. But how do you get them over that line? Any thoughts there?

George: I don’t know about for you guys, but I think every single customer we do a proof of concept for, I would assume that’s going to be the case for a very long time. So, do a proof of concept. One of the things we like to do is add value during that proof of concept process. So, the way our migration function works, it does an, essentially an hourly capture of the entire VMware environment. And so, while you’re doing testing, if you have an outage in your VMware environment, you’ve got, essentially a ready to go copy under VergeIO.

We’ve actually had situations where, we didn’t plan it this way, but where a customer had a, the one customer I’m thinking of had a, dedicated storage array, did a firmware update on that array. Did controller A, it went fine. Did controller B, it went fine. About 30 minutes later, the firmware locked on both controllers and they were done. I always remind people, 7x24x4 is a 4 hour response, not a 4 hour fix. So, the vendor was out for four hours. It took them about 48 hours to fix it. It happened to be a massive hospital. So, that’s a problem. So, the ability to just go, okay, we were testing, but now they’re in production. It becomes very valuable.

Bruce: Yeah, we do the same thing. Proof of concepts. Demo is proof of concept. Check out the software, see how it works. It just gives them the comfort factor. Like, yeah, this is not scary any more, let’s go do this thing.

George: I think most vendors today  in our space, have the migration, we could argue over whose migration is better, but most have some migration and usually it’s pretty good. So, it’s not that. There are a couple of emerging vendors that can migrate between everything, which is kind of interesting. I think the other advice, and this is sort of counterintuitive, given who I work for, I would never be in a situation where I’m solely dependent on one particular vendor for a key component of my infrastructure. I’d at least always be looking at something else because you just never know.

And I’m sure you guys face this too. Customers will say to me, well, you know, you guys are small or, we never heard of you before or, VMware is big and I’m like, well, okay, but how did that work out for you? Because any company can get purchased. That was the lesson. So be ready. I’m not saying necessarily, always run another  hypervisor because I think it would be hard.

But make sure you have tools in place that allow you to move. The advantage of a VM is you’ve encapsulated a server. It’s movable. Make sure you have the tools available to you to be able to move it in case you need to.

Bruce: Yeah, we get that question a lot, how do we know that if we go with you that you’re not going to get bought and we’re going to be in the same shoes three, five years from now? The answer is, we don’t know. But we’re going to help you because a VM is a VM is a VM, and we’re based on KVM, I’m not sure what you are, I assume is something similar. That’s a pretty common core of what virtualization is. So, moving off of us, we don’t want you to. But if you need to, we’ll help you do that and we’ll make it easy. Right. That kind of thing.

So, we clearly can’t have a conversation without talking about AI. I did see that you guys made some announcements a couple of weeks ago, so, what are your views on AI in this world with virtualization and HCI, do you see it becoming a big driver? Is it driving business? Is it something interesting? What are your takes on how this impacts the world that we play in?

George: To some extent, it goes back to the on prem conversation that we were having, in that, AI, there’s a large swath of, let’s call it enterprise or organizational data that, using AI against that data could have some value. And you’re not going to load, or at least you shouldn’t load your proprietary information up into ChatGPT and see what happens.

And so, this idea of privatizing AI, is very interesting. I think, honestly, I said this to our board the other day, that I think that private or enterprise AI or sovereign AI, whichever word you want to go with, could actually inspire companies to get servers on prem for the first time. A small company, and examples of this already, small company, 100% cloud, had all their data essentially on laptops locally. They did government contract, RFP work, things like that. And they are now feeding all of that into what is essentially a private AI engine, to help them respond to future RFPs faster. Well, they don’t want their confidential RFP responses up on ChatGPT, and I’m not picking on ChatGPT, but pick your thing. They’re literally going out and buying 4 servers specifically for this use case. And it’s going to cost them $75,000, but they don’t care. For them, that’s worth it. And so, I think, well actually some of that, which is kind of interesting, but I do think there’s an untapped section of the market now.

At Verge we tend to be very opinionated,  if you will, on the way infrastructure should work. But there’s no denying that every vendor should have a, feasible ability to roll out AI in a relatively quick fashion on prem. How do you do that? We could argue about, but I think that that’s coming. And it’s not going to be like tomorrow. A thousand people are lining up to get it. It’s going to take time, just like everything else. Frankly, just like a more alternative thing. But as that rolls, it’ll be pretty big.

Bruce: We’re seeing similar. Our world’s a little bit different in that it’s small site. It’s edge, but what we’re starting to see a lot of is,  AI is real, as we all know. It is not hype anymore. I would say 6 or 9 months ago it felt a little hype like, but it is real, and it just gets better every week, every month. So, we’re starting to see demand from our customers to make decisions fast on prem quickly. They can’t depend on cloud for everything AI related anymore.

So, we are starting to see some demand to run, maybe the learning still done in the cloud, but local inferencing for fast decision making. But what we’re finding is at the edge. And this is still a debate, I’d say, or it’s still brewing, but do you really need really expensive Nvidia GPUs in all of your servers at the edge to do that? Maybe not. Nowadays, a DL145 with a nice fast AMD processor in there can do some  good damage to what you need to be done, for small sites it can do a really good job. So, I think that whole market is just now evolving and people are figuring it out. But, it’s good for vendors, right? It’s opportunity for all of us.

George: Well, I agree 100%. And I think the other thing is you’ve got to put this into scope. The ChatGPT, I don’t know how many GPUs they have, probably more than one, they service 125 million requests a day. Last I heard. Well, if you’re running AI at the edge or even AI in a decent sized enterprise data center, you’re not going to serve 125 million requests.

And, there’s also some practicality. One of the top use cases we’re seeing in data centers is the ability to have AI help people write code. And the private part comes in there because you’re not going to take your source code, put it up in the cloud, because, for the companies we’re dealing with, that’s their product.

But if it takes

a minute for that thing to help you write that section of code versus five seconds, you probably don’t care. Because it’s going to take you a minute to see what it’s outputting anyway, so if we can output it faster than you can read it right, who cares? And so, we’re even seeing some customers, I’ve got a few customers, there’s like this external GPU thing that plugs into a USB-C port. I see people doing, I would assume in the mini PC area where you guys play, that’s pretty popular. There are also guys that are doing, GPU-less implementations. And so, that might take two minutes. But again, if I’m saying, hey write a Python script that will automatically trigger a snapshot replicated here and update my Grafana, and it takes five minutes for it to write. Do I care? It would have taken me longer than five minutes to write it. So, I don’t know if I care. And by waiting that five minutes it saves you $50,000 in GPUs. Yeah, that makes a difference.

Bruce: Especially if you have 10,000 of those sites. You can’t afford to put that GPU in every single one. I only have two more for you. One is just a general question. What’s next in the industry? In the market? For Verge? I don’t care, answer it however you want. But give me some future thinking, George.

George: Well, I think AI, we’re still in the very, very, very early stages of AI, I jokingly say all the time, most people have used AI to turn an email into a country music song.

Bruce: I haven’t done that yet, but that’s a good idea.

George: My sister is an educator, and I love educators, so please don’t send me emails. But the education system has no idea what to do with this. Because this is more than when handheld calculators came available. And telling people not to use it, isn’t the right answer.

I don’t know about you guys, but our hiring practice now is, because we encourage the use of AI throughout the company, but we expect five x the output. So, you’re not going to work 20 hours a week, you’re going to still work 50, 60 hours a week. You’re just going to put out the results of 5 people. But to your point, this is a very early stage. I think we’ve got a whole decade of stuff to learn about. Before we get anywhere with this really. I don’t think it’s going to get as smart, as fat.

I’ll get in trouble for this, too. But I don’t think it’s going to get as smart, as fast as the AI fanboys say. I think we’ll start to see a levelling off. We see that, generally speaking, in almost every technology. But I think the future is clearly AI as a foundational element. I think, a rebalancing of infrastructure, where it’s a balance between, on prem and cloud capabilities in a more intelligent fashion. I do think edge or remote office,  whatever you want to call it. I think that continues to be a thing. Because connectivity is there.

But other than that, the problem right now is AI is such a big thing that it’s not, I’ll tell you what my personal  opinion is, it’s not robotics because, robots are bad. We’re so far behind where the science fiction said we were going to be. I saw a video the other day, on YouTube or something, of these robots playing soccer I’m sorry, they referred to it as football. I was like, that’s horrible. Four year old kids would do a better job playing soccer, than these robots.

Bruce: I don’t understand it. Maybe it’s entertaining, I don’t know.

George: Robots are.

But there are elements of the human experience that robots are nowhere near close yet. So, it’s interesting to see how that evolves. So, I don’t know if it’s robotics, but I do think it’s this, you don’t think any more when you pull up a calculator on your phone or your laptop or whatever. Oh wow. Look at this calculator. I think we get there with an AI. We don’t think about it anymore.  It’s just there. I don’t think we’re going to work less.

Every time technology is promised to work less, we all end up working more. So, I think that’s going to happen. And I think that’s the reality. We’re going to still put in 50, 60, 70 hours a week and we’re going to be, but we’re going to be putting out five times as much stuff.

Bruce: You’re only working 70? Come on George, what the heck.

George: I’m getting old, I got to slow down a little bit.

Bruce: And then, my last question is an easy one. You can choose not to answer it if you want, but do you miss the days of being an analyst? How are you enjoying your job? Do you like the world that you’re in?

George: I do, I actually really like, what I do right now at this moment in time. It’s very exciting. You’ve got the VMware thing. You’ve got the AI thing. I think, being aligned with, for me, it’s very helpful that our product does everything we say it does. And I would not be a good, I have a friend who’s a CMO who literally said to me,  he doesn’t care if the product works or not. And I’m like, I l don’t know if I could do that, so it’s very helpful.

Bruce: It sounds like EMC from the old days.

George: He said I don’t care if it’s built by a hamster, literally what he said and I’m like, wow. But yeah, I really like what I do. I love the people I work with. I like being on a team. All of those things that you’d  want to get out of that. So, I don’t really miss being an analyst, I do sometimes think, with the ability to do AI and things like that, what my job would be like now.

One of the things you get  asked to do the most, six years ago, seven years ago was somebody called me up and said, hey, I want a competitive paper of EMC versus Compellent. That’s hard. Now with Perplexity and all these different tools, it’s not that hard. You can get all the data you need. You can limit the search. You can do all the things you need to do. I don’t know about you guys, but when we have to put together a competitive document on, say, Nutanix, it’s relatively easy to do. And you can triple check it and feel pretty good that you’re accurate. So, it’s kind of interesting. But long answer to now, I love exactly what I’m doing right now.

Bruce: All right George,

thanks so much for the time. I appreciate you jumping on PodMagic and being here. And hopefully we can do this again in another world someday soon.

George: Thank you for having me, I appreciate it.

Bruce: And for those of you that are listening, thanks for joining PodMagic. Real conversations about  solving real IT challenges. If you enjoyed the conversation. Like it. Subscribe. Tell your grandma to subscribe. Whatever you need to do, give us some airtime. Thanks very much. Until next time.