On this week’s edition of MI&S Datacenter Podcast, Moor Insights & Strategy co-hosts Matt, Will, and Paul analyze the week’s top datacenter and datacenter edge news. This week they talk Cisco, Google, Nutanix & Dell.
Watch the video here:
Listen to the audio here:
Cisco Launches AI Defense
https://moorinsightsstrategy.com/research-notes/cisco-ai-summit-highlights-need-for-safe-secure-ai-enterprise-controls/
Compute = Quadrillions X
https://www.forbes.com/sites/moorinsights/2025/01/28/googles-105-qubit-willow-chip-achieves-major-quantum-milestones/
Nutanix and Dell Get Cozy
https://moorinsightsstrategy.com/research-papers/research-paper-ai-in-the-modern-enterprise/
Disclaimer: This show is for information and entertainment purposes only. While we will discuss publicly traded companies on this show. The contents of this show should not be taken as investment advice.
Matt Kimball: And by the way, our dear friend Paul, we are live, but I just want to point out that Paul has a new camera, and man, it brings out his features, doesn’t it?
Will Townsend: It’s beautiful.
Matt Kimball: He is a-
Will Townsend: Very, very nice.
Matt Kimball: … quite a handsome guy. All right. Hey, good afternoon, good evening, or perhaps good morning to everybody in Moor Insights land. My name is Matt Kimball. I am joined by Will Townsend and Paul Smith-Goodson with the bi-weekly Datacenter Podcast. Today, while we usually go for about 45 minutes or so, today, we have a shortened version of it, some quick hits on some thoughts we have, and we will resume our regular programming in two weeks. With that said, we’re going to have some fun, we’re going to talk some interesting stuff, but do not take anything we say as investment advice or any hard decision-making, life decision-making kind of advice from us. This is for entertainment purposes only.
However, if you’d like more information, you can always reach out to us online. Before we start, like us, love us, and make sure that you tell all your friends to listen to us in two weeks as well. So quick one, we’re going to talk about Cisco, we’re going to talk about Google, and we’re going to talk about Nutanix and Dell and their dating relationship, which is going to be a lot of fun. But I think we’re going to kick off with my dear friend Will Townsend, Mr. Networking and Plumbing, to talk about some Cisco stuff. Right, Will?
Will Townsend: Matt, you’ve been calling me the plumber lately. I don’t know if I like that or not. Just don’t call me Roto-Rooter Man and I’m good with that. Hey, so yeah, I’m going to hit this really quick. So I want to talk about Cisco and they launched something called AI Defense, and I published a research note that’s on our website if you want to go a little bit deeper. It was based on an event that occurred in Palo Alto in mid-January that I was invited to. It was Cisco’s AI Summit and AI Defense was a big part of that. And what AI Defense is, “It allows enterprises,” according to Cisco, “to deploy, develop, and secure AI applications with confidence.” And there are three things that it focuses on: AI application discovery; runtime security, which is pretty critical and hard to do; and model validation. So again, if you want to learn more, go hit my research note.
I do want to talk about the model validation. I think this is really interesting and Cisco is claiming that it is unique to what they are doing. Cisco acquired a company called Robust Intelligence last year, and basically what Robust Intelligence allows Cisco to do is automate the whole notion of red teaming, which is a very traditional cybersecurity practice where ethical hackers can attack a certain application or a large language model to determine its weaknesses. And what Cisco claims is that with this Robust Intelligence integration, that they can scale it well beyond human capabilities into the billions and trillions. And so it’s interesting. So the solution will be GA in March. So I’m not going to call it a winner yet, but I believe it’s got great promise. So Matt, what do you think about all that?
Matt Kimball: I love that. I love the model validation and I love the fact that they’re using automation to do their red teaming. Red teaming, you’re right, it’s been around forever and it’s a great technique. Red team, blue team, find, remediate, find again, remediate, go back and forth. The problem is it’s being run by humans, right?
Will Townsend: Yeah.
Matt Kimball: And while you have great hackers that are, usually great hackers, that are ethical in nature and looking for all your vulnerabilities, they’re not perfect. If you can build this into an automation tool, my guess is it’s going to be far more thorough in finding vulnerabilities and exposing those vulnerabilities, which is going to lead to a much stronger defense. So love the story. That’s fantastic.
Will Townsend: Yeah, yeah. Jeetu Patel that leads product for both networking and security was the MC, and what he shared was there has been some early customer preview, but that… And I liked the term that he used, dynamic guardrailing. And so this model validation capability allows this constant fine-tuning and flexibility to react to new threats. And AI is going to be used within this tool as well, so you got to figure it’s going to become smarter and more intelligent as well, so.
Matt Kimball: Yeah, I love it. Great story.
Will Townsend: Thanks.
Matt Kimball: Good. Man, we’re going like boom, boom. Paul, did you have any thoughts before we move on to you and your-
Paul Smith-Goodson: No, I just… Automated red teaming is going to cut down the time and effort required pretty significantly, I think.
Will Townsend: Yeah.
Matt Kimball: Oh, yeah, yeah. And I’m guessing it documents everything in real time automatically as well, so yeah, that’s pretty cool. Very cool. All right, moving on to the brains of the operation, Paul Smith-Goodson talking about compute equaling quadrillions X. Okay? All right.
Paul Smith-Goodson: Yeah. No, I wanted to revisit a couple issues on the Google Willow chip. There’s been some confusion about that. Not that it’s wrong, it’s just confusing. So one thing I wanted to talk about was the critical quantum error correction threshold and then also talk about the benchmarking that was done and try and make some sense out of that. But the critical quantum error correction threshold is basically your error correction has got to be able to correct more errors than is happening. Obviously, if you’ve got… I mean, this is not the right metric, but if you can correct five errors every minute and your system is generating six errors every minute, you’re not going to get ahead of the game. So Willow is able to do this. So you absolutely need that, obviously, for or a fault-free machine. But anyway, that’s one thing I wanted to clear up. There has been some confusion about that.
The other thing is when they ran the random circuit sampling and then they compared it to the random circuit sampling that was done back when they ran it on the Sycamore back a couple of years ago, it kind of didn’t make sense. Random circuit sampling is a way to measure the quantum computer’s ability to perform things that are believed to be intractable for a classical supercomputer, so it’s a good metric to see how much progress is being made. It’s basically the benchmark is running a quantum circuit and they’ve got randomly chosen gates and then they measure the output, and then as a number of qubits grow, then the complexity of that circuit grows exponentially, which makes sense. In 2019, the Sycamore processor had 53 qubits and it ran that benchmark in 200 seconds, which was theoretically, it would take a classical supercomputer 10,000 years to complete.
Matt Kimball: Geez.
Will Townsend: Wow.
Paul Smith-Goodson: Now, when they ran the RCS on the Willow chip, now the Willow chip’s got 105 qubits in it.
Matt Kimball: So much more, yeah.
Paul Smith-Goodson: The Sycamore had 54, but they only use 53 because there was a fault with one of them. And every time you add a qubit to a quantum machine, it doubles the power, it’s exponential. In other words, you have one qubit and you add another qubit. It’s not twice as powerful, it’s four times as powerful. So if you look at the difference between 105 qubits and 53 qubits, quite a bit of difference in it.
Matt Kimball: Oh, yeah.
Paul Smith-Goodson: So anyway, the benchmark was that it took five minutes for Sycamore to run it and it took 200 seconds for the Willow to run the benchmark. And like I said, the supercomputer was calculated it would need 10,000 years, but with the Willow, it would need 10 septillion years to run the same benchmark.
Matt Kimball: My goodness.
Paul Smith-Goodson: And it’s kind of confusing, but it’s because of the difference, if you subtract 53 qubits from 105 qubits.
Matt Kimball: And it’s double that, double that, double that. It’s an exponential kind of.
Paul Smith-Goodson: Yeah, it’s like 2 to the 53rd power. So it’s actually 4.8 quadrillion times more powerful than what the Sycamore was. And that complexity is what made it, extended the time it would take us a supercomputer to run it. So that was basically-
Matt Kimball: Hey, Paul-
Paul Smith-Goodson: … that’s all I wanted to clear up because it didn’t make sense if you just said one of them took 200 seconds and the other one took 10 seconds to do it. You know?
Matt Kimball: Yeah. So Paul, with Sycamore and Willow, these are Google projects, right, Google chips?
Paul Smith-Goodson: Yeah.
Matt Kimball: If you look at the arms race of quantum, where do you place a Google versus a IBM versus a… I’m sure Microsoft is working on something. I mean, how do you rank these? And I don’t mean one’s better, I just mean, you know?
Paul Smith-Goodson: Yeah. Well, I mean, there’s different benchmarks that they can run to evaluate one against the other, and each of these systems have got their own advantages and disadvantages, so I wouldn’t want to compare. The Google chips got 105 qubits, IBM has got a chip coming out with over 1,000 qubits, but the architecture’s different.
Matt Kimball: Yeah, that is a fantastically political answer.
Will Townsend: And I didn’t excel in math and so I lost Paul after about the second or third word there.
Paul Smith-Goodson: Well, more forced time.
Will Townsend: But I think it sounds badass.
Paul Smith-Goodson: You kind of get a feel for that. I mean, that’s how much Musk is worth, I think.
Matt Kimball: It’s just incredible to think about five years from now, 10 years from now, what the compute landscape looks like as quantum goes from concept to nascent technology that’s starting to get used. And I mean, a 10,000-year benchmark being performed in 200 seconds is just incredible.
Will Townsend: It’s mind-blowing, yeah.
Paul Smith-Goodson: Yeah. I mean, it depends on the error rates, the coherence time of how long that quantum state’s going to last.
Matt Kimball: Sure.
Paul Smith-Goodson: The quality of the quantum operations, the error correction. You know?
Matt Kimball: Okay. So let’s be general, let’s be stingy and say 5,000 years instead of 10,000 years performed in 200 seconds. It’s still beyond the realm of logical grasping of what this can do and what it will do and how it’ll impact the way we live our lives. It’s incredible. Incredible.
Paul Smith-Goodson: That’s even fun to say. 4.5 quadrillion years, that’s fun.
Matt Kimball: Some days it feels like I’ve been living that long, so for what it’s worth.
Paul Smith-Goodson: Yeah.
Will Townsend: Amen, brother.
Paul Smith-Goodson: So anyway, I just wanted to clear that up.
Will Townsend: Cool.
Matt Kimball: Hey, that’s fantastic. These are always fun discussions and we’re going to go from really smart stuff to maybe not as smart, but quite interesting. Nutanix and Dell get cozy, I think this one is me, if I remember correctly. That was a joke. All right. So here’s let me give you a quick setup and then a quick what’s going on. So a couple of years ago, as we all know, Broadcom bought VMware, and I’m not here to say it was a good move, a bad move, but it was a disruptive move and it has caused virtually every enterprise organization to rethink their virtualization strategy. And there’s-
Will Townsend: I actually wrote an analyst insight on that very point, Matt, last year, so another gratuitous plug for my writing.
Matt Kimball: Yeah. Yeah, it is pretty gratuitous. No, I’m kidding. No, and it’s a good write-up and we have quite a few up on Moor Insights that cover the market dynamics, but there are three kind of outcomes for enterprise IT organizations. One is you stick with VMware, you go VCF 9, and you’re all in on VMware’s private cloud strategy, which is a good and legitimate strategy. The second is you look for an alternative because maybe you’re not as big or you’re not going to consume as much of the capabilities of VCF. So you go to an alternative HCI cloud-like solution, Nutanix, Red Hat, HPE has their own stacks, Citrix, Hyper-V, obviously, from Microsoft. And then the third is you say, “You know what? We’re moving away from virtualization entirely. We’re going to cloud-native,” still a cloud operating model, but maybe it’s a slightly underlying technology, technological difference. You can still use Nutanix, can still use Red Hat, but just a little bit different.
I say all that because in the wake of all of this disruption, Nutanix, of all the alternatives out there, has probably been, not probably, it has been the, I would say, the best mover in terms of offering an alternative without being overly aggressive and overly in your face. They build tools, they build channel relationships, they build partnerships to go to market and make it easy to decide, figure your costs, do the migration, make the move, and then live happily ever after. And they’ve always had strong relationships with Lenovo and with HPE, but they’ve had somewhat of a, and strained is maybe the wrong word, but I’m going to use strained relationship with Dell because of Dell’s entanglement with VMware for many years before Broadcom came in. It pushed Nutanix to the back and Dell helped VMware sell more. VMware helped Dell sell more.
Guess what? Now that the breakup happened between Dell and VMware, Nutanix and Dell have renewed their relationship and have gone to great lengths to build out these very bespoke platforms. It’s called the Dell XC Plus, and I’ve written a paper on this on Moor Insights. That’s my gratuitous plug. I wrote a research paper on it that goes in-depth as to what this is, but it’s a bespoke cloud platform that will drive everything from traditional legacy workloads to cloud-native to even AI. They have something called Gen AI-in-a-Box.
So really kind of interesting platform. It’s going to be fun to watch where this goes because Nutanix is a very strong channel presence. Dell has a strong enterprise presence. Dell can leverage Nutanix up market or down market and Nutanix can leverage Dell up market. So it’s going to be fun to watch, but it’s been an interesting ride, and man, this is a really cool platform they’ve put together. And of course, there’s the Intel derivative and the AMD derivative, or I should say offering, but really cool stuff. Anyway, I thought it was worth talking about because it’s a… Literally two years on, people are still talking about the VMware disruption. It’s been that big of a move. So it’s fun to see what’s happening and how companies are responding to all of this. And that’s been-
Will Townsend: And Matt, yeah, question for you. From your perspective is what’s driving this, it’s cost containment, right? So how much more aggressively positioned is Nutanix with these partnerships with Dell and others relative to the VMware solution? I’m assuming the OpEx profile is super aggressive, right?
Matt Kimball: Price-wise, Nutanix is very attractive relative to VMware. That’s part one. Part two, when you go with VCF, you’re building out a private cloud. Now, you can have that private cloud reside on-prem or you can have it reside on AWS or Azure, but it’s your private cloud. I mean, it is a… Think of it as sovereign to your company, right?
Will Townsend: Mm-hmm.
Matt Kimball: And the connective tissue to cloud services, even though it’s residing in AWS, it’s your siloed environment and connectivity to other services is limited. Nutanix is far more open. And I say it’s a cost-effective, more cost-effective. It really depends on how much you leverage the technology. If you leverage all of VCF, all of its capabilities-
Will Townsend: Then it’s probably a great value, right? Yeah.
Matt Kimball: It’s a great value. But you have to consume all of its capabilities and you have to be of a certain scale to really see those kind of TCO benefits. So I think there’s always been talk about, “VMware is going to go after the top 1,000, top 2,000 and just focus on them.” I don’t know that that’s the case, but it’s certainly is a solution that is more attractive. The larger you are, the more attractive its solution is.
Will Townsend: Okay, that makes sense. Yeah.
Paul Smith-Goodson: I got a question. If an enterprise customer is moving from VMware to Nutanix, isn’t that quite a bit of change and change management, retraining, and maybe even re-architect?
Matt Kimball: Yeah, that’s a good question, Paul. So it’s a continuum, right?And this is what I think has held a lot of… That inertia or gravity, if you will, has kept a lot of very large customers running on VMware who might otherwise consider alternatives because VMware, over the last almost 20 years or 20 years or so, has become so deeply embedded in the enterprise. It’s not just I’m running VMs on VMware, it has become the control plane for my data center. Compute, networking, storage, workloads that run within my cloud environment, it is all, everything is built around VMware and that is product, that’s people, that’s processes. So simply saying, “Well, I’m not happy with VMware so I’m going to go to Nutanix or Red Hat or Hyper-V,” it’s nice to say that, but it’s not an easy move. I mean, I’ve seen estimates where a company say it’s going to take them a full five years to migrate off of VMware onto an alternative platform. That is crazy, right?
Will Townsend: Yeah.
Matt Kimball: So that’s, to what you said, that’s the stickiness that VMware has within these organizations and Broadcom will play that well. I expect at some point you’re going to see a copilot or an assistant that enables. In fact, AWS has a, I forget what they call it, but they have their version of Copilot that helps, that will look at your existing VMware environment and tell you how you take that and you remove your dependencies and move everything up into AWS. I expect to see similar kind of copilots come out or AI assistants to come out for Hyper-V, for Red Hat, for VMware or for Nutanix. It only makes sense, right? Because otherwise, it is, yeah, it’s a bloody mess. Yeah.
Will Townsend: All right. Well, look at this.
Matt Kimball: I know.
Will Townsend: We got three topics in 23 minutes.
Matt Kimball: 23 minutes, man.
Will Townsend: Crazy.
Matt Kimball: I know. Oh, by the way, we should have said this at the beginning. This is episode 37, which is really cool. Again, a prime number. I’m a huge fan of prime numbers. Episode 38 will be twice as long because we’re going to cover twice as many topics. But anyway, before we get going, any last thoughts on what you’re seeing out there, guys? We have about two or three minutes to play around with if you want to throw any hot topics.
Will Townsend: Yeah, I’ll throw something out. So news broke around SolarWinds and a private equity company taking it private, and I shared some initial thoughts on my X handle, on my LinkedIn handle. Not surprised to see that. From my perspective, the company did its apology tour but it never really did anything beyond that to communicate what it was doing to basically address that issue and prevent future occurrences. And so yeah, that’s a thought I thought earlier.
Matt Kimball: Man, how the mighty have fallen, huh?
Will Townsend: What’s that?
Matt Kimball: How the mighty have fallen.
Will Townsend: I know. I tell you what, boy, they were… Back in the day, man, I would drive down Capital of Texas Highway in Austin and see their logo on the building, just with VMware as well, and things have changed. So I think they can come out of it. I think going private’s going to allow them to really do some introspection, figure out what they need to address. And I think a lot of this was just how they message this to the market and I had spent some time with them following that big supply chain hack and I just didn’t get a sense that it was sort of… I don’t think they were in denial, but I just got a sense that they weren’t recognizing how dramatic it was and it was just business as usual and they were reaching out to me. It’s like, “Hey, we’ve got a great observability solution.” And then you’ve got other companies like Dynatrace and Datadog and Gigamon and Cisco that are just full pedal to the metal. So yeah, I mean, I wouldn’t count them out, but it’s going to require a lot for them to rebuild their go-to market.
Matt Kimball: Yeah, yeah. It felt like, to what you’re saying, it felt like they were in an identity crisis even before that happened, like who are we and what do we do? Yeah, it’ll be interesting to see how that plays out. Any last thoughts from you, Paul?
Paul Smith-Goodson: This OpenAI came out with a Deep Research which will produce analyst-grade reports.
Will Townsend: Uh-oh!
Paul Smith-Goodson: What would take us three or four days, yeah, what would take us three or four days to do can be done in an hour. It’s going to cost you money to do it.
Matt Kimball: What takes you three or four days takes me three or four weeks, so it’s even better for my..All right, listen, guys, we’re up against the clock here. I know that we have some tight schedules. Everybody’s got things they have to fly to. Thank you for taking this shortened approach. This was a lot of fun actually. I liked the pithiness and y’all have a great weekend. And before we close out, just a reminder, hit the like button, subscribe. If you don’t like us, get on Twitter or X and complain about Will Townsend’s stuff and do not take anything we say as investment advice. And other than that, have a great weekend and we’ll see you in two weeks.
Will Townsend: TGIF.
Matt Kimball: Right on. Thanks, guys.
Paul Smith-Goodson: Yeah.
Will Townsend: Yeah. Thanks, Matt.
