HPE Discover 2024 | theCUBE Collective Analysis

Clips
News
More from HPE Discover 2024

Zeus Kerravala

Founder & Principal Analyst

ZK Research

Three insights you might have missed from theCUBE’s coverage of HPE Discover

Artificial intelligence took center stage at HPE Discover last week. For three days, company executives captivated audiences with keynote sessions focused on Hewlett Packard Enterprise Co.’s cutting-edge AI solutions and the robust infrastructure it’s building to support them.New solutions for AI deployment and how HPE’s enterprise customers can take advantage of them were discussed at length by theCUBE, SiliconANGLE Media’s livestreaming studio, through interviews with company executives, clients and industry analysts at HPE’s signature event

AI’s uncertain path: Companies weigh strategy and implementation in 2024

When taking stock of the artificial intelligence revolution, so much has changed even since the beginning of 2024. Despite that, companies still need to fully grasp the implications of the technology, as well as how to approach AI investment.It’s not quite the Gartner Trough of Disillusionment, according to Zeus Kerravala (pictured), founder and principal analyst at ZK Research. But companies have kicked the tires, and the next steps are far from certain.ZK Research’s Zeus Kerravala talks with theCUBE’s Dave Vellante at HPE Discover.“Every CIO tells me, their bosses are screaming, ‘We need an AI strategy,’ so they’re investing in the technology,” Kerravala said

play_circle_outline AI Dominates HPE Discover 2024 Amid Worries of Industry Oversaturation with Nvidia Partnerships

play_circle_outline Comparison of HPE's AI stack with VCE for private cloud

play_circle_outline Exploring Challenges & Benefits of AI Adoption Amidst Rapid Growth in Cloud Market

play_circle_outline Analyzing Duopoly Influence on HPE's Market Position in Network-Centric IT Shift

play_circle_outline Navigating SaaS Challenges in AI Wave: Nvidia's Influence, Broadening Impact, and HPE's Growth Potential

Info

theCUBE Collective Analysis

Zeus Kerravala

Founder & Principal Analyst ZK Research

search

>> Welcome back to HPE Discover 2024 from Las Vegas. We're live. My name is Dave Vellante. Zeus Kerravala is here. He is the principal at ZK Research, member of theCUBE Collective, friend of theCUBE, many time guest contributed to siliconangle.com. Good to see you, my friend. How you doing?

Zeus Kerravala

>> Great. It's been a great week.

>> How'd you hit them today? You were golfing this morning with some CIOs.

Zeus Kerravala

>> Yeah. Yeah. Shot 75, couldn't be happier about that. CIOs played well as well. So when the CIOs leave happy, I'm happy.

>> That's fantastic.

Zeus Kerravala

>> Got a little mini CCIO summit.

>> Well, thanks for making some time to come back inside this hall. We're winding up, as you know, the day three, I guess for us, day four of the conference. To start, what were your big takeaways?

Zeus Kerravala

>> Well, obviously AI is the big takeaway here, and HPE jumped on that everyone else has. I'm a little concerned though that Nvidia today is the cool kid in high school and everybody wants to be their friend. And I'm finding that across all the shows that you and I have even been to this year, everyone's got this Nvidia relationship, and they all kind of look the same. And that's a concern for the industry, because with the lack of differentiation, then everybody winds up just being a channel for Nvidia. And I asked several HPE people, including Antonio in our analyst Q&A, and he focused a lot on the services and support and sales. And while I think that's notable, I do think HPE could have focused, he could have focused his answer more on some of the technology like ops ramp and the networking that HPE has, especially with Juniper coming. And so it'd be good to see them be able to differentiate themselves from a technology perspective more so than service and sales.

>> Well, so a couple of differentiators that I'd point out, and then to get your reaction on. John Furrier has been tracking how long Jensen has stayed on stage for each of these events. HPE was the longest, notwithstanding GTC, like 25 minutes. Mostly, he's 10 to 15 minutes

Zeus Kerravala

>> Was it because we're in The Sphere and he wanted to hang out in there more?

>> I think so. I think probably the first keynote ever in The Sphere. Antonio evidently mentioned that at the HPE studio, but of course he didn't attribute Silicon Angle and John Furrier, so we'll just make that on the record here.

Zeus Kerravala

>> Okay.

>> And we'd like attribution going forward. So that's funny, tongue in cheek. Okay. The other piece of it is the messaging that this is not a reference architecture, that this is a complete system.

Zeus Kerravala

>> Yes.

>> And we will see how much of a difference that makes. It should make a difference. If in fact the competitors are only doing reference architectures and not full systems, that's going to be a differentiator. Having said that, Jensen's always emphasizing that we're a systems company, we build systems, and then we break them apart, and then we allow you to put them together. When you do put them back together, we've done all this systems work. So what are your thoughts on that? How much of a differentiator do you think it will be for HPE?

Zeus Kerravala

>> Yeah, if it really is as simple as they say three clicks and you're up and running, then I think that is a differentiator. I don't know if any of the other OEMs have a three click roll it up and turn it on system. So certainly not Dell. Cisco's still building theirs. So I do think that here, when you think about what the long-term differentiation of this industry could be, it will wind up being ease of use, I think in a lot of cases. And I think for a lot of enterprises, building your own AI stack certainly isn't going to be easy and the things are more complicated. The risk reward here is pretty high for enterprises too, which makes ease of use more important because everybody wants AI. Most companies I talked to aren't sure how to deploy it. They all tell me complexities a lot higher than it was even two years ago. And so if there's a way to minimize the complexity by getting a turnkey system... It reminds me a lot of the early days of private cloud when VMware, Cisco and EMC had the VCE stack.

>> Right.

Zeus Kerravala

>> That was incredibly popular because it was an engineered system. And then eventually we got good at best practice and we didn't need them anymore. And so when you look at HP's AI stack with Nvidia, I think we could almost look at it as the AI equivalent to what VCE was to private cloud.

>> So the trade-off is granularity because they did the T-shirt sizes, the small, medium, large and extra large. And based on what you're saying, I think that's a good trade-off. I would rather lean towards simplicity, like Snowflake and database, than have infinite granularity and knobs to turn. I think that the former is going to play better, at least initially in this AI land grab.

Zeus Kerravala

>> Oh yeah, 100% I agree with you on that. If my choice as a CIO is take my fully integrated box, three clicks, turn it up, maybe I give up some optionality, or go spend months, maybe years tweaking and tuning, turning knobs and levers, that's a no brainer. Especially now, time to market's so important here.

>> And this is where the converged infrastructure... You're bringing up an interesting point about converged infrastructure. David Flynn, who was the CEO of Fusion IO at the time said something that I thought was pretty prescient at the time, now it's pretty obvious, is that the hyperscalers, they'll spend engineering time. They throw engineers at it-

Zeus Kerravala

>> Oh yeah....

>> to save money. Enterprises will spend money to save time because they don't have the resources. And it's true. If they can find a three click solution and it maybe costs more, or maybe it doesn't even cost more, they'll go for that every time.

Zeus Kerravala

>> Yeah. Well, why do you think White Box never really took off an enterprise? I remember talking to one of the telecoms even about they actually looked a white box, writing the code, supporting it themselves, and they said the TCO was 5x than what it was to buy just a Cisco switch. Now that's just one switch, right? So you think about doing that with a full AI stack... I think today, Dave, as early as we are with as few best practices we are and as immature it is in the channel, I don't really understand how it would make sense to try and build it yourself, unless you really feel like you have the engineering chops to differentiate. But then your IT costs are out of control at that time, but that's going to be okay for some companies, but it's a very small percent of companies.

>> Now, I have never attended an Atmosphere event. I thought it was really smart that they put the two events together this year. There about 15,000 people here, which is-

Zeus Kerravala

>> Yeah, and about a third were Atmosphere, I was told. So that was a pretty impressive number.

>> So smart move. I actually think it's about time that they made that move, but it was a very well attended event. It's quieting down here on the getaway day, but day one and day two were packed. What was Atmosphere like? What were the key messages? Any discussion on the Juniper acquisition and how that all fits in?

Zeus Kerravala

>> Well, certainly the Airheads audience is very big. They're very passionate about the HP products. In fact, I thought it was funny during Antonio's keynote when he mentioned Airheads, and then he was almost defensive about it saying, "No, no, they like to be called that just to make sure that the rest of the HPE audience knew that that was the groupthink."

>> He wasn't a pejorative. It was like alpha geek. That's a compliment.

Zeus Kerravala

>> Yeah. And they really couldn't say a lot about Juniper. I think the key message to HPE customers is, look, there's no concern about buying Aruba products today. And Phil Mantra was very, very clear about this. Whenever we make a product decision, we give our customers years of notice. And I've talked to some customers that are a little concerned about continue to invest in central. It looks like mist is coming in, and that might wind up being the roadmap for the future. And Phil wanted to make sure everybody... If that's the case, it's going to be years from now when that happens. And in fact, I think over time we will see is you'll have missed in central. They'll eventually converge years out, and everyone will wind up in the same place. You just won't know that it's happening because they're taking different roadmaps.

>> And that's the right strategy. Very clearly, it's like give customers plenty of time.

Zeus Kerravala

>> It costs the vendor more to do that.

>> Yes, but they got to do the engineering is the point I was going to make.

Zeus Kerravala

>> Right, but it's better for the customer. And so I think in that case, it was a good message for the Aruba audience to calm them down and let them know you can continue investing in us and you are not going to be put at risk. And that seemed to make a lot of them happy. And I do think they like the integration to rest of HPE. I think it's been a long time coming for that, and the whole edge to cloud story does involve more than just networking.

>> But I would like to see a statement of direction bringing this and central together and have that engineering work start. Because as we know, if you get multiple single panes of glass, it causes other problems for customers.

Zeus Kerravala

>> Yes. And I think they will do that once the deal closes. The rubs are going to come, Dave, from a sales perspective. If I'm an HPE salesperson and it's greenfield, do I sell Mist or do I sell Aruba? And in a lot of cases, I think it's going to come down to if I'm a legacy Juniper salesperson, I'm going to have an affinity to Mist, and if I'm a legacy Aruba salesperson, I'm going to have an affinity to that product. But that, to me, is I think where a lot of the work would need to be done to educate the salesforce where you sell, what, and why?

>> What's the right strategic fit.

Zeus Kerravala

>> And I do think there's some obvious ones like HPE is very big in gaming and hospitality, right? They're part of hotels, they're part of the Golden State Warriors, Tottenham Hotspur, a lot of... Where Juniper doesn't really have much of a stadium footprint. Juniper is very big in a lot of universities. And so you could see that there's some natural synergies between the two, but there's obviously going to be a lot of overlap as well.

>> Okay, so we've been perpetually on the road for the last six months, it feels, right? This is the last day-

Zeus Kerravala

>> See more of you than my wife.

>> It's like, don't worry, we are going to be home for a while.

Zeus Kerravala

>> Yeah.

>> But so you're thinking about... I think this is the last day of the spring, right? This is the first day of summer, last day of spring.

Zeus Kerravala

>> Oh yeah.

>> It's getting closer.

Zeus Kerravala

>> Happy summer.

>> Summer solstice is today, right?

Zeus Kerravala

>> Yeah.

>> Okay. So wow, looking back six months, I don't even want to go back a year. Sometimes we go back to November '22 when ChatGPT was announced. Forget that. So much has happened since then. But if you go back from the beginning of this year to where we are now, how would you describe the AI sort of progression? And where do you think we're going from here over the next six months?

Zeus Kerravala

>> I think customers now... I don't want... If you used the Gartner Profit Disillusionment, I don't think we're quite there, but I do think customers have now kicked the tires enough, and they're still not sure what to do with it is the problem. And I'm a little worried that we're going to have a situation of... Earlier this year or last year, Cisco had claim digestion issues, customers overbought.

>> Yep.

Zeus Kerravala

>> And I do wonder if we're going to see that here where customers were so thirsty and they're getting every CIO tells me their bosses are screaming we need an AI strategy, so they're investing in the technology, but are we going to wind up in a situation where, without a lack of a roadmap of what to do with AI, you bring the technology and you spend the money, and then it looks like a bit of a failure because you didn't know how to implement it property, or at least what problems it is you're trying to solve, right? And I think there's a real risk that could happen. And so for the OEMs and the vendors, it's very important that they help their customers understand what the right use cases are and what the right use cases aren't, because you don't want customers investing all this money. This is not cheap stuff. And CTOs are now telling me for the first time in a decade, they're buying more data center space.

>> Yep.

Zeus Kerravala

>> Right? So you've got a lot of conflicting forces in trying to save money in trying to be more sustainable. But I'm buying more AI and I'm buying more data center space, and something's got to give. And I'm concerned we're going to see that in the back half of the year.

>> And the data center builders are saying, "I don't have enough power."

Zeus Kerravala

>> Yeah.

>> And that is a real legit problem and I got to figure it out and bring it in. Charlie Kawwas at the financial analyst meeting earlier this year, Broadcom, I thought laid out very succinctly. He said, "Look, the ROI in consumer for AI, it's a no brainer." They're trying to build... We know they're working toward building million GPU clusters. They're nowhere close to that today, but the bigger the GPU cluster they can build, the more ads they're going to sell, the better targeting they're going to have. They're going to compress the time to value. No

Zeus Kerravala

>> And they can measure the value very easily.

>> Yes, and it's a no brainer. And these hyperscalers, the internet giants, they're awash with cash. It's on their balance sheet. They might as well invest. It's a huge arms race. Okay, and that's what's been powering all this momentum. Nvidia earlier was up today again. But I think to answer the point that you just brought up, the question becomes is Jensen's law, does that hold true? I.e., if you buy more-

Zeus Kerravala

>> The more you buy, the more you save?

>> The more you buy, the more you save. If that's true, then this thing has some more legs. Certainly will be the case with the big guys doing training who have a lot of cash. Whether or not that trickles into the enterprise, I think it's clear. It has not trickled into the enterprise quite yet. It was interesting. We just had Deloitte on, and I sort of was laying out that scenario and said, "Look, we actually see it differently. We have big customers coming to us in things like financial services, and they're going hard after this." I don't think most customers are doing that. Most customers are experimenting, hitting singles, doing chatty like applications, being careful about where they allocate the budget, stealing from other budgets. Now, whether or not that shifts and changes in the second half remains to be seen. I think the summer, probably not. Maybe Q4, we start seeing some of those bigger. And if we don't, I think to your point, things are going to get tighter.

Zeus Kerravala

>> And I think part of this is, like I said, an uncertainty of what to do with it and when to roll it out. I was at a conference earlier this year where I was talking with a bunch of contact center managers about AI in the contact center, which is one of the low hanging fruit use cases. And one of them was saying the large language models have been evolving so fast that you think you've finished testing and you're ready to roll it out, and boom, a new one rolls in, and then you rerun all the models, or you put all your data in with the new models and you go through that whole process again. And then he says, "I'm never sure when to give it the go because it's changing so fast." And he said, "I'm perpetually in the state of almost ready to go, but I can't ever seem to push it over the finish line." And I hear that over and over and over. And so there's some obvious use cases. It is... I think contact center is a good place for it. I think as an AI ops tool, but... And I do worry sometimes that IT pros wait for perfection. Oh, it made a couple of errors. It answered 18 of my 20 questions correct. Well, that's still better than people. And as long as your threshold's better than people, I think you're a winner. But I do think there's a lot of trepidation about getting it wrong, and that's where the hesitation comes to play. If I put it in and I... You remember the early days of chatbots and things, that was a disaster for a lot of companies.

>> Yeah.

Zeus Kerravala

>> And nobody wants a repeat of that because you're going to drive customers away if you deploy it wrong. And then a lot of people are going to be losing their jobs if that happens, right?

>> And then the legal and governance and compliance issues, people are rightly concerned.

Zeus Kerravala

>> The question is there's no AI better than bad AI.

>> That's not a clear cut answer.

Zeus Kerravala

>> Yeah.

>> It is one of those consultant it depends things.

Zeus Kerravala

>> And who's got a good handle on their data? How many enterprises really have a good handle on their data?

>> So we know broadly that the answer is few and far between, but if you have isolated data sets, then potentially you can have a good handle on data, but then the question becomes, well, what's the real value? What's the actual size of that benefit that you're getting? You might get an ROI of 200%. So what if it's five bucks? It doesn't matter, right? And so that's not going to fund future AI investments. So some of those bigger MPVs have to start hitting in the second half, or I think people will tighten up. You will hit that trough of disillusionment. I don't think we're there yet, I agree. But part of that is because the hyperscalers are putting in so much funding to this thing, and that's like a gift to all of us. We're like, yeah, keep spending. Yeah, Blackwell's coming great. Spend more on that. HPE, same thing.

Zeus Kerravala

>> Yeah. I think a decade from now, AI is going to be built into everything we do. I look at it like the early days of the internet, where as an IT pro, I used to go to internet conferences. There are no internet conferences anymore because I don't need the value to be proven to me.

>> Right.

Zeus Kerravala

>> There are AI conferences today, and a decade from now, we won't have them anymore. Beginning from here to there, even in the early days of the internet was very problematic. What do you use it for? Should you take credit cards there? There's a lot of ups and downs, and we're going to go through that here. The long-term benefit, it's a no-brainer, but how you fund it, where you apply it, how you measure the ROI, I think we're going to see a lot of change in leadership in the S&P and things like that based on it like we did with the internet and cloud. These transitions always create a big upheaval in the business world. And so I think long-term trends are positive, but expect some bumpiness along the way.

>> Well, and the prevailing law determines those winners and losers. It was Moore's Law, it was Metcalfe's law in the internet phase, and that I would argue that really actually powered a lot of the cloud, the network effects, and certainly social media, big data clouds, social, and mobile. That was Metcalfe's law, was really the prevailing force at the time. And we will see if Jensen's Law holds true. He's been right a lot.

Zeus Kerravala

>> Yeah. But I think from an HPE perspective, they have such a big footprint with a lot of leading companies, they should be in just a prime position to be able to help take a lot of the early adopter customers that are using AI and be able to replicate that across their broad customer base. Their customer base is literally who's who of the business world.

>> Well, it's interesting, these infrastructure companies, it's almost like Dell and HPE have kind of a duopoly for on-prem. Certainly IBM's in there, but they're not really an infrastructure player.

Zeus Kerravala

>> I think the advantage HPE has, and I'm hoping Antonio meant this when he said it when they acquired Juniper and they had the Rami Antonio analyst Q&A, and he said, "We want to be known as a network first company." Right? And I think in fact, Sanjeev Katwa, the CIO of Tottenham Hotspur, I went to the session yesterday that he was in, and he said, "The network is the most important asset we have today, because without it, we can't do anything. We can't process tickets, we can't sell people stuff. We have no 911, we've got no," or whatever their version is over there. And so I think there is a pivot going on within corporate IT to be in a world that's more network centric. I think HPE, being a traditional compute company, is in a good position to make that change more so than even Cisco, because Cisco, it'll seem somewhat self-serving. Dell doesn't really have any network assets, and so-

>> They're taking a different strategy, building the networking into the service.

Zeus Kerravala

>> I'm really hoping that that message continues to carry through to next year's Discover and across other HPE activities, because I do think with the Juniper acquisition, they have just a very, very good portfolio of network products now, and they can actually do end-to-end networking today.

>> Right. And that's where this duopoly thing is interesting, but it's diverging, because like you say, HPE becoming a networking company, in many respects, Dell is much more storage heavy than say HPE is, so they've got some potential advantages there. And then you've got the cloud guys who are still growing at 20% a year,-

Zeus Kerravala

>> Yeah, it's crazy....

>> which is collectively 200 billion. And so you hear a lot about repatriation, but the numbers still aren't propelling

Zeus Kerravala

>> Well, you can repatriate and grow, right?

>> Absolutely. They're not mutually... They both can be true. We know anecdotally, repatriation is happening. It's just pales in comparison to the growth that you're seeing in the cloud. But it's like right now, it's good to be in tech, except SaaS is showing some cracks in the armor. So we'll see how that plays out, because they haven't really benefited from the AI wave. Not withstanding ServiceNow is obviously-

Zeus Kerravala

>> Actually, what's fascinating to me, Dave, is if you look at the capital markets, Nvidia has created this massive tailwind, but it's lifted very few boats. Arista has gotten a big AI bump. Pure Storage has. How many other infrastructure vendors have?

>> Well, Dell has. Dell's cranking this year. Super Micro has, and most recently HPE. HPE is up about, I want to say 22% in the last 30 days. So HPE is kind of... Their momentum's catching up. And I think HPE is being very conservative about their guidance. They beat and they raised. They're only... They're guiding to like, I don't know, one to 3% growth. I think they're going to grow fast. I think their AI server backlog is much bigger. And I think coming out of this show, they're going to beat again. But they're smartly, I think being conservative. So you're seeing a little bit of broadening beyond the Nvidias, and the Qualcomms, and the Broadcoms, and the TSMs to some of the infrastructure plays, but it certainly hasn't hit the software stack yet for the Mongos, Snowflakes kind of bouncing around right now. Databricks is private. I think they're growing actually very fast, I think for other reasons. Salesforce got hit a little bit. Like I said, ServiceNow is killing it. So that is the one anomaly.

Zeus Kerravala

>> We are seen have and have nots, so that's-

>> Yeah, very much so. All right, Zeus, hey, awesome to see you.

Zeus Kerravala

>> Yeah, always a pleasure.

>> Thanks so much for making some time for us.

Zeus Kerravala

>> Yeah.

>> All right, keep it right there. We're wrapping up day three of HPE Discover, live from Las Vegas. You're watching theCUBE.

theCUBE Collective Analysis

search

>> Welcome back to HPE Discover 2024 from Las Vegas. We're live. My name is Dave Vellante. Zeus Kerravala is here. He is the principal at ZK Research, member of theCUBE Collective, friend of theCUBE, many time guest contributed to siliconangle.com. Good to see you, my friend. How you doing?

Zeus Kerravala

>> Great. It's been a great week.

>> How'd you hit them today? You were golfing this morning with some CIOs.

Zeus Kerravala

>> Yeah. Yeah. Shot 75, couldn't be happier about that. CIOs played well as well. So when the CIOs leave happy, I'm happy.

>> That's fantastic.

Zeus Kerravala

>> Got a little mini CCIO summit.

>> Well, thanks for making some time to come back inside this hall. We're winding up, as you know, the day three, I guess for us, day four of the conference. To start, what were your big takeaways?

Zeus Kerravala

>> Well, obviously AI is the big takeaway here, and HPE jumped on that everyone else has. I'm a little concerned though that Nvidia today is the cool kid in high school and everybody wants to be their friend. And I'm finding that across all the shows that you and I have even been to this year, everyone's got this Nvidia relationship, and they all kind of look the same. And that's a concern for the industry, because with the lack of differentiation, then everybody winds up just being a channel for Nvidia. And I asked several HPE people, including Antonio in our analyst Q&A, and he focused a lot on the services and support and sales. And while I think that's notable, I do think HPE could have focused, he could have focused his answer more on some of the technology like ops ramp and the networking that HPE has, especially with Juniper coming. And so it'd be good to see them be able to differentiate themselves from a technology perspective more so than service and sales.

>> Well, so a couple of differentiators that I'd point out, and then to get your reaction on. John Furrier has been tracking how long Jensen has stayed on stage for each of these events. HPE was the longest, notwithstanding GTC, like 25 minutes. Mostly, he's 10 to 15 minutes

Zeus Kerravala

>> Was it because we're in The Sphere and he wanted to hang out in there more?

>> I think so. I think probably the first keynote ever in The Sphere. Antonio evidently mentioned that at the HPE studio, but of course he didn't attribute Silicon Angle and John Furrier, so we'll just make that on the record here.

Zeus Kerravala

>> Okay.

>> And we'd like attribution going forward. So that's funny, tongue in cheek. Okay. The other piece of it is the messaging that this is not a reference architecture, that this is a complete system.

Zeus Kerravala

>> Yes.

>> And we will see how much of a difference that makes. It should make a difference. If in fact the competitors are only doing reference architectures and not full systems, that's going to be a differentiator. Having said that, Jensen's always emphasizing that we're a systems company, we build systems, and then we break them apart, and then we allow you to put them together. When you do put them back together, we've done all this systems work. So what are your thoughts on that? How much of a differentiator do you think it will be for HPE?

Zeus Kerravala

>> Yeah, if it really is as simple as they say three clicks and you're up and running, then I think that is a differentiator. I don't know if any of the other OEMs have a three click roll it up and turn it on system. So certainly not Dell. Cisco's still building theirs. So I do think that here, when you think about what the long-term differentiation of this industry could be, it will wind up being ease of use, I think in a lot of cases. And I think for a lot of enterprises, building your own AI stack certainly isn't going to be easy and the things are more complicated. The risk reward here is pretty high for enterprises too, which makes ease of use more important because everybody wants AI. Most companies I talked to aren't sure how to deploy it. They all tell me complexities a lot higher than it was even two years ago. And so if there's a way to minimize the complexity by getting a turnkey system... It reminds me a lot of the early days of private cloud when VMware, Cisco and EMC had the VCE stack.

>> Right.

Zeus Kerravala

>> That was incredibly popular because it was an engineered system. And then eventually we got good at best practice and we didn't need them anymore. And so when you look at HP's AI stack with Nvidia, I think we could almost look at it as the AI equivalent to what VCE was to private cloud.

>> So the trade-off is granularity because they did the T-shirt sizes, the small, medium, large and extra large. And based on what you're saying, I think that's a good trade-off. I would rather lean towards simplicity, like Snowflake and database, than have infinite granularity and knobs to turn. I think that the former is going to play better, at least initially in this AI land grab.

Zeus Kerravala

>> Oh yeah, 100% I agree with you on that. If my choice as a CIO is take my fully integrated box, three clicks, turn it up, maybe I give up some optionality, or go spend months, maybe years tweaking and tuning, turning knobs and levers, that's a no brainer. Especially now, time to market's so important here.

>> And this is where the converged infrastructure... You're bringing up an interesting point about converged infrastructure. David Flynn, who was the CEO of Fusion IO at the time said something that I thought was pretty prescient at the time, now it's pretty obvious, is that the hyperscalers, they'll spend engineering time. They throw engineers at it-

Zeus Kerravala

>> Oh yeah....

>> to save money. Enterprises will spend money to save time because they don't have the resources. And it's true. If they can find a three click solution and it maybe costs more, or maybe it doesn't even cost more, they'll go for that every time.

Zeus Kerravala

>> Yeah. Well, why do you think White Box never really took off an enterprise? I remember talking to one of the telecoms even about they actually looked a white box, writing the code, supporting it themselves, and they said the TCO was 5x than what it was to buy just a Cisco switch. Now that's just one switch, right? So you think about doing that with a full AI stack... I think today, Dave, as early as we are with as few best practices we are and as immature it is in the channel, I don't really understand how it would make sense to try and build it yourself, unless you really feel like you have the engineering chops to differentiate. But then your IT costs are out of control at that time, but that's going to be okay for some companies, but it's a very small percent of companies.

>> Now, I have never attended an Atmosphere event. I thought it was really smart that they put the two events together this year. There about 15,000 people here, which is-

Zeus Kerravala

>> Yeah, and about a third were Atmosphere, I was told. So that was a pretty impressive number.

>> So smart move. I actually think it's about time that they made that move, but it was a very well attended event. It's quieting down here on the getaway day, but day one and day two were packed. What was Atmosphere like? What were the key messages? Any discussion on the Juniper acquisition and how that all fits in?

Zeus Kerravala

>> Well, certainly the Airheads audience is very big. They're very passionate about the HP products. In fact, I thought it was funny during Antonio's keynote when he mentioned Airheads, and then he was almost defensive about it saying, "No, no, they like to be called that just to make sure that the rest of the HPE audience knew that that was the groupthink."

>> He wasn't a pejorative. It was like alpha geek. That's a compliment.

Zeus Kerravala

>> Yeah. And they really couldn't say a lot about Juniper. I think the key message to HPE customers is, look, there's no concern about buying Aruba products today. And Phil Mantra was very, very clear about this. Whenever we make a product decision, we give our customers years of notice. And I've talked to some customers that are a little concerned about continue to invest in central. It looks like mist is coming in, and that might wind up being the roadmap for the future. And Phil wanted to make sure everybody... If that's the case, it's going to be years from now when that happens. And in fact, I think over time we will see is you'll have missed in central. They'll eventually converge years out, and everyone will wind up in the same place. You just won't know that it's happening because they're taking different roadmaps.

>> And that's the right strategy. Very clearly, it's like give customers plenty of time.

Zeus Kerravala

>> It costs the vendor more to do that.

>> Yes, but they got to do the engineering is the point I was going to make.

Zeus Kerravala

>> Right, but it's better for the customer. And so I think in that case, it was a good message for the Aruba audience to calm them down and let them know you can continue investing in us and you are not going to be put at risk. And that seemed to make a lot of them happy. And I do think they like the integration to rest of HPE. I think it's been a long time coming for that, and the whole edge to cloud story does involve more than just networking.

>> But I would like to see a statement of direction bringing this and central together and have that engineering work start. Because as we know, if you get multiple single panes of glass, it causes other problems for customers.

Zeus Kerravala

>> Yes. And I think they will do that once the deal closes. The rubs are going to come, Dave, from a sales perspective. If I'm an HPE salesperson and it's greenfield, do I sell Mist or do I sell Aruba? And in a lot of cases, I think it's going to come down to if I'm a legacy Juniper salesperson, I'm going to have an affinity to Mist, and if I'm a legacy Aruba salesperson, I'm going to have an affinity to that product. But that, to me, is I think where a lot of the work would need to be done to educate the salesforce where you sell, what, and why?

>> What's the right strategic fit.

Zeus Kerravala

>> And I do think there's some obvious ones like HPE is very big in gaming and hospitality, right? They're part of hotels, they're part of the Golden State Warriors, Tottenham Hotspur, a lot of... Where Juniper doesn't really have much of a stadium footprint. Juniper is very big in a lot of universities. And so you could see that there's some natural synergies between the two, but there's obviously going to be a lot of overlap as well.

>> Okay, so we've been perpetually on the road for the last six months, it feels, right? This is the last day-

Zeus Kerravala

>> See more of you than my wife.

>> It's like, don't worry, we are going to be home for a while.

Zeus Kerravala

>> Yeah.

>> But so you're thinking about... I think this is the last day of the spring, right? This is the first day of summer, last day of spring.

Zeus Kerravala

>> Oh yeah.

>> It's getting closer.

Zeus Kerravala

>> Happy summer.

>> Summer solstice is today, right?

Zeus Kerravala

>> Yeah.

>> Okay. So wow, looking back six months, I don't even want to go back a year. Sometimes we go back to November '22 when ChatGPT was announced. Forget that. So much has happened since then. But if you go back from the beginning of this year to where we are now, how would you describe the AI sort of progression? And where do you think we're going from here over the next six months?

Zeus Kerravala

>> I think customers now... I don't want... If you used the Gartner Profit Disillusionment, I don't think we're quite there, but I do think customers have now kicked the tires enough, and they're still not sure what to do with it is the problem. And I'm a little worried that we're going to have a situation of... Earlier this year or last year, Cisco had claim digestion issues, customers overbought.

>> Yep.

Zeus Kerravala

>> And I do wonder if we're going to see that here where customers were so thirsty and they're getting every CIO tells me their bosses are screaming we need an AI strategy, so they're investing in the technology, but are we going to wind up in a situation where, without a lack of a roadmap of what to do with AI, you bring the technology and you spend the money, and then it looks like a bit of a failure because you didn't know how to implement it property, or at least what problems it is you're trying to solve, right? And I think there's a real risk that could happen. And so for the OEMs and the vendors, it's very important that they help their customers understand what the right use cases are and what the right use cases aren't, because you don't want customers investing all this money. This is not cheap stuff. And CTOs are now telling me for the first time in a decade, they're buying more data center space.

>> Yep.

Zeus Kerravala

>> Right? So you've got a lot of conflicting forces in trying to save money in trying to be more sustainable. But I'm buying more AI and I'm buying more data center space, and something's got to give. And I'm concerned we're going to see that in the back half of the year.

>> And the data center builders are saying, "I don't have enough power."

Zeus Kerravala

>> Yeah.

>> And that is a real legit problem and I got to figure it out and bring it in. Charlie Kawwas at the financial analyst meeting earlier this year, Broadcom, I thought laid out very succinctly. He said, "Look, the ROI in consumer for AI, it's a no brainer." They're trying to build... We know they're working toward building million GPU clusters. They're nowhere close to that today, but the bigger the GPU cluster they can build, the more ads they're going to sell, the better targeting they're going to have. They're going to compress the time to value. No

Zeus Kerravala

>> And they can measure the value very easily.

>> Yes, and it's a no brainer. And these hyperscalers, the internet giants, they're awash with cash. It's on their balance sheet. They might as well invest. It's a huge arms race. Okay, and that's what's been powering all this momentum. Nvidia earlier was up today again. But I think to answer the point that you just brought up, the question becomes is Jensen's law, does that hold true? I.e., if you buy more-

Zeus Kerravala

>> The more you buy, the more you save?

>> The more you buy, the more you save. If that's true, then this thing has some more legs. Certainly will be the case with the big guys doing training who have a lot of cash. Whether or not that trickles into the enterprise, I think it's clear. It has not trickled into the enterprise quite yet. It was interesting. We just had Deloitte on, and I sort of was laying out that scenario and said, "Look, we actually see it differently. We have big customers coming to us in things like financial services, and they're going hard after this." I don't think most customers are doing that. Most customers are experimenting, hitting singles, doing chatty like applications, being careful about where they allocate the budget, stealing from other budgets. Now, whether or not that shifts and changes in the second half remains to be seen. I think the summer, probably not. Maybe Q4, we start seeing some of those bigger. And if we don't, I think to your point, things are going to get tighter.

Zeus Kerravala

>> And I think part of this is, like I said, an uncertainty of what to do with it and when to roll it out. I was at a conference earlier this year where I was talking with a bunch of contact center managers about AI in the contact center, which is one of the low hanging fruit use cases. And one of them was saying the large language models have been evolving so fast that you think you've finished testing and you're ready to roll it out, and boom, a new one rolls in, and then you rerun all the models, or you put all your data in with the new models and you go through that whole process again. And then he says, "I'm never sure when to give it the go because it's changing so fast." And he said, "I'm perpetually in the state of almost ready to go, but I can't ever seem to push it over the finish line." And I hear that over and over and over. And so there's some obvious use cases. It is... I think contact center is a good place for it. I think as an AI ops tool, but... And I do worry sometimes that IT pros wait for perfection. Oh, it made a couple of errors. It answered 18 of my 20 questions correct. Well, that's still better than people. And as long as your threshold's better than people, I think you're a winner. But I do think there's a lot of trepidation about getting it wrong, and that's where the hesitation comes to play. If I put it in and I... You remember the early days of chatbots and things, that was a disaster for a lot of companies.

>> Yeah.

Zeus Kerravala

>> And nobody wants a repeat of that because you're going to drive customers away if you deploy it wrong. And then a lot of people are going to be losing their jobs if that happens, right?

>> And then the legal and governance and compliance issues, people are rightly concerned.

Zeus Kerravala

>> The question is there's no AI better than bad AI.

>> That's not a clear cut answer.

Zeus Kerravala

>> Yeah.

>> It is one of those consultant it depends things.

Zeus Kerravala

>> And who's got a good handle on their data? How many enterprises really have a good handle on their data?

>> So we know broadly that the answer is few and far between, but if you have isolated data sets, then potentially you can have a good handle on data, but then the question becomes, well, what's the real value? What's the actual size of that benefit that you're getting? You might get an ROI of 200%. So what if it's five bucks? It doesn't matter, right? And so that's not going to fund future AI investments. So some of those bigger MPVs have to start hitting in the second half, or I think people will tighten up. You will hit that trough of disillusionment. I don't think we're there yet, I agree. But part of that is because the hyperscalers are putting in so much funding to this thing, and that's like a gift to all of us. We're like, yeah, keep spending. Yeah, Blackwell's coming great. Spend more on that. HPE, same thing.

Zeus Kerravala

>> Yeah. I think a decade from now, AI is going to be built into everything we do. I look at it like the early days of the internet, where as an IT pro, I used to go to internet conferences. There are no internet conferences anymore because I don't need the value to be proven to me.

>> Right.

Zeus Kerravala

>> There are AI conferences today, and a decade from now, we won't have them anymore. Beginning from here to there, even in the early days of the internet was very problematic. What do you use it for? Should you take credit cards there? There's a lot of ups and downs, and we're going to go through that here. The long-term benefit, it's a no-brainer, but how you fund it, where you apply it, how you measure the ROI, I think we're going to see a lot of change in leadership in the S&P and things like that based on it like we did with the internet and cloud. These transitions always create a big upheaval in the business world. And so I think long-term trends are positive, but expect some bumpiness along the way.

>> Well, and the prevailing law determines those winners and losers. It was Moore's Law, it was Metcalfe's law in the internet phase, and that I would argue that really actually powered a lot of the cloud, the network effects, and certainly social media, big data clouds, social, and mobile. That was Metcalfe's law, was really the prevailing force at the time. And we will see if Jensen's Law holds true. He's been right a lot.

Zeus Kerravala

>> Yeah. But I think from an HPE perspective, they have such a big footprint with a lot of leading companies, they should be in just a prime position to be able to help take a lot of the early adopter customers that are using AI and be able to replicate that across their broad customer base. Their customer base is literally who's who of the business world.

>> Well, it's interesting, these infrastructure companies, it's almost like Dell and HPE have kind of a duopoly for on-prem. Certainly IBM's in there, but they're not really an infrastructure player.

Zeus Kerravala

>> I think the advantage HPE has, and I'm hoping Antonio meant this when he said it when they acquired Juniper and they had the Rami Antonio analyst Q&A, and he said, "We want to be known as a network first company." Right? And I think in fact, Sanjeev Katwa, the CIO of Tottenham Hotspur, I went to the session yesterday that he was in, and he said, "The network is the most important asset we have today, because without it, we can't do anything. We can't process tickets, we can't sell people stuff. We have no 911, we've got no," or whatever their version is over there. And so I think there is a pivot going on within corporate IT to be in a world that's more network centric. I think HPE, being a traditional compute company, is in a good position to make that change more so than even Cisco, because Cisco, it'll seem somewhat self-serving. Dell doesn't really have any network assets, and so-

>> They're taking a different strategy, building the networking into the service.

Zeus Kerravala

>> I'm really hoping that that message continues to carry through to next year's Discover and across other HPE activities, because I do think with the Juniper acquisition, they have just a very, very good portfolio of network products now, and they can actually do end-to-end networking today.

>> Right. And that's where this duopoly thing is interesting, but it's diverging, because like you say, HPE becoming a networking company, in many respects, Dell is much more storage heavy than say HPE is, so they've got some potential advantages there. And then you've got the cloud guys who are still growing at 20% a year,-

Zeus Kerravala

>> Yeah, it's crazy....

>> which is collectively 200 billion. And so you hear a lot about repatriation, but the numbers still aren't propelling

Zeus Kerravala

>> Well, you can repatriate and grow, right?

>> Absolutely. They're not mutually... They both can be true. We know anecdotally, repatriation is happening. It's just pales in comparison to the growth that you're seeing in the cloud. But it's like right now, it's good to be in tech, except SaaS is showing some cracks in the armor. So we'll see how that plays out, because they haven't really benefited from the AI wave. Not withstanding ServiceNow is obviously-

Zeus Kerravala

>> Actually, what's fascinating to me, Dave, is if you look at the capital markets, Nvidia has created this massive tailwind, but it's lifted very few boats. Arista has gotten a big AI bump. Pure Storage has. How many other infrastructure vendors have?

>> Well, Dell has. Dell's cranking this year. Super Micro has, and most recently HPE. HPE is up about, I want to say 22% in the last 30 days. So HPE is kind of... Their momentum's catching up. And I think HPE is being very conservative about their guidance. They beat and they raised. They're only... They're guiding to like, I don't know, one to 3% growth. I think they're going to grow fast. I think their AI server backlog is much bigger. And I think coming out of this show, they're going to beat again. But they're smartly, I think being conservative. So you're seeing a little bit of broadening beyond the Nvidias, and the Qualcomms, and the Broadcoms, and the TSMs to some of the infrastructure plays, but it certainly hasn't hit the software stack yet for the Mongos, Snowflakes kind of bouncing around right now. Databricks is private. I think they're growing actually very fast, I think for other reasons. Salesforce got hit a little bit. Like I said, ServiceNow is killing it. So that is the one anomaly.

Zeus Kerravala

>> We are seen have and have nots, so that's-

>> Yeah, very much so. All right, Zeus, hey, awesome to see you.

Zeus Kerravala

>> Yeah, always a pleasure.

>> Thanks so much for making some time for us.

Zeus Kerravala

>> Yeah.

>> All right, keep it right there. We're wrapping up day three of HPE Discover, live from Las Vegas. You're watching theCUBE.