As an apprentice, an engineering apprentice in the Argentine Navy. You from there, you studied engineering at the National Technological University , where I also believe you studied art as well. And very early in your career, you left Argentina and moved to Italy to join a small IT company. You then moved on to join HPE, or HP as it was at the time, in 1995 in Amsterdam. You worked through a number of very, very senior roles, ultimately culminating in, in February of 2018, you were announced as the CEO of HPE, replacing Meg Whitman, and then you also joined the board of directors. You have been credited with spearheading some of the key technologies, Apollo HPC portfolio, the Superdome X.
And today, beyond being the CEO and president of HPE, you're also on the board of directors of Elevance Health and on the advisory board of Maserati, which sounds like a very, very cool responsibility. Antonio, we don't have a great deal of time. We've only got 30 minutes, so I'm gonna dive right in. And I think the most logical place to start is a week ago. You released an exceptional set of results, revenues up 15%. What was sort of really the underlying drivers of that extraordinary performance that really, I think, validated your strategy in many regards?
I think the results of last week is a testament that the strategy that we put in place when I became CEO is working, and we have been curating that strategy for that period of time. But it's not. It wasn't no one thing. It's multiple things that we think are playing now in our favor based on the mega trends of the market we see today in networking, in hybrid cloud, and AI. And so when you look at our performance, obviously very pleased at the company level, we grew 15% year over year on revenues. We delivered record-breaking performance and operating profit of $938 million. We achieved growth on every business on a sequential basis. We grew orders year over year pretty much in all our businesses. We delivered the best financing volume for HPFS.
And when you look at the full 2024, we exceeded our revenue guidance. We exceeded our GAAP and non-GAAP EPS guidance and commitments. And we delivered record-breaking free cash flow performance of $2.3 billion, the highest in the history of the company. And underneath that, what we see is that the networking business is on a recovery from an orders perspective. We have transitioned a lot of that business in a subscription-based model. So over time, you're gonna see that revenue climb. Remember, we are going through the period of digestion after COVID where the Aruba business.
Yep.
We closed that transaction in 2015, and Barclays was our leading banker at the time. It has been an incredible success. Obviously, we have now the pending, you know, acquisition of Juniper, which will double the business in our networking business. It will be an $11 billion business, with very strong structural margins. Very excited about that. What it tells us is that the networking market recovery is gonna be a tailwind for us as we go 2025 and 2026. On the hybrid cloud, which has been quite interesting, people were a little surprised at our performance. I mean, if you look at our customer adoption on the platform, it should tell you that this is not a coincidence. We have now 39,000 customers on that platform. The underlying growing drivers are four. Number one is our storage performance.
It was very, very strong on our IP products. And Alletra is now a billion-dollar annualized revenue. And we introduced a series of new offerings in just the last month, targeting AI specifically, with object storage and file. When I think about private cloud, it's a driving growth for us. And we have fantastic offerings both on the virtualization side with our own HPE VM Essentials software.
Yep.
That gives an alternative to the VMware stack, but at the same time allows you to manage the VMware stack, and other stacks. And then, the other piece of this was the private cloud for AI, which is now ramping very nicely. And last but not least, it's all the adjacent software , and services that is driving two-thirds of the annualized revenue run rate, which now, is at $1.9 billion, again, growing at 48% year over year. So all those are driving, elements, plus all the subscription to networking and servers, because each of them requires a software which is disaggregated inside the GreenLake platform. And so for me, that's very important because networking is the core foundation from edge to cloud. Hybrid cloud is necessary to deploy AI 'cause data lives everywhere. And then we have the AI business.
In the AI business on the server side, there is a very strong momentum on the traditional server side, but the AI servers are obviously growing very rapidly. And in the quarter, we booked $1.2 billion before the booking. We felt it was absolutely necessary because the customer was not at the level we wanted in terms of delivering on their commitments. But that said, today we are at $3.5 billion of backlog, which is very strong. And then, you know, we continue to have a pipeline that's a multiple of that. And so very exciting.
I think when you look at AI, there are four different segments that HPE has the ability to play: enterprise to the private cloud AI enterprise solution, the sovereign, because we have both sovereign in terms of supercomputing and sovereign in terms of generative AI, and then the model builders and the service providers.
Yeah. Yeah. I mean, on the back of such an outstanding set of results internally within the organization, you know, what are you talking about as being the top two, three priorities that you're gonna really focus on in the next three?
I think I said it in my remarks during the call. For us, I can wrap the year with one word, which is innovation, so that's one of the key messages across the company. When you look at the innovation that we brought in the market in 2024, but also in 2023 and 2022, first through the focus on the edge, then focus on the service, then focus on storage, and then focus on AI, we have tremendous momentum.
I just had a conference in Barcelona two and a half weeks ago where we made a series of announcements again, and there were 5,000 customers there, all very excited about the innovation, because the way we create value for our shareholders, yes, of course, there is capital deployment and, and the like, but ultimately it's about relevancy with customers that are willing to buy your solutions that generate profitable growth. And we see the number of customers growing very rapidly. In fact, in 2024, we had 9,000 new logos to our portfolio across every segment.
Yeah.
And so that tells me we are becoming way more relevant. They're all very excited about the Juniper acquisition, and that drives value for our shareholders. And that value is shown in the profitability of the business and the free cash flow that we generate. So that's why I am very excited about 2025 and 2026. And our organization has now the highest engagement scores I ever had in the organization. We are at 85%, you know, the employee engagement index. You know, I just had a call the other day with all the team members, and, you know, when you ask a question, you know, is the strategy relevant? Do you see us as a winning organization? And that is in the mid-90s, which is very, very high. So and culture plays a role.
To me, culture is everything in the end, in the end of the day, because without the right culture, the right, you know, the right behaviors, you can't do anything. That's why innovation and talent management and mobility are the two key priorities. Why? We continue to find tremendous opportunity to streamline and simplify the organization. Obviously, we are deploying AI ourselves. I can tell you we have more than 200 use cases that we are either already deployed, at least close to 20% of them, or we're going to ideation and testing so that we can rapidly turn to that. Then we can also consult with our customers, how to deploy it.
Yeah. Yeah. I just wanna pick up on what you talked about around innovation. As the CEO of the organization, how would you describe sort of the model of innovation inside HPE, and how do you sponsor it, and how you use it as a means by which you differentiate yourself from competitors?
Yeah. I think there's multiple ways. So first of all, innovation comes in three forms. Organic innovation, where you invest a dollar from your balance sheet, is a better return for shareholders because, you know, I subscribe, you know, dollars per gross margin basis, how much that dollar can generate, and then making sure there is a hell of a return on that. Then there is the inorganic innovation, which you need because you can't innovate everything yourselves. And that's where, you know, Juniper or Aruba or Morpheus Data or Cray or many others. I have done 35 acquisitions in the last eight years to accelerate that innovation so that we can take that unique IP and the unique talent and scale it to our go-to-market. And clearly, that has resonated with customers and partners, but it needs to be integrated. It cannot be a widget.
And that's why GreenLake is so important. And last but not least, these very unique curated partnerships. So those are the way I think about it. But ultimately, this mentality of fail fast and improve is so important, and the future belongs to the fast, and the fortunes favor the bold. So when I think about that, you know, in the context of innovating in GreenLake, we have to take a different approach. When you are such large businesses inside the company, what tends to happen culturally is that everybody wants to de-risk their business.
Yeah.
But now, in order to deliver a solution, you need to innovate horizontally. And when we started the GreenLake journey, we had to select what I felt was 30, 35 people out of their jobs and focus on that transformation. Not commingling, nothing to do with the rest of the business. And once you scale it, then you bring it back. And so the approach has evolved over time, Paul, but fundamentally, you know, it's not just technology. It's business model innovation and culture and operating model shifts. In fact, in our sales model, we have shifted over the last seven years multiple times from HP, obviously split it. We had to reduce the number of layers. We went to a geo model. Now, between me and the seller, it's only two layers.
Yeah.
So I get the truth for the most time until you inspect, right? Then we went to a common segmentation. Then we shift the model to orders-paid , to our sellers. Then we went through a specialized model. Each of these has been a turn of the crank.
Yep.
Based on where we were on the portfolio innovation and ability to go sell it against the competition.
Yeah. Yeah. AI servers, obviously, a very, very hot area. How are you at Hewlett Packard Enterprise leveraging your leadership in supercomputer and high-performance computing to win business?
It's a core tenet, right? So when I think about the requirements of generative AI, I'm not that significantly different than a supercomputer. You're talking about accelerated computing capacity, parallel computing. Ultimately, the magic, yes, there is this silicon discussion that all we are having. And honestly, there will be many more in the future years. But the reality is to bring it all together, you need a unique expertise. And supercomputing does that. You know, whether it's an NVIDIA GPU or whether it's an AMD GPU or whether it's a NAS array of sorts, the magic of that happens in the infrastructure design and the networking fabric. And so we design architectures that are very large, sustainable.
Yep.
And this is why HPE has been in the liquid cooling for decades. Obviously, we acquire a lot of IP through the Cray acquisition. And we have done liquid cooling for at least 20 years+ . And we have a lot of patents. But we are the only company that today is the first of a kind that has 100% direct liquid cooling. So when you go to direct liquid cooling, what happens is that you can save 90% energy and you can save up to 50% space. But today, when the people refer to liquid cooling, they refer to what I call the hybrid. You cool the GPU or the CPU, but then the rest of the system still air-cooled, which means you need fans. When we do liquid cooling, we don't need fans. It's actually the quietest system you can be in.
There is no blowing air anywhere. Therefore, we can increase the density and the energy efficiency. So HP has that expertise at the infrastructure level. At the same time, you need to have a lot of services capabilities because when you go to liquid cool, now it's like, I always make the joke, it's like, how many of you have aquariums here? You have to maintain the aquarium and the bacteria in the aquarium. Those systems run millions of gallons of water a day. So it's a very unique chemistry. You have to be able to maintain. The maintenance cycles are very, very different. Then this networking fabric is very special because ultimately, you have to be able to connect 40,000, 50,000, 60,000.
We just introduced the next generation of that fabric with 400 gig, which is liquid cool. HP has the only liquid cool switch on the planet. And fundamentally bring out, in a way, the access of one system. In the cloud, if a server failed, nobody cares. It's just, it's virtualized, moves on. But if one GPU fails in a run, you're gonna have a problem.
Right.
That's what HP knows how to do.
Yeah. Yeah. Antonio, let's turn to the acquisition of Juniper.
Yeah.
I might get you just to talk a little bit about how that acquisition that you announced in January, how does that fit into that overall vision that you just articulated?
Yeah. I think when the acquisition is closed, at the core, HP will be a networking company. That's the message number one, and fundamentally, it's gonna have every aspect of the networking stack to challenge the incumbency, which I believe was here, an hour or so ago, and do it in a modern, secure, and AI-driven way because the networking infrastructure is not keeping up today with the demands of AI.
Mm-hmm.
I just explained some aspects of that.
Yeah.
So fundamentally, HP will have an edge-to-cloud architecture from silicon. And I wanna remind everybody that in the campus switching, HP has its own silicon. I do not buy 90% of the silicon from Broadcom. It is our HP Aruba networking silicon, but that's an example of it. Juniper has a lot of silicon and high-performance fabric with routing and data center switching. So those are two aspects: silicon, infrastructure, operating systems. They are proven both on the ArubaOS-CX, which has now more than 30 million ports deployed, and then obviously the Junos OS on the Juniper side, both for switching and routing. Then security, traditional firewall with Juniper and what I call cloud-native approaches with SSE with the SD-WAN. We have an SD-WAN business, which I acquired in 2020, with Silver Peak. And then all the AI ops capabilities.
So I feel HPE plus Juniper will give you customers a much better architecture that will catch up with the demands of AI, plus the HPE Slingshot architecture, which is the liquid cool architecture for large systems. That will become the core of the company, as we go forward. But that technology will find its place in other places, including in the storage business, because in the storage business, you need a switch to connect disks or flash to controllers. And as we already are in the 800 gig, we can actually replace third-party switches in our storage arrays with our own switching. But when you think about these AI data centers, the first thing you need, simply put, you need a big pipe into the data center. That's a routing. And Juniper has already the 800 gig router, high speed.
Once you are inside the data center, you need to lay all the pipes.
Yep.
To connect all these halls and racks, and that's the Juniper QFX fabric with 800 gig. Once you go to the rack, obviously you can adopt different architectures, but with our Slingshot fabric plus high-performance silicon for Juniper, we can collapse that fabric and make it more CapEx-low, savings in CapEx and a lot of OpEx savings because you don't have to manage two control planes or sometimes three control planes, and that's the opportunity for us, in addition to drive high performance, and then HP has silicon photonics, which we intend to transfer to that networking business so that we can also capture the opportunity at the interconnect level, even outside the data center, so that's why the opportunities are pretty significant beyond just the synergies of the deals that, you know, generally financial people tend to understand.
Yeah.
But what excites me is the ability to really capture these massive inflection points with synergies of the deal that makes this a no-brainer for shareholders.
Does it talk about, does this introduce you to a new client set that you can now bring the full suite of the consolidated HP?
Absolutely. I mean, obviously, HP has been with Aruba business pretty much contained in the, what I call the campus and branch segment of the market. And there we have done a fantastic job by unifying the connectivity layer between Wi-Fi, which obviously now we are going into the Wi-Fi 7 cycle, campus switching, SD-WAN, private 5G, and HP has a fantastic private 5G offer through another acquisition we did with Athonet and then IoT. On top of that, we built the security layer with access security and then the AI ops. But that was campus and branch. But now I extend all the way to the cloud. And fundamentally, whatever you want to size the TAM opportunity, we go from approximately $35 to $40 billion total addressable market to $135 billion addressable market.
Yeah.
So we more than tripled the addressable market. And the combined business is still below 20% market share, which means the runway is pretty significant. But that runway comes with a complete different margin profile than the current margin profile HP has as a whole company.
Yeah.
And that's why, I'm excited, again, from an innovation perspective, from a customer relevancy perspective, from a shareholder perspective. This is a very unique combination.
Yeah.
That's very strong.
Yeah. Yeah. Terrific. Maybe I might get you to talk a little bit about partnerships and how partnerships feature in the growth strategy of HPE. And in particular, if you could talk a little bit about the partnership that you recently announced with NVIDIA.
Yeah. I mean, obviously, Jensen and I have spent a lot of time thinking about what the combination of NVIDIA HPE can do better for our customers. And we both agreed from our history, you know, we have participated together in the sovereign space with some of the supercomputing or some of the generative AI clouds we are building today for both the University of Bristol or the Japan AIST entity. These are NVIDIA GPUs with the HPE infrastructure, servers, and everything else we talk about, including the HPE Slingshot. By the way, the Los Alamos system is exactly that. It is a NVIDIA GPU with everything else being HPE. But we said, "Hey, ultimately, all of this is wonderful, but if you don't sell and you deploy this technology at enterprise level.
Yeah.
What it's for, right? One thing is consumer and one thing is commercial. And so we felt that let's go and architect a value proposition for enterprise that addresses their unique pain points. And those are pain points at the core in enterprise is time to value and return on investment. And we felt that the simplicity of the value proposition needs to be as simple as three clicks and less than 30 seconds to deploy an AI application. And that's what we did with NVIDIA AI Computing by HPE , which really brings our HPE GreenLake cloud with our control plane, all our services in the cloud, including the data fabric, including the data lakehouse , including all the infrastructure lifecycle management for server storage, networking, plus NVIDIA stack, which people used to call it CUDA, but it's NVIDIA enterprise software, plus all the NIM agents.
Then we added on top of that, since the announcement. Now the configurations are available since September, and people are already buying and deploying is the unique partnerships with people like, companies like Deloitte.
Yeah.
Because ultimately, it's a business process optimization where the focus is and business productivity realization when you deploy AI. And so that's why I'm excited that this is not about giving them a reference architecture and a bunch of widgets. It's about giving them a complete integrated value proposition that addresses those, those two fundamental problems.
Yeah. Yeah. How are you thinking about the outlook for 2025, sort of from a macroeconomic perspective? And what is that telling you around sort of the visibility of sort of orders growing as you head into 2025?
I think, first of all, if you take it by market, I will say, you know, in general, the United States has been very solid.
Yeah.
But we'll see what the new government transition will happen. You know, sometimes there is, you know, some people are thinking, "Hey, let me see what interest rates does, what the policies are." But if you go back to the incoming administration, which was here eight years ago, you will see it was the first year was very positive. Remember, we had the tax reform.
Yeah.
This year, we may have a think different thing with tax and tariffs. But fundamentally, I'm positive. Now, then you go underneath each of the segment. I think the server segment is in a positive upswing on the cycle.
Mm-hmm.
We at HPE have 500,000 servers that we have an opportunity to go modernize by shrinking the footprint 10 to one , reduce energy by 65%, and then capture the unit plus the services attached. And by the way, when you add our competitors, that's a bigger, bigger number. Why that's important is because many customers need to free up data center space and power. When you were on the CPU base, you were running single digits kilowatts per.
Yeah.
Per rack. Now you're running at least double, in some cases triple. And so they understand that AI is not just a computing thing, because in the cloud space, when you went through the cloud journey, it was about, "Okay, let me ride the wave of innovation in the public cloud, and let me get a commodity computing unit." This time it's about where I'm gonna host the data to train or fine-tune that model. So it's a data-first approach, computing second. So that's why, you know, I think the server will continue to be solid. And then obviously, our ramp will be continue to be accretive.
Yeah.
We have not seen cannibalization between one and the other one.
Yeah.
The number shows that because we have had 3/4 of double-digit orders growth on our traditional CPU while we ramp the AI business. On the networking, I think the market is on a sequential recovery, and we have had three consecutive quarters of order growth. As I said earlier, I think networking is gonna be the upside for HP today and after the Juniper acquisition. Then on the AI, I think, you know, like I said earlier, I think about AI in four unique segments. There are the model builders and the hyperscalers. They will continue to drive a lot of demand because they are in the business to build these large language models and integrate the models in their applications where they make sense.
You have the tier two, tier three, which are there to support either sovereign requirements or being hosting of people that need capacity on demand or enterprises that may need to use a colo because they don't have space. You have the sovereign, which obviously there are at least 15 countries taking the lead, European countries, Middle East, very strong. Obviously, they're going through the process for the importation license. You see more positive news on that. Fundamentally, every big geography wants to control their AI models for culture, for geopolitical reasons, and governance and compliance and security. Last but not least is enterprise. It's funny because you look at this model, you can define it in two axes: number of customers, number of GPUs.
So there is the first segment one, which is the model builders and hyperscalers. Few customers. You can count it with a finger or two hands, but they drive a million GPUs, right? You go to the tier two, maybe 20 to 40, depending on the capital, capitalization of this company, and then drive maybe into the tens of thousands that go into the hundreds of thousands GPUs. You go to sovereign, they're very bespoke systems to begin with, and those that will be 15 to 25, maybe in the tens of thousands GPUs. They go to enterprise, tens of thousands or hundreds of thousands of customers. But you know what? They're gonna consume hundreds of GPUs because you don't need all these GPUs. You need few GPUs, that they are doing this rack and fine-tuning and inferencing. And inferencing obviously will continue to grow.
Sometimes you do it at the center, sometimes you do it at the edge, and that's another opportunity with Juniper, with the presence on, you know, the core, obviously, but then moving to the metro and the edge, that's where we can expose some of these inferencing capabilities through the network connectivity.
Yeah. Yeah. Antonio, unfortunately, we're nearly at time, but I did wanna ask you one final question. Like, what's your big prediction for 2025?
You know, I think we are living a remarkable time. I think, you know, there will be some, you know, rationalization going on into this AI space because obviously there was a lot of hype to the beginning, but the technology has tremendous potential. I see it everywhere. I spend 50% of my time talking to customers. But we have to make it simple.
Yeah.
We have to make it super simple for the enterprise side. But I think over the next three years, I do believe IT will continue to outpace GDP. And that has to do with a lot of things we see today around data, modernization, and capabilities that ultimately everybody needs today. Without IT, you can't run any company. And so, those are very positive. And I'm more positive on our hand because HP has a unique set of assets that no one else has in the traditional space we are in.
Yeah.
Some have unique things in one area, but no one has.