What's Next After AI Adoption? HPE CEO Antonio Neri on the Future of Enterprise Transformation

All of that has come together in one unified operating model where the intelligence has been built at the core and that's why we believe as a company we are ahead of the game when it comes down to providing these solutions to enterprises of that unification.

Thank you. Thank you for having me. And thank you for joining us at HPE Discover this very, very critical week. I would say a fun week.

Well, it was a great day. Obviously, we had the keynote early in the morning and we had record attendance. It was unbelievable to see that room. It was almost two football fields. Think about how wide that room was. More than 9,000 people came to see it. But, you know, wrapping the day with my fellow compatriots, which I know because I met him a couple of times. It's remarkable to see what he's doing. It's just the best ever.

It's the best in the world. I remember four years ago sitting there watching that final. Do you think they could do it again?

Yeah.

And if he's on form, it's going to be hard to beat him.

Yes. I can't...

We're tailgating. That's what we're doing.

And you have a lot of that here on the show floor. I mean, we have seven acres of technology on display and the story is being told by customers which is what I love. One thing is you standing up there and sharing your point of view. One thing is a customer telling how they're using this technology for their advantage. And look, I spent more than 50% of my time talking to customers and partners. And when I think about the adoption of AI, it really hit an inflection point six months ago with the advancement we see with Agentic AI. And that's why we spend a lot of time around the Agentic Enterprise. And look, we see it ourselves. We have 200 use cases under a lifecycle process and 250 already in deployment. The value of the agentic AI is really bringing together AI in a process-oriented approach where you hyper automate using AI that workflow and we see it across all the functions: finance, HR, legal, supply chain services, and then of course in engineering you see the value of this technology to assist coding and application development. So clearly enterprises are seeing the value and then there is one metric I watch all the time: the growth in inferencing. Because the moment that inflection point continues to grow, inferencing is going to be the largest part of what we're going to see by the end of the decade and that's why I'm very encouraged about the adoption but most importantly I'm encouraged about what we have to offer. That's why our AI factory for enterprise is so critical.

Well, I mean look, I understand the hype around accelerated computing which of course was the GPU progress we saw tremendous progress in the last two years but everything else in the stack has to catch up. And the next layer of that stack is memory. And to me, storage is an extension of memory. That's why you see now things like KV cache and other things growing very rapidly. But ultimately, the biggest bottleneck is networking.

Yeah. Well, I mean, look, the networking is the core bottleneck to get the maximum amount of productivity for this accelerated computing. And so whether it's CPU now for inferencing and we made a number of announcements together with Nvidia or whether it's alternative training stacks like AMD Helios, and the need to really drive the maximum productivity for this amount of capex because we believe by the end of the decade we're going to basically have 250 gigawatts of compute power. You don't want any GPU or CPU to be idle. And that's why our thesis on networking is growing up very, very rapidly because networks for AI have to catch up to the other demands. And I believe with Juniper we have the best products to compete in that market.

And then networking represents between 15 and 20% of that.

Yeah.

Multiple.

Which you talked about being inside the Helio stack. Scale out because you now need to connect massive amount of GPUs across hundreds if not thousands of racks and then scale across because you have to drive the data center interconnect. Juniper brings all elements of that thesis to bear with the latest time-to-market technology, fully automated solutions with AI.

No, we have exposure. It's just intentionally we select what to play because look, in any customer segments the one thing I try to reinforce all the time is you need to think about the use case: training versus inferencing. And then you need to think about customer segmentation underneath. Obviously hyperscalers, service providers and model builders is one segment. Sovereign is another segment and enterprise is another segment. In any case, in any of these whether it's training or inference or whether it's segment one, two or three, we lead with networking. We have it all. That's where we can get tremendous value for our shareholders and we bring the right solution to our customers. Then for compute we prioritize the full stack integration with Nvidia for enterprise our AI factory including software because there the value proposition is time to value.

Right. And sovereign because we have a large supercomputing business which the adjacency to makes total sense in these large service provider is all about capital deployment.

Yeah.

Yes. That's sustainable and durable which is I think in a very important world and then bring the right relevancy to each of these customer segments and use cases. Look, time will tell but I think we are on the right path and the fact that we got six quarters ahead and we are pulling forward two years our original commitment and paying down the debt a year earlier it shows that our strategy is working.

Exactly. And so you've got to build something that allows because people we don't have enough hands. We don't have enough hands to manage all of this infrastructure. There's too much. You're very focused on security. The surface of risk is too big for people to try to get a human being can deal with that data.

Well, obviously we're using AI to our own benefit and we have this concept of self-driving and you talked to Rama about the self-driving network which obviously at the core is that AI ops and we have expanded a self-driving concept to all networking domains: started in the campus and branch, we went to the data center and we took it even to the data center interconnect with routing. I mean if you look at some of the demos with the routers, the self-optimizing capabilities we build are insane.

But because the technology is there and we have the data to deal with it. But we also extend it to the rest of the portfolio leading with GreenLake intelligence and that's the self-drive intelligence we built in everything we do. So the first principle, whether I was right or lucky or a true visionary doesn't matter, we build everything underneath GreenLake, the infrastructural layer to be cloud native. By being cloud native, it allows me to put a lot of that value into the cloud and a lot of the data to be in the cloud which means I can observe. And the OpsRamp acquisition has been a phenomenal success because that allows me to be more intelligent around that cloud native infrastructure and then I can package that infrastructure for the use cases: it can be virtualization, it can be AI, it can be just networking deployments, it can be just compute and storage deployments. It doesn't matter. All of that has come together in one unified operating model where the intelligence has been built at the core and that's why we believe as a company we are ahead of the game when it comes down to providing these solutions to enterprises through that unification. And so we are very happy because that allows us to cross-sell and upsell across the enterprise.

2019. You know, remember when I was here at Discover the first thing I said was in 2018 that the enterprise of the future will be edge-centric, cloud-enabled and data-driven. In 2019 we said we're going to offer everything as a company as a service. And then in 2020 we came out with our GreenLake cloud and since then we have added...

Yeah, it was before, before the pandemic. It was right before. But it's still sort of like you said, this is all sort of bubbling up and with what accelerated compute and AI has done, it's made all that look even more visionary. You know, when I acquired Cray supercomputing and we have the Cray one next to a quantum computer right to show that these systems will coexist for a long period of time, it is about economics, it's about scale, it's about all of that. When we did the acquisition of Cray which was in 2019, we understood that the largest workload that will run on a supercomputer would be AI.

Exactly.

But it becomes even more important now. I would say absolutely. And when we think about security, it's quite fascinating to me. Obviously we put guardrails around it and governance around it and that's what we announced yesterday and Fidelma today is going to elaborate further what that means from a technical point of view. But we build all of that inside our GreenLake cloud control plane so when you deploy the AI factory, all those capabilities come with it. You don't have to worry about it. And we even observe what the agents are doing. So you put the guardrails around the agents and if an agent goes rogue, we can bring it back and we can bring the entire environment back. No different than a human being will make a mistake sometimes through the keyboard. But when I think about security, the next evolution and you're going to see us knowing that the network is the on-ramp to use these models, and inference is going to be done everywhere. What if we build security at the core of the network? So we understood networking and security are converging. We are taking it one step further. We are actually doing it at the silicon level, not at the software level. So the next generation of our campus and branch switches which we now own, own silicon, is going to have both converged at the silicon layer, not at the software layer. So when a token gets through that silicon, it is being protected at the root of the infrastructure, not at the stacked up layers of the stack.

All right, so Argentina is going to win this year. Messi has a shot to repeat. But let's talk about AI in three to five years.

Well, I think we all say the same thing generally speaking, which is this is a new industrial revolution. I don't think about AI just as a technology. I think about AI as a way we live and work, whether in the business or in the society. So what we're going to see in the next three to five years, and look some people are going to be left behind. It's all about choices and we have a say. The future belongs to the fast. We are racing to that by basically adopting this technology to change everything. You know how you interact with us, how you place orders, how you get service, how you develop product is going to change dramatically. We already see it. I give an example: when I do my forecast calls with my team, it used to take three days for people to bring all the data from all over the place and as you know when you are a CEO normally you get a version of the truth that has been judged before it gets to you. Now not anymore. You're seeing it raw, pure. I see it and I see it in a way I can digest it. Although being in the company so long I have instincts when people are telling me the truth or not because I always said the truth is in the cold facts. And these models are cold-blooded. I give you a little example. The other day I was looking at something and there's a tiny little thing in the portfolio which is not big in the context of what we do and the agent said, "Look, you are behind and the reality is that you suck." What do you mean you suck? I mean literally you're not doing what you're supposed to be doing. So they tell you exactly what it is. So I think that's what we're going to see. That's the measure of success: how we change how fundamentally we operate and how we live.

Yeah.

I love it. I always say to my team: hyper-personalized, contextually aware. That's what I need to see. So I don't need to see a generic dashboard in CRM. I need to see the view I want to see and AI, if your data and all the other things are done right, presents an opportunity to do what we just never could do before.

At a scale we've never seen. Antonio, I want to thank you so much. I could chat all day, but I'm betting you've got a few other things you got to do around here.

Well, congratulations. We'll be watching close and we'll catch up. We always catch up around the quarter, so we'll definitely keep catching up. But congrats on a great event and thank you.