What does networking look like today? HPE Discover Barcelona 2025

I joined Juniper pretty much 30 years ago. I had a very simple role back then in the silicon team developing, or helping verify actually, small logic blocks for our ASICs back then. But then I took a career path through technology. I was a silicon engineer. I moved into management. Then I moved into product management. Then I moved into general management. Then I became the CEO of Juniper. I was very humbled and honored to have that responsibility. And I worked my way up from the most junior person on the team to CEO.

Thank you. Now, you've just finished your speech. How did it go?

Absolutely. And I believe this is your first HPE Discover. How are you finding it?

Actually, it's my second HPE Discover. I did attend last year's but as a guest because the deal had not closed yet.

In disguise. Actually, not so disguised. But this year it definitely hits different because I'm an HPE employee, and obviously I had much more access, a lot of great meetings, customer conversations, the big stage appearance, and we're not done yet because there's more to come tomorrow.

So I'm a geek. I love technology. I love products. So the highlight for me is just the products. The technology, not just within HPE networking, but if you walk on the show floor here and you see the unbelievable technology across compute, storage, networking, hybrid cloud, software, artificial intelligence, it's pretty amazing. It's like you are a kid in the candy store. And of course in HPE networking we have been working very hard since the close and combining our portfolios and doing something that's really special for our customers, and the progress we've made is pretty amazing. All right, let's clear up some common misconceptions because I think when a lot of people think of networking, they think of Wi-Fi, but networking is far more than just Wi-Fi, right?

Yeah, I mean, networking is everything. And now networking to HPE networking is Wi-Fi, wired switching, security, wide area networking, data center networking, AI data center networking, routing infrastructure solutions. It's a full stack of capabilities and solutions. I mean, we have the ability to address any networking use cases and any security embedded in networking use case for practically any industry, cloud providers, telcos, enterprises, you name it.

I think the best way to think about it is in four distinct technology pillars that we really focus on at HPE networking. There's data center networks, of course, routing infrastructure solutions that are the most mission-critical network carrying wide area network for telcos and large cloud providers and even enterprises that choose to run their own wide area networks, campus and branch wired and wireless infrastructure, and of course security which includes all aspects of security embedded in networking.

Well, it's a totally different business right now. It's far more comprehensive in terms of what it looks like. I mean, it looks awesome first. Second, I think it's good to start with our vision. Our vision is around self-driving networks. Networks that truly have the ability to delight customers or those that use the network to do practically anything today, and network operators, the heroes that are running these networks each and every day. So the best way to achieve that objective is through a self-driving networks. Networks that essentially configure themselves, run themselves, diagnose themselves, optimize themselves. You know, when there is an issue, they will self-mediate and address these issues well before anybody that uses a network has the ability to even detect that there was a problem to begin with. That is the vision. And the good news is that that vision is becoming more and more of a reality because of the technology that we have now inherited from both HPE Aruba Central and also Juniper Mist. The AI ops that we have developed and the agentic AI we've developed as part of these platforms is second to none in the industry, and we're applying it across every layer of the technology stack. And that's something that 10 years ago just wasn't even conceivable, that sort of self-driving networks. It was very labor intensive, very just hard to wrap your head around. The progress that has been made towards the vision of a self-driving network over the last 10 years has been staggering. Starting with data collection, to insights, to assisted self-driving, and now with agentic AI, really the opportunity is completely self-driving where the network operator can literally take their hands off the wheel and let the network run itself.

I mean, there are areas where if you look at it from a distance they might look like they overlap, and the one obvious area that customers ask me about would be in the campus and branch where HPE Aruba networking and HPE Juniper networking have access points, have campus switches. But if you take a closer look, really these two platforms have unique strengths. Now Mist was developed as a cloud platform, and Aruba Central has much more of a diverse deployment model from virtual private cloud, on-prem, etc. So together we can support every single deployment model under the sun. And the beautiful thing, this is what I talked about in my keynote just today, is that the approach we're taking is to cross-pollinate the best of each platform onto the other. And in so doing, we're accelerating the pace of innovation for all of our customers without leaving any customer left behind. That message, by the way, was incredibly well-received. And that's what we're executing on. And you know, some customers were saying, 'Well, that's years out.' No, it's not years out. We're actually introducing the first set of microservices that we're cross-pollinating just weeks away from here. You've worked in the world of networking for a fair few years. How have you seen networking change in the last 30 years?

I mean, I had a front row seat way back when I joined Juniper at the explosion of the internet. Back then there was nothing in the industry that could keep up with the pace of growth of the internet. And this is why many customers had no choice but to bet on a small upstart, a startup company like Juniper that could develop the technology that can keep up with the scale of what was happening in the network. But the network has evolved a lot since then. We've gone from the internet to mobile computing to cloud computing, and now we're in the era of AI. And as important as the network was back in the era of the internet, networking now in the era of AI is even more important. I mean, it really is an exciting time to be in networking today just because of how critical everybody sees the network is to practically doing anything today.

Yeah, I mean AI workloads are just fundamentally different. The amount of pressure that networks are feeling, experiencing right now because of AI workloads, both within the data center and in the wide area, is stunning. And this is why some of the technologies that we've sort of been dabbling in the labs like liquid cooling technology and co-packaged optics have now moved from a realm of nice to have into the realm of must-have. Like the new QFX 5250 switch we just announced, over 100 terabits per second in a single switch, 100% liquid cooled, and we were able to tap into the liquid cooling capabilities that HPE has developed for decades as part of the supercomputing technology that HPE has been innovating in and apply it to this switch, gives us a significant advantage in the market.

It's a lot, and there will be customers that will need even more than that soon.

Yes.

Yes.

So it goes back to your question about AI workloads and how AI sort of affects the requirements of networks. I mean low latency is one attribute. High performance, no congestion. Congestion is absolutely evil for networks in the AI context because at the end of the day, congestion results in using your very expensive GPUs less efficiently. Nobody wants to waste GPU cycles today. And the way to not waste GPU cycles is to have high performance, low latency, and congestion-free data center networks, which happens to be exactly what we know how to do really, really well. Because your GPUs that maybe cost tens of thousands or hundreds of thousands of dollars, they're a prime asset in your organization. You want to be using them flat out. You don't want any congestion. Nobody wants to spend millions of dollars in GPUs and then stinge out on a bad network that's going to use those GPUs at like 50% utilization. That's what we're doing. We're building the networks of the future that can keep up with all of the attributes of AI networks. Low latency, high performance, congestion-free. We've innovated not just in the hardware, but in the protocols, the software that can achieve all of these attributes.