 Welcome back everyone, live here at theCUBE in Barcelona for MWC, Mobile World Congress. I'm John Furrier, Dave Vellante. We have a special guest analyst host, Aziz Carvella. Aziz, great to see you. Hey, always great to be on theCUBE. Let's put this together. We've got a distinct CUBE alumni, Rami Raheem, CEO of Juniper Networks. Soon to be HPE with the acquisition, not yet closed, but great to have you back on theCUBE. It's been a couple, been a New York minute. Thank you. Thanks for coming on, appreciate it. Thank you, appreciate it, it's been too long. Well, Aziz put this panel together, I appreciate that, Aziz. So let's get into it. So my first question before Aziz takes it over is the innovation around Silicon, the innovation around AI and networking with your acquisition of MIST has been really kind of pointed to as a future, and we actually brought up an opening segment here, that that's the kind of thing you're going to start to see with AI. AI for networking and networking for AI. It's an area that has been waiting to be innovated for in the cloud native world. It's like the last piece of the puzzle on cloud native. But yet the AI native world is coming super fast. So what does that all mean? What is your perspective on this? What is AI for networking and what is networking for AI? What does that mean? Yeah, okay, it's a great question to start with. And let's start with AI for networking because that's actually an area where we have invested and seen incredible progress. And I know there are a lot of people at the show that are talking about AI. We're delivering on AI and we're delivering in a way that's actually provided incredible financial results for Juniper. Last year our AI driven networking business grew 70% year over year from the year before. How do we do this? We're offering solutions to our customers that essentially transform the way they interact with their networks. It's reducing the cost of running networks. It's enabling them to delight the end user with an exceptional experience. All of this is being accomplished by replacing human effort with artificial intelligence that's delivering value day in and day out. Transition to the next big opportunity that we are pursuing at Juniper and that's more around networking for AI. Here you have to understand that AI doesn't just appear out of thin air. AI has to be offered from rich high performance infrastructure. This is where silicon high performance software comes into play to enable AI for the variety of applications that we see around the globe today. Both of these opportunities are multi-billion dollar opportunities for Juniper that we are pursuing. How big do you think that those opportunities are there? If you think about the five year TAM, I've heard networking for AI alone be anywhere from five billion to 20 billion. So if you think of those two as separate markets, how would you size the five year TAM on those? So the AI for networking, essentially replacing human effort in running day to day networks into robots, software running these networks, it's as big as the TAMs are today because if you think about our campus and branch, that's a 25 billion dollar TAM. Data center is also 20 to 25 billion dollars. There's 50 billion dollars of total addressable market that is ripe for disruption by applying artificial intelligence to the management, control, day to day operations, reduction of travel tickets, simplification and delighting the end user. The newer opportunity is on the networking for AI, that's around offering these high performance data centers to enable artificial intelligence. There, I've heard the TAM is today roughly around a couple of billion dollars. I'm specifically talking about Ethernet because the technology of choice today is still in Finneban. We expect that's going to transition rapidly to Ethernet over time. I've heard predictions of eight billion dollars within three years. That's a massive opportunity and it's rare that you see off markets grow that fast. On the whole, and we'll click on the Infiniban Ethernet question. We're seeing Infiniban for purpose built clusters and Ethernet much more for a broader market. Is that how you see it? Because people are discussing is Infiniban only for specific architectures or clusters or will it prevail on the open side at scale? What's the difference? Where's the trade off? Yeah, well, let's face it, Infiniban and Ethernet are competing for the same addressable market inside the AI data center opportunity. We've been talking about this a long time, right? Ethernet eating away Infiniban, but it hasn't happened. So it convinced me it's going to happen this time. Well, I don't know how long we've been talking about it. I do believe it will, it's just a matter of time. And it really comes down to economics. When you have a number of different technology providers, Juniper, Cisco, Arista, others, that are all competing and innovating to address this opportunity in an open way, it's just a matter of time before Ethernet starts to become a superior technology to Infiniban from an economic standpoint, from a speed of innovation standpoint. I think it's already happening. As we get into 400 gig and 800 gig, I don't know, I think Infiniban will be left behind. Even in videos in that mix doing Ethernet. That's true, that's true, but you know what? Game on, at the end of the day, I believe that those that bring to the table the best solutions from the silicon, the software, the automation capabilities, the visibility will ultimately win this market and we're in it to win it. Well, again, you add them to the list it's obvious that Infiniban is the open standard. I want to come back to when you were talking about the market. Is there an analog to cloud? When you think about the cloud market, the infrastructure market, let's say it was 100 billion, maybe it was 200 billion, whatever number you choose. And then AWS made famous the line undifferentiated heavy lifting and they basically replaced a lot of human labor. It sounds like you're doing a lot of that with AI for networking. A lot of journalists freak out about, oh, losing jobs, it's an important conversation. But did the cloud, it replaced a lot of human labor, but the market didn't shrink. People didn't lose jobs because of it. Is there an analog there with AI for networking? You know, as I travel around the world as I am doing right now and talking to IT professionals, CTOs, CIOs, there is a common theme of the kinds of challenges that they are facing. Number one is a skill shortage. Number two is a talent shortage. Number three is typically 70 to 80% of the effort that is exercised by their team at any given time is not going into innovation. It's not going into avoiding the disruption that's happening from their competitors. It's going into rudimentary keeping the lights on, just keeping the network running. And we believe that job should be solved, should be accomplished by the robots, not by human beings, so that we can free up the time of human to actually work on digital transformation efforts and actually innovative new capabilities and revenue generation opportunities. That is what AI is solving today. My advice to IT pros right now is if you're doing things as part of your job that aren't going to advance your resume, forget about what it means to the company. If you're doing things that won't advance your resume, stop doing them. The only way to stop doing them is to automate them out of your job. Or the quality of their work. If they're grinding away, they can, that toil, they just, does the machines do it, they can move to a higher value piece of work. But I think Dave's point is a good one though because there's far more compute engineers today than there were pre-cloud, right? And if you think of with hybrid work and IoT and things like that, we're going to need more network engineers. But we can't continue to run the networks with the way we did before by hunting and pecking at CLI. Look, it might change the nature of the work, but it doesn't eliminate the need for the work. I honestly believe that every customer that has seen the light, basically there were a lot of skeptics about AI when we first started to offer MIST. Today, we're offering MIST across all aspects of networking from wireless to wired to WAN to security. Every customer that has seen this has walked away absolutely thrilled and delighted because it has simplified their life and it has given them so much more time to focus on what's actually truly important in their business. We did a panel on theCUBE with Bob Friday and the folks over there on the team in 2019 unpacking AI at that point. So you guys saw it early, congratulations, great investment, it panned out. And it's leading the industry conversation of the future of operationalizing some of these abstraction layers and automating it with the human, very key. The question I want to ask you is, at what point did you realize that the data was very valuable and that you can go there? And then the second part of that question is, as telco, and this is kind of a telco enterprise show, let's just be, it's a cloud data show for telcos and enterprises. What's the value of the data because people here is trying to assess what data do I own? What data is available? If I own my data, what IP is that? So when did you realize you had the magic and the data? And then how do customers figure out how to use their data and networking? So, I mean, AI obviously thrives on data. If you don't have the data, the AI is actually quite useless. And in fact, the big differentiator in the missed solution is not the AI algorithms themselves, although we have some amazing data scientists that are implementing these algorithms in amazing ways. However, it's the ability for MIST to consume vast amounts of data and store it in a scale out matter in the cloud, like basically infinitely scalable across the largest networks in the world. Only then can you apply the algorithms and then close the loop with real time response to add value to the end user. Without data, none of this would work. So that's a data flywheel right there. You got that going on? I love the term flywheel because just like people talk about the network effect where networks become sort of exponentially more valuable as you add more and more users, I believe that there's a similar flywheel effect with data. As you accumulate data, you learn from that data, you gain insights from that data. It allows you to enhance the product. When you enhance the product, what happens, your customers are happier, you get more customers. With more customers, what happens? You get more data. And that is a positive flywheel effect. That's incredibly powerful that's driving our business today. I think the other underappreciated aspect of MISTO, and we talked a little bit about it this morning, is MISTO was founded in an era where cloud native was becoming the norm, where if you look at a lot of your competitors, and it's not nothing against the competitors, but when they created those products, there really wasn't the concept of cloud native. So that was more the analogy of they took a controller and they did a lift and shift into the cloud, which is different than what you have, correct? Absolutely, we have the advantage of an amazing team, but also the advantage, as you mentioned, is the use of time. We invented the solution, the MIS team, invented the solution at a time that cloud native backends actually existed, and that has given us a massive moat right now around the solution. I also want to address your question about what does it mean for telcos, because it's really important, but at the end of the day, we're at level of Congress. There are a lot of telcos here that are looking at the MISTO success and asking how do they take advantage of it? Well, first and foremost, there is no reason why you can't apply the AI ops capabilities to a telco network. We've already expanded MISTO from being a primarily Wi-Fi solution into Wi-Fi wired WAN and security. We can include wider routing, and we will include wider routing. Second, I do believe on the other aspect of AI, which is more about networking for AI, the edge of the network, right? Whether it be in the central office or in the enterprise, it's still white space. It's still an opportunity for telcos to come in there and offer edge services, potentially with AI inferencing, as a new revenue stream generation approach in their business. So how do they feel about that? Because if you look at all the cycles we've had, and Dave, you mentioned a few cloud and trends like that, telcos really were scared of the technology, late to adopt cloud, late to adopt software, and they missed out on the opportunity to be able to monetize that, right? Today, when you look at Edge and AI, are they on board with that? Do you think they're mindsets actually in a place where these mythical new revenue generating services will actually be realized? I think that's the case. I don't think everybody believes it, but you only need a few telcos to provide an example. And I view our role at Juniper Networks as an enabler, as somebody that can work and partner with our telco partners to show that this is in fact an approach that can drive new revenue streams for them. The data, the data between networks, whether it's private 5G or private hybrid, public private networks on mobility, and then you got the telcos have all the data, not just customer data, network data. So imagine putting a foundation model around that. That's where they could have a net new business model. Definitely a data play here. I mean, a whole nother data play for the... Telcos are sitting on a goldmine of data about their customers. They don't know how to get that to their customer, that's the problem. If you can get to that data. They know they're sitting on a goldmine, they don't know how to get it out. There is a way to get to the data and to leverage that data, and they're also sitting on prime beachfront property for Edge Cloud services, right? Central offices and the relationships that they have with their enterprise customers, all of the components, all of the pieces are there. They just need to be combined, assembled, and formed into an amazing offering. I think it's possible. Antonio was on earlier today and talking about the Edge opportunity. Can you help us understand the rationale for the acquisition, irrespective of the financial outcome, which is incredible. I'd love for you to take us inside the boardroom, which you probably won't, but what were your strategic alternatives that you thought about? You could have made some other big moves that were super risky. I'm sure HPE the same. I mean, they've made a lot of little tuck ins. This is a big move for them. Explain the rationale and what it means for customers. So Juniper was born in the era of the internet and we bet big that IP networking was going to create this massive global network. And ultimately that bet paid off. Fast forward to today. The big inflection of our time is artificial intelligence and I believe that incremental moves towards capturing this market opportunity are not going to be sufficient. You need bold moves. And this is an example of a bold move. This combination of HPE and Juniper is incredibly exciting to me and honestly in talking to many of our customers, very exciting to them as well. We have an ability here to take the innovation that comes from both sides across compute storage networking in pretty much every single domain of the network from client all the way to the cloud and the application layer and form turnkey solutions for our customers and have the global reach and scale necessary to reach more customers internationally. It's just the sky's the limit in terms of possibilities. It's funny when you talk about turnkey though because this is one of those pendulums that swings back and forth where five, six years ago everybody wanted everything disaggregated but now nobody can put the stuff together. And everyone, more and more customers including the service providers are looking for these turnkey solutions because infrastructure today isn't just network, it's not just servers, it's not just storage, it's all of it working together. First of all, you can have turnkey end-to-end solutions and also be open and give your customers choice. And that's exactly what we will do. But I think as you think about the AI opportunity and in particular networking for AI where now you really need or there's an opportunity here to offer end-to-end solutions that include all aspects of learning and inference at the edge all over the place. This presents an incredible opportunity for the combined company. When you think about intelligent applications and a real-time representation, digital representation of your business, people, places and things in real-time, you need a network that is so fast to be able to do that. So reliable and that's a new layer of value of application value that's going to be developed. Well, I can't agree more and I think this is ultimately the opportunity that we can pursue better as one company. Of course, we've got to get through the regulatory process that's going to take a bit of time but I'm really excited about the potential of this. So if we accept the premise that you're a head on AI for networking because you've been doing it longer than anybody, do you feel that gives you an advantage in the networking for AI though? And we were kind of joking before the thing started that you need AI for networking to do networking for AI because there's a lot of moving parts and so are you able to take your strength in the AI for networking and be able to take a leadership position in networking for AI? So Zia, I think it's a great question and I actually think it's the combination of the strength and the leadership we have in AI for networking which absolutely is applicable to building the AI cluster data centers because ultimately you want them to be optimized of self-running just as any other network you want it to be. But also- The risk of doing that wrong is huge, right? It becomes astronomical when it comes to learning and inference. Add to that our pedigree in high performance networking our silicon development capabilities our Juno's operating system the features that we've developed to mimic the capabilities of InfiniBand that combination is incredibly powerful. Now potentially in the future include things like the slingshot technology that comes from HPE's HBC high performance computing background I mean things get really exciting. The HBC AI combination is going to be super computing with networking it's going to be kind of the holy grail for this next gen. Question that comes up that I want to get your thoughts on you mentioned hybrid well we talked about hybrid and public private networks across tracking like cloud we'll see a lot of hybrid stuff going on in telco markets and other networking. The comments that's coming on the internet is things like statements like partnerships remain critical for adoption of AI solutions. Part the partner ecosystem is developing with these solutions from silicon you got to pick your clusters foundation models are emerging that has specialty around them. What is your view on how partnerships are going to work in networking and in this area because you need to connect things. You got to have open gateways big discussion they're having as one example. What do you see changing or evolving in how partnerships company to company, company to customer integration. What are some of the hotspots and the opportunities? Yeah I mean if I understand the question I think what people are asking for are partnerships that make that give them the comfort level and the peace of mind that when we offer an end to end solution that delivers on all aspects of AI networking compute storage networking the software the automation the visibility the peace of mind that this thing is just going to work. Look either you're going to need to have all of the components in the house that you can assemble together and pre-test or you're going to have to have strong partnerships that will allow you to assemble these components and to test them to give your customers the confidence that it's just going to work. And on the developer side how do you see the developer market emerging? AI native networking for developers like a DevOps or integrating into the application where the network policies could be implemented. What's the developer angle on this? I mean I've pretty much any market is ripe for AI solutions addressing. Whether it be the developer market or any other market for that matter. You know I do believe the opportunity is just massive and the benefits of artificial intelligence is pretty much profound across the board. Given your background in Silicon how do you see what's going on in Silicon? How do you see that playing out? I mean it's video is two years out. You see their results growing like crazy super high gross margin, et cetera, et cetera. Everybody else trying to get into that business. So it seems to be a gate right now to AI acceleration. How do you see that playing out? Is it as much of a gate as I believe it is or we believe it is and how do you see that progressing? I mean obviously it's hard not to see what NVIDIA has done and be in absolute awe and respect for the technologies that they have developed and the market share that they have captured. Having said that I think they're going to have more and more competition down the road. Pretty much all the hyperscalers are developing their own Silicon. Of course you've got AMD, you've got Intel that are coming into the mix. There are a lot of players that are going to come in and basically compete for that GPU opportunity. On the networking side, I can say something very similar here. I believe part of what Juniper brings to the table that provides us with significant differentiation is our ability to choose the best of breed, merchant Silicon where it makes sense, but then also custom Silicon with specific hooks that can be very valuable for AI cluster networking that ultimately build these end-to-end data centers. Silicon is going to play an incredibly important role here. It wasn't that long ago that this industry stopped believing in Silicon. We felt that software defined is everything, that the hardware is meaning. Well, White Box is going to rule the world, right? And honestly, it feels good to see that people have started to wake up and to realize just how important Silicon is to the ultimate end-state solutions that we need to offer to our customers. Is there a lesson though from the success of Nvidia though, where if you look, they have good Silicon, but what they've really done is changed the Silicon industry where you need to deliver the software. You were asking about developers, that developers can use to interact with the Silicon versus here's your chips figure out to do with it. Look, like I said, Nvidia has figured out the recipe and you're absolutely right. It's not just about the Silicon technology, but it's a software that enables the end users, the developers to tap into that Silicon. Yeah, I don't know, that's how you've been building boxes for a long time. Well, premise is that Silicon is not going to be the gate to AI acceleration. You believe that barrier goes away in the near to midterm? Is that a fair assessment or? Well, hold on, I believe that we can never take a step back or relax and say, okay, the Silicon is here. We will need to continue to innovate and iterate with new process nodes and architectural advantages and so on to stay ahead of the game. Actually, at Supercomputing 23, we're just at the rise of what we call clustered systems. I mean, the clustered system is the definition, but you're starting to see the trend of clusters being built around combination of chips, like pairing chips with this software for workloads that like, say, GPU clouds are emerging. That is clearly becoming a Supercomputing like architecture. Let's put some clusters together. So I think we're in an era now of clustered systems as a category of building clusters and connecting with NICs and networks. Absolutely. Absolutely, which, by the way, is why it is so important to think about the Silicon certainly for the performance aspects, but then also the software that sits across all of the various different components, compute storage networking, and gives you the confidence that this is actually working well as a single system, a single Supercomputer, if you will. So we're at a whole other revolution. It's a systems revolution. Systems mindset is the new normal, clustered systems. Things that the pendulum goes back and forth. Five years from now we're going to say we want to disaggregate everything again. It sounds like we need a new operating system to run all this. Robbie, thanks for coming on us. Thanks so much, Zia, for your time. Dave, guys, we're live here in Barcelona, bringing you all the action on theCUBE. Four days of wall-to-wall live coverage only on theCUBE. We'll be right back after this short break.