3339: How Cisco Is Preparing for a World Powered by Agentic AI and Quantum Computing

[00:00:05] Where does the future of computing intersect with breakthroughs in quantum networking and agentic AI? Well today I reconnect with my good friend Vijal Pandey from Cisco. He's the driving force behind OutShift at Cisco. And we recorded our conversation today live at Cisco Live. And OutShift serves as Cisco's internal incubator. And pushes the boundaries far beyond the company's traditional core.

[00:00:32] And bravely tackles ideas that carry both technological and market risk. But today he's going to be sharing how his team balanced that risk. By nurturing early stage concepts from ideation to real custom attraction. So today we dive into Cisco's progress with quantum networking. Including the recent unveiling of the quantum network entanglement chip. And the new quantum labs in Santa Monica.

[00:00:59] All which are building blocks for a secure, scalable quantum internet. And he'll also talk about Agentic AI, Cisco's vision for an open internet of agents. And how global collaboration partners are shaping a new era of machine-driven, human-guided automation. And with quantum teleportation, eavesdrop-proof security and coordination at the speed of entanglement. These ideas might sound like science fiction.

[00:01:27] But they are already becoming a reality. But which of these advancements could rewrite how your organization thinks about computing and secure collaboration? Well it's time to get my guest on now. So a massive warm welcome back to the podcast. Can you just remind the listeners with a little of who you are and what you do? So first of all, thank you Neil for having me. So I am VJoy. I run this group called OutShift within Cisco.

[00:01:56] Which is an internal incubator. And it's an incubation engine for Cisco. So the whole idea behind OutShift is to think about products that are adjacent to Cisco's core businesses. So these are personas, both buyer and user personas, that are adjacent to who we talked to today. And can we build some products and solve their pain points? And the other thing that we look at is products that are a bit further out and a bit riskier for Cisco right now.

[00:02:23] So it could be technology risk, the tech is not yet mature, or it could be market risk. The tech is mature but we are still figuring out whether there's a market, how big is the market, is the market stable? And are people willing to pay for using it? And so those are the kinds of spaces we get into. And we go all the way from ideation to customer traction. And once it starts getting customer traction, we work with the BU's to figure out how to scale it within the BU.

[00:02:52] Oh, wow. So you're in the exciting part. Well, it's exciting and it's heart wrenching because one of the things that I keep reminding the team is that 99% of what we do will not see the light of day. Oh, so it's an emotional tax, but that 1% that succeeds is really breakthrough. That's the game changing stuff. And of course, it's been 12 months since our last conversation at Cisco Live. So how is the quantum network envision at Cisco?

[00:03:18] How has that evolved during that time since we last spoke and what's driving acceleration now? So on the quantum side, what I would say is we've announced a few pretty neat things. And I'll get to that in a second. But the thesis behind what we do at Cisco in quantum is we are a networking company and a security company, first and foremost. So when it comes to quantum, we are building the quantum network and a secure quantum network.

[00:03:46] So quantum safe, quantum networking is where we play. And the vision is that while we see many companies out there building quantum compute nodes, our job in this space is to connect these compute nodes so that we get distributed quantum computing. And by doing so, we achieve scale out and we can pull in pragmatic, practical quantum computing by five to 10 years.

[00:04:14] Because if you think about even the classical world, you have servers and you can build bigger and bigger servers, but that's not what people do. We connect these servers through a network, a classical Cisco network, and you build these distributed computing environments within an on-prem site or you build a cloud. And both are distributed computing environments. And the network plays a pretty critical role in enabling that distributed system.

[00:04:41] And so we are building out the same philosophy on the quantum compute side. And by doing so, we can pull in the size and scale that matters for practical quantum computing, which is you're looking at hundred thousands to a million qubits of compute power. And right now we are at 200, 250. So the only way you can achieve that with speed is to connect these smaller compute nodes through a network.

[00:05:09] And here at Cisco Live, it was recently unveiled that the quantum network entanglement chip. So what does that milestone actually represent for the industry and how close are we to scalable quantum networks? It all feels incredibly exciting. We've been talking about it for years, but we're getting closer, right? We're getting closer. And so I think one of the things that is clear is the rate of innovation is actually becoming exponential in the quantum space.

[00:05:35] And we saw a similar rate happen in the AI space. I mean, if you think about neural networks, they were announced or they were invented 1960s. And you used to see innovations happen in decades. And then it went to years and then quarters and then months. And now it feels like it's every day there's something new happening in the AI space.

[00:06:03] There's something similar happening on the quantum side where you we started working as an industry on quantum a long, long, long time ago. But you're seeing that rate of the pace of innovation actually pick up and becoming exponential. And you're actually seeing like Microsoft and Google and Amazon and IBM and us, all of us announcing breakthroughs in the quantum space pretty rapidly. So I would say the timeline is shrinking.

[00:06:30] And just like AGI or artificial general intelligence, there's no strict and hard prediction that somebody can make. Similarly, on the quantum side, I can't give you an answer that, hey, we'll have pragmatic computing by this time frame. But if you see some of the roadmaps from, let's say, IBM published one just two days ago that talks about their fault tolerant quantum computing hitting hundreds of thousands of qubits by 2029.

[00:07:01] And we can bring that in even closer by connecting those compute islands through a network. So that's our job and that's what we're trying to do. So I think it's pretty close where we can get pragmatic quantum computing. But having said that, what is truly exciting is that the foundational technologies that we're building for quantum networking, which is the chip that we just talked about. And I'll get into the details there.

[00:07:27] But that chip, the network protocols that we need to redo for the quantum world and then the software controllers that we need to build for the quantum computing, quantum networking domain. All of these foundational technologies have use cases in the classical world today.

[00:07:45] And so whatever your belief system is and whatever you think quantum practical quantum computing is at, like five years out, 10 years out, you have use cases for the quantum network today in the classical world. And that's what's truly exciting. I mentioned the chip a moment ago and the new Cisco quantum labs in Santa Monica. Collectively, these things signal a serious investment.

[00:08:11] So tell me more about that chip, the kind of research and collaboration, everything that's being prioritized here. Yeah, so we just announced the Santa Monica lab and people asked me why Santa Monica. And so Southern California is at the nexus of all quantum research and development in the US, at least. So you've got Santa Barbara where a lot of work happens, UCLA, Caltech, all of these universities are here.

[00:08:35] And then also the labs from Amazon and Microsoft and Google and Cisco sitting in Southern California. So there's a hotbed of talent and then there's a hotbed of research in Southern California. And what we're doing in the lab is building out the components for a secure quantum network. And so these are, again, the chip, the protocols, the software controllers and the compilers that are needed to enable distributed quantum computing.

[00:09:03] So the first step in that journey is this chip that we announced, which is the quantum network entanglement chip. And what that does very simply is it creates entangled photons at a really high rate. So 200 million entangled photons a second. And so, yeah, so it was a big deal. I mean, why do you create these entangled photons? So what you do is you send one of the entangled photons to, let's say, you, Neil, and one with me sits with me.

[00:09:33] And using these entangled photons, we leverage a property called quantum teleportation. And we can exchange state information between you and me. And by doing that, you being a compute node, a quantum compute node, and Vjoy being another quantum compute node are now at the same quantum state. And so we operate as a singular logical entity when it comes to quantum computing.

[00:10:00] So we are actually building a network using entangled photons and using this method of teleportation to exchange information. And that way, we are enabling distributed quantum computing. And so this chip is actually generating these photons at a high rate over telecom frequencies. That means you can use existing fiber. And it operates at room temperature. So you don't need these big cryo fridges to enable this.

[00:10:27] And this is the first building block, and we are super excited about this one. I have to add that this is a research prototype, and this is not available at scale yet. You had me at quantum teleportation, if I'm honest with you. But as we look to the future, everyone's making predictions around AGI, quantum computing. But if we just come back to the present day for a moment, the big talking point here at Cisco Live is around agentic and agentic AI.

[00:10:54] Tell me a little bit more about some of those announcements and what excites you. Yeah, so I think agentic, of course, is a here and now problem. But also agentic is a decade-long problem as well. I mean, it's here and now. You have some capabilities today. But to truly make agentic work, you need to have these agents which behave like humans.

[00:11:19] So they have human-like attributes and capabilities, but they operate at machine speed and scale. So they are a blend between human ways of thinking and probabilistic inputs and outputs and communication. But they are operating at machine speed and scale. So they are really, really, the throughputs are really high. So that combination is really powerful, but it's also really scary.

[00:11:44] So you need to make sure that you can bring them together as subject matter experts, help them collaborate in a trusted environment, evaluate them so that you know that they're doing the right job and they're doing it the right way. And that's when you get magic. And so what we are doing within OutShift and with the larger Cisco force behind us is we are building out this notion of an Internet of Agents.

[00:12:09] And what that is, is a platform or a framework for agents from diverse vendors sitting on diverse clouds with diverse identity and authorization buckets, different profiles. They can all come together on this platform. They can be discovered. They can have identity and authorization. They can have reputation. They can communicate. And then they have observability and evaluation built in.

[00:12:37] And through this, this platform allows them to communicate in a trusted environment and solve for a larger business score, which is what you want to get towards. So that is a big deal for us. So the Internet of Agents is one place where we're focusing. And we believe that just like the original Internet, this new Internet of Agents needs to be open and interoperable. So which means what is open and interoperable in the modern day and age? It's not standards only. It's open specifications.

[00:13:06] It's open source. And it's free services that are available to everybody. So we formed an open source collective. We call it Agency. It's A-G-N-T-C-Y. So agency.org. You can go there. You can download code. You can look at documentation. But it's 55 plus partners working together to build out the Internet of Agents in an open source, open spec, open services manner.

[00:13:34] And what excites you about that project, that collaboration and all this talent coming together to work on this? Yeah, so it's truly exciting because everybody comes at it from different viewpoints. And one of the things that we believe in is that if you want to truly build the Internet of Agents correctly, you need vendors and you need developers and you need operators and you need consumer customers and B2B customers.

[00:14:01] You need the scientific community because foundation models are going to be heavily used for science and mathematical evolution as well. You need people building robotics and physical embodied AI. And so you need all of these people coming together. You need folks that are building social avatars, for example.

[00:14:20] So all of these people from all of these spheres need to come together and build out an Internet that works for operators, developers, consumers and vendors. And so what is truly exciting is that we've built up that pretty diverse crowd of contributors and adopters and developers in agency. And they're looking at problems from different angles.

[00:14:45] So when we're solving for their needs, we're solving for the community as a whole. And before sitting down with me today, you've just come from on stage where you were talking about agentic AI amongst other things. What were you talking about there? What kind of questions did you receive? And also what kind of feedback are you getting from customers around agentic right now? So I think agentic is a new tool. It's where we are moving from deterministic computing to probabilistic computing.

[00:15:14] And that's a new tool that can be leveraged pretty heavily and solve for new problems that you could not do in the past. The one journey that we all have to go through is make sure that we don't retrofit the new tool in the old world.

[00:15:33] And if you remember where cloud was maybe 10, 15 years ago, people were taking their on-prem infrastructure and just doing a carbon copy of that infrastructure on the cloud. That's a great place to start. But that's not a great place to end at because you're not leveraging the cloud in the right way. And so when we talk about what these capabilities look like with probabilistic computing, that's what's very exciting.

[00:15:59] But you need to also leverage that probabilistic computing in a good way. And so the Internet of Agents actually helps you leverage that probabilistic computing in a good way, in a trusted, evaluated, safe manner. And so what we're hearing from customers is truly excited. We are throwing some use cases at them because so what? So we are throwing some networking use cases. You saw AI Canvas being talked about in the keynote by DJ and team. You've seen AI Defense in the past.

[00:16:30] So there are both products that help you deploy agents in your environment. But also we are showing use cases, real practical use cases in the SRE space. We have an open source project called SRE Jarvis. We are working with our service providers, partners, and customers to build our network configuration validation pipeline.

[00:16:51] They're very valid use cases that we are showing out there that show you how to build multi-agentic apps and how you can leverage Internet of Agents components to build those apps out. So there's excitement, there's potential, and there are real-world use cases. And between those two, there's imagination, and then you need to solve for those problems that you have.

[00:17:12] I think the biggest gap that exists today is there is a lot of potential, but then making it work in the enterprise is where Cisco can come in and help you. As someone that's watched far too many Spider-Man movies, with great power comes great responsibility. There's a big focus here on the importance of human in the loop when talking about technology.

[00:17:34] And I know you've previously drawn parallels between the early days of classical computing and where quantum stands now, and maybe even agentic too. So what lessons do you think we can learn from the past before going forward? So one of the things that with technology, I would say there are certain step function changes that happen in the technology landscape every few decades.

[00:18:00] And the Internet was one such change, which actually changed how all of us do business and how do we interact with businesses. So the entire e-business paradigm occurred after the Internet was formed. With cloud computing, we had another step function change where we had mobile apps and we had social networks, and how we interacted with each other actually pivoted quite a bit after the cloud transformation.

[00:18:28] This is probably bigger than any one of those combined. And so agentic AI is where everything that we do, the way humans work, is going to change. And how we interact is going to change. So like we already discussed scientific discovery and physical work through robotics, and of course B2B and how we interact through social avatars. I mean, all of that is going to change. So we need to embrace it, and we should not resist it because it's something that is going to happen,

[00:18:58] and it's accelerating what you do. Whether it's physical work, B2B, scientific, social, it is going to accelerate what you do. So leverage it, make it work, and if there are problems around trust or observability or evaluation or discoverability or communication, let folks like us help you overcome those challenges. And we can show you through real-world use cases how we've overcome those in the products that we are building. So we are here for the journey together.

[00:19:27] Let's make it happen. But there are a few step function changes that happen, and you need to ride that curve. And the same thing is going to happen with quantum. I think we all had our chat GPT moment. We are going to have a chat GPT moment for quantum. And this one is probably going to be bigger than anything else that we've seen, even in Gentic, because this is a transformation happening all the way from the bit level. So you're going from bits to qubits with our quantum bits.

[00:19:57] So everything from computing, from networking, from software, from algorithms, from computer science theory, everything is going to change. So this is something that you need to start preparing for right away. So if you're thinking about it, please come and talk to us. The big theme here seems to be about being open and open to collaboration as well. So how would you say Cisco's vendor agnostic approach to quantum network, how does that benefit the broader ecosystem,

[00:20:26] especially companies that are building different types of quantum processes, etc? Yeah, I mean, we are, I mean, our mission is to power an inclusive future for all. And I think if you think about we are in the secure connectivity space, we are in the collaboration space, it goes back to that mission statement, which is we are in the business of connecting diversity.

[00:20:49] And if you think about even bare bones internet and where it all started, we connected different clouds, different organizations with different policies, and we made e-business happen. We are not the only ones, but we are big participants in that journey. With agentic AI and with the Internet of Agents, we are doing something similar, where you're going to see agents from different vendors, different clouds, different profiles.

[00:21:18] And there's going to be a lot of heterogeneity and diversity in that ecosystem. And we want to build a collaboration platform that brings all of those things together. The same thing applies for quantum computing, where even today you are seeing eight plus ten, maybe physics-based technologies on the quantum computing side, transform, trap-tion, all of these technologies, photonic.

[00:21:43] And so the way we are building out the quantum fabric or the quantum network is it's agnostic to all of those technologies because truly, and you can ask any one of those vendors as well, we don't know which tech is going to win yet. And maybe there will be a world where different technologies are suited for different kinds of problems. And so you'll have lots of technologies coexist on the quantum side. And so the way we are building out the fabric is we are pushing that heterogeneity at the edge.

[00:22:13] And then we are transforming that to a photonic fabric using quantum teleportation, the magic of quantum teleportation, that leverages existing fiber. So you can leverage existing infrastructure, but push out that complexity of quantum computing to the edge. And that's always been our goal. And that's what we are going to push for. And if we look beyond quantum applications, you've also mentioned advantages for classical systems right here, right now.

[00:22:40] So what would you say are some of the short-term enterprise use cases that you see gaining traction? Yeah. So this is, I mean, this is where we believe that quantum networking use cases are here and now and today. And so they're happening right now. And we're working with a whole set of design partners on some of these problems. I can't talk about those names yet because we are still proving the entire stack out, but I'll give you some classes of problems that we can solve. So the simplest one to understand is decision coordination.

[00:23:09] So let's say you had a starting gun that you would want everybody on planet Earth to hear at the same time, to act on some piece of information. So Neil, you might be sitting in Tokyo. I might be sitting in San Jose and we are maybe starting a gaming session.

[00:23:34] And there should be fairness because if you start even a picosecond before me, maybe you get to see things that I don't get to see. And so how can I ensure that there is fairness between you and me when we start playing this game? And through this means of quantum teleportation, we can signal a state change, a starting gun between you and me instantaneously, literally instantaneously. There's no speed of light hurdle to cross.

[00:24:01] That way we can both start our gaming session at the same time. Now, a more concrete use case is in, let's say, in high frequency trading, where you might have a data center in New York, another one in New Jersey or Pittsburgh. And even at those distances, the speed of light signaling adds a certain amount of delay and a cost to fairness.

[00:24:26] And so using this notion of entangled photons through an entanglement chip and teleportation, you can signal both of those data centers to act on an algorithm or a piece of data at the same time instantaneously. So that's a pretty easy use case to understand. And you don't need quantum computers to do this. You can just use this chip and a little bit of software that we're working on. But it's possible today.

[00:24:53] The other one, which is actually pretty intriguing, is figuring out if somebody is eavesdropping in your network communications. So we're calling that eavesdropper proof security. And the way we do that is, let's say you and I are talking to each other over a fiber link. And there's an intruder who's eavesdropping in between. How can you provably tell Neil, you or me, Vijoy, that somebody is snooping in?

[00:25:21] And so what you do is, again, you send entangled photons between the two of us. And if there is an intruder in between, they measure that photonic traffic between us. And the moment you measure that entangled state, and if you go back to quantum physics, the cat is dead or alive, or dead and alive at the same time. The moment you open that box, the Schrodinger's cat box, you know whether the cat is dead or alive. And so this intruder in some ways opens that box.

[00:25:51] And it gets to know that the cat is dead or alive, which means you and I know instantaneously that somebody is intruding in our communication. And this you can layer on on any kind of communication that is happening between us. It could be standard TLS, HTTPS. It could be IPsec that happens today. It could be the NIST compliant post-quantum cryptography software standards. It could be anything that you might be using, but I can still tell you that somebody is intruding in our communication.

[00:26:21] So that is, again, pretty exciting. And finally, if I was to ask you to do a bit of crystal ball gazing here, we look into the future. We've talked a lot about agentic, about quantum. How do you or what excites you about these two worlds converging in the future? So the way I think any piece of technology gets used in the technology stack itself is that we create a blend of all of these capabilities, because all of these technologies are good at something or the other.

[00:26:51] So let's take the current world of CPUs and GPUs or accelerators in general. Any supercomputer out there is a combination of CPUs, GPUs, DPUs, all sitting within one logical environment. And based on the needs of the algorithm that you're running, some parts of that run on the CPU. Some parts get accelerated through a GPU or a DPU.

[00:27:17] And so to us, what we think about, and especially because we are building the network that connects all of this, QPUs or quantum processing units are going to be yet another capability that we add to the equation. So the supercomputers or the on-prem data centers of the future or the cloud data centers of the future are going to be a combination of CPUs, classical accelerators like GPUs and

[00:27:43] QPUs sitting in a combined environment, in a distributed environment through a network from Cisco. And your algorithm will actually figure out what needs to be sent to CPUs because it's good enough, what needs to be accelerated through a GPU because it's a little bit more expensive, but it's specialized. And then what needs to be sent to a QPU because it can solve these really hard problems that take exponential time in the classical world.

[00:28:11] But through qubits, you can bring it down to more measurable outcomes in timelines that you and I care about. So that combination is where the world is headed. And that's the combination that computing will be for all of us in the future. Well, thank you so much for sitting down with me today. You've just come from being on stage. You've now completed a podcast. You've got to dash again in a moment. Just thank you for talking with me again.

[00:28:37] Hopefully we can meet next year and further explore the world of quantum teleportation. That seems like a topic for another day, but thank you for joining. Thank you for having me. Absolutely loved unpacking Cisco's dual focus on quantum networking and agentic AI. And for me, it makes it clear that we're standing on the edge of this next big leap in how networks, machines and humans all work together. Whether it be fostering a thriving open source ecosystem to building an infrastructure that

[00:29:07] embraces diverse quantum technologies. I think Cisco's vendor agnostic approach positions the broader industry to innovate freely and securely. And short-term gains like entanglement-based security and fair decision coordination all hint at what's achievable even before full-scale quantum computing arrives. So as we look ahead, I think blending CPUs, GPUs and QPUs in unified distributed architectures

[00:29:35] all promises new frontiers for solving problems that were once thought impossible. And as this pace of discovery accelerates, the challenge for every organisation will be knowing how to harness these tools without losing sight of trust, governance and collaboration. But which part of this vision will your team be exploring first? As always, email me, techblogwriteroutlook.com, LinkedIn, X, Instagram, just at Neil C. Hughes.

[00:30:06] I'll be back again soon with another guest and another topic where we will explore together. But that's it for today. Bye for now.