Good morning. How's everyone doing? I'm excited to be here today, you know, for three specific reasons. One is I think we're living in a very special time where the kinds of problems that we're going to be able to solve because of AI are going to actually surprise us in a very, very positive way over the course of the years. Number two, I'm really excited because the class of innovation and the tempo with which we've, with the team, has innovated for you is something that I'm super excited to share with you. In fact, we have so much innovation that we had to break this session into two parts: one that we're going to do today, one that we're going to do tomorrow. So, really, really excited to share that with you. And number three is none of this really matters if it doesn't resonate for you. And what I think is going to happen is I'm hoping that we'll deliver some technology to you that allows you all to do your life's best work for your customers. And so I'm really, really excited to see if we can bring that to you.
So with that, what I thought I'd do is before we go into the details around the innovation that we've had and the payload that we have, I thought I'd just zoom out and take a step back and say what's happening in the market right now. Because we're in a pivotal moment in the market where we're moving from this era of chatbots that intelligently answered questions for us to agents that are like digital co-workers that are going to conduct tasks and jobs for us almost fully autonomously. And we are right around the corner because we're not stopping there. We're right around the corner, like Chuck mentioned, also in this era of physical AI where robotics and sensors are going to be actually sprinkled all around us. And these shifts are actually creating infrastructure requirements that are changing in pretty substantial ways.
And if you just take a look at the patterns of consumption of infrastructure right now, the infrastructure demand patterns are fundamentally different even from chatbots to agents. Chatbots were very spiky because they were largely human-led. A human asks a question to a chatbot, the chatbot responds back with an answer. And so the patterns of infrastructure consumption were very spiky. When you go to agents, agents are working 7 by 24. And so there's a much more persistent, sustained demand signal for the infrastructure where it's consuming things at a very, very kind of consistent pace. But there's more than that. These agents are also operating at machine speed. And so what you're starting to see is the sheer volume of infrastructure that's going to be needed is meaningfully higher than what we've had in the past. And the way I like to think about this is humans click but agents swarm, you know. And so this demand that we're seeing, interestingly enough, is not just confined to data centers. It's also confined to the workplace. In fact, what you're starting to see as a pattern right now that's really interesting is you have these Mac minis and Mac Studios that people will have and they'll be sitting right by their desk. In fact, there's a whole new category emerging of deskside computing, right? Where you have these Mac minis that are sitting by your desk and what you're essentially doing is having hundreds of agents that are on that Mac Mini that are starting to work on your behalf. And what this allows you to do is solve problems with these agents who become teammates for you as digital co-workers. And it's not just one Mac Mini or one deskside computer. It's actually thousands of deskside computers because everyone's going to have one of these by their desk at some point in time. And as you have tens of thousands of these in your workforce, what's that going to mean? It means that you're going to have trillions of agents that are going to be proliferated everywhere that are working with each other. They're going to be working for us and that are going to be making sure that they make decisions on our behalf once we start feeling comfortable that we can delegate work to them.
Now, what's really interesting about these agents is not just the sheer volume of these agents. What you're also starting to see is these agents are far more consumptive in their patterns of network bandwidth that's going to be needed. So on average, we just recently did a study, every agent generates about 450% more traffic than a human for conducting that same task. So what you're going to start to see is the demand signal is going to be much, much higher. And infrastructure is no longer just something that's bound by compute and memory. It is actually very significantly network-bound. And so as a result, what we're seeing right now is we are entering into a massive what we call networking super cycle. And this networking super cycle is not just upgrades of infrastructure in your data center which is actually happening at multi-trillion dollar scale. And that's going to continue for the foreseeable future. But you're also going to start seeing those refreshes of infrastructure and reconfiguration of infrastructure happening in the workplace and happening with service providers.
So what we need to make sure that we're doing is getting our organizations effectively prepared that as you have this massive upsurge of agents that are going to be conducting and automating workflows that we have the right infrastructure that we can provide our business users so that they don't actually get constrained by infrastructure to go out and get the full potential harnessed from AI that they need. Now this super cycle that we talk about isn't just limited to networking. In fact, we need to fundamentally reimagine security as well. Now, why is that? If you think about the agentic era, what happens? A user delegates a task to an agent. That agent then might actually talk to a bunch of other agents. It might be a local agent. It might be a third-party agent. Those agents then will make sure that they're talking to an MCP server. They might talk to a skill. They might talk to CLI and access tools because one of the big differences between a chatbot and an agent is agents have tools access. They use tools on our behalf to get jobs done. And when you start looking at this entire architecture, it creates a whole new class of attacks.
Now, it turns out that Cisco is at the center of addressing both of these infrastructure scaling requirements as well as security requirements for AI. And this has put us in an incredible position to drive three very, very specific and strategic outcomes for you that Chuck touched on briefly. So I'm just going to dig into that a little bit more, which is the first one is making sure that we can help you in building out AI-ready data centers because that's where the digital workers live and so the apparatus that you're going to need to make sure that those data centers are operating at the speed and throughput and reliability and security and resilience that's needed that we make sure that we can make that real. Number two, we want to make sure that we can future-proof the workplaces because those workplaces are going to be where the human workers are, but you're also going to have agents working in the workplace. And that's going to require an uplift of the infrastructure in the workplace to be agent-ready, to be AI-ready. And by the way, these human workers need to talk to these agents within the data center. And so you're going to need that secure global connectivity fabric.
And lastly, all of this is going to be extremely important that it's done in a way that's resilient from an infrastructure standpoint. What we call digital resilience, which is the ability for an organization when they experience an outage, when they experience a breach, to be able to make sure that they can stand back up right away and not spend an enormous amount of time figuring out what actually happened, but in fact compress the time to investigate, respond and remediate as quickly as possible and get to what we need to get to so that the business can be back again and operational. And all of this we want to make sure that we accelerate with AI. But here's the thing. When you think about the agentic era, it comes down to this one thing which is every agentic action is going to be a routing challenge. It's going to be a trust decision and it's going to be a telemetry event. And this is the very reason that we're focused on actually building out this vertically integrated platform that actually starts with the foundation of silicon and optics. And on top of that, what we have is our systems that actually drive the networking switches, routers, access points. We have compute, we have data, we have the models, and then we have applications and agents. And it's wrapped around with security and observability so that you have full visibility on your estate and you can make sure that you can enforce policy so that you can remain safe. This co-designed full stack is something that we've worked very hard to go out and build because when you start doing piece parts it's actually pretty hard to integrate. And so when you start thinking about building a platform, we wanted to make sure that we built a platform from the physics, the silicon to the semantics, the meaning that actually is driven by the agents.
Now as we built these platforms, one of the things that had to be kept in mind is the core characteristics we wanted to have in the platform to focus on you. And so the first one was we can't be arrogant enough to think that the only technology that you're going to use within your organization is Cisco. In fact, you're going to have a whole slew of other technologies and vendors within your organization. And so we need to make sure that we have an open ecosystem. We also need to make sure that the technologies that you buy from Cisco are tightly integrated but loosely coupled. You don't have to buy everything all at once from Cisco, but boy, when you do buy two things together, they should work like magic. And third is we wanted to make sure that we could reduce your marginal cost of ingestion of new technologies that you purchase from Cisco. It should get easier and easier and easier as you buy more things from Cisco to make it easier to ingest and it should add compounding value to you which means it's not just adding value for what we've delivered it also adds value for everything else that you have from Cisco that starts getting compounding value.
So today what I wanted to do was I wanted to focus on three specific areas where we are going to be helping out all of you in bringing technology to bear. And those three areas are one, making sure that we can scale network infrastructure for the agentic era for you. That's number one. Number two is we want to make sure that we can secure your agentic workforce. And number three is we want to make sure that we can fundamentally reimagine operations and really drive a whole lot of simplicity in there. So let's start with scaling networking for the agentic era. Now as I talked about earlier, inferencing is happening everywhere and at the core of this is Silicon One. We've actually, we make our own networking silicon on top of which that's the foundation on top of which everything else gets built. And so let's take a look at what we've built.
So, you know, thank you. About 10 years ago, and I take no credit for this decision. I wasn't even here. 10 years ago, Chuck and Tain made a strategic decision which was, hey, we need to make sure that we get into building our own silicon. And what was set out as a vision was we need to make sure that we have a common architecture across all our chips that go across all of our devices. They have to be high performance and there has to be programmability that's infused in there. So not every single time where you need to go out and change something or reconfigure something, do you need to do a new tape out of a chip? You can just have programmability be something. And through software you can make some of those changes. Today I'm delighted to share with you that we've actually got a full family for a continuum of use cases for both data center and the workplace. In fact, I'm holding two chips over here in my hand. This is the G300 chip, which we'll talk about a little bit later. And this is the P200 chip. These two chips, this one is a scale-out chip and this one is a scale-across chip. And we'll talk about these a little bit later, but these are the foundation of the things that we do to make sure that we can move AI traffic reliably, quickly, and securely, you know, within the world. So I'll just leave it over here. Don't let me forget them. They're a few thousand bucks.
So let's start by talking about what these things do in the data center and then we'll talk about what they do in the workplace. So in the data center, when you start thinking about the data center, Cisco is foundational to the world's data center buildouts. There are, you know, four to five trillion dollars worth of buildouts that are happening right now. There's constant capex that's being raised. In fact, we are very fortunate that we actually serve the largest of the largest hyperscalers. We serve neoclouds. We serve sovereign clouds. We serve service providers. And the fact that we serve all of those four entities at very large volumes and scales really helps us make sure that we can bring some of those innovations to the enterprise so that you all can start having the right level of technology built for your data centers as well. Now one of the key problems that the industry is focused on is this notion of how do you scale up these data centers and there's three patterns that have emerged. So the first pattern is when you actually have scale up, what they call scale-up networking where you might have racks of servers that are together and that's what you scale up. The second pattern is when a rack itself which has lots of GPUs might not be sufficient enough to go out and fulfill the workload. And so what you have is multiple racks that get clustered together within a data center. But now what you're also starting to see is because of power constraints because you can only have 80 to 100,000 GPUs typically in a data center. What you're starting to see is this pattern of scale across data centers where two data centers might be hundreds of kilometers apart and they actually start operating like one giant computer and all of these things require very, very sophisticated networking at high volumes. And to address these needs, Cisco has been extremely focused on ultra-high performance scale-out and scale-across solutions.
And like I mentioned earlier, here's what we are starting to do with delivering scale-out solutions. And I'm going to start with what the chip is and then we'll go into the systems because the chip is the G300 chip that I showed you earlier. This chip is a revolutionary chip because it has 102.4 terabits of Ethernet switching that's happening at it. It's a successor to the G200 chip which had 51 terabits of speed. So this is twice the speed, right? It's a 3-nanometer process node. And here's the crazy part. That chip actually has a quarter trillion transistors in it. 246 billion transistors on this 100 square millimeter package that you see the package substrate. So it's a tremendous piece of, you know, computer science problem that's been solved. And it's basically the G300 puts us in the heart of some of the largest AI clusters in the world. And literally, you can't get more performant switch chips in the world today than this one, right? Yep. Thank you.
And so that chip then gets placed into systems. And so we will have both air-cooled and liquid-cooled variants of systems with the G300 chip. We have the N9300 system which is actually focused on enterprises and we have the Cisco 8100 system which is focused on hyperscalers, right? And we also have new optics. These are 1.6 TB pluggable optics and 800 gig linear pluggable optics that make sure that, you know, we no longer need just copper. You have photonics that are actually communicating packets at very, very high speeds across clusters within a data center which is pretty amazing to see. So that's how we're innovating for scale-out.
Now, here's what we're doing to innovate for scale-across. For scale-across, just like you had the G300 chip, we've actually built the P200 chip. The P200 chip is a 51.2 TB chip which can actually process 20 billion packets per second, but it has this specific and special capability which is deep buffering. And what deep buffering allows you to do is imagine two data centers that need to act like one logical computer that are 800 km apart from each other. When you send a packet over, you need to make sure that if there's packet loss, you don't have to restart your training run. So what this chip does is actually has deep buffering capability in it. And it powers new systems that are both 51.2 TB fixed systems. These are the pizza boxes that can get racked, or it powers 28.8 TB modular systems through line cards, right? And basically this is a chip that is a routing chip. So we have this 8223 fixed router that both runs Sonic operating system as well as IOS XR and it's got both 400 gig and 800 gig pluggable optics. Now, why am I giving you all this detail? Because these pieces of infrastructure and data centers are really complicated to build. And the GPU is great, but the GPU is no good if it's not networked. And what we are doing is not just networking the GPU within the data center, we are now starting to network data centers across from each other so that they can act as one logical unit. So that's what the scale-out solutions are for the agentic era. Now, we're also partnering with Nvidia on a number of fronts. And so, we've actually got, we started with the Cisco Secure AI factory with Nvidia, which is a reference architecture. We now have our switches that are fully integrated in the NVIDIA Spectrum X architecture. And what this means is that our switches work with Nvidia's NVLink to deliver low latency and intra-cluster communication at very, very fast speeds, right? And then what we also have is Nvidia silicon that also gets used in Cisco's chips. So, one of the things we want to do is we want to make sure that we continue to stay very open in the market and we want to have the full stack that's well integrated, but we also want to use what else is there in the market from others. And so, that's some of what we're doing in data centers.
But, as we discussed earlier, it's not just about data centers. We want to make sure that we fundamentally think about reimagining the workplace for networking as well. And so in the workplace, I'm going to take a break for a sip of water. There's a lot over here. I didn't realize my mouth will be so dry. Chuck, you got a joke. Thank you. All right. So, essentially, we need to make sure that we can fundamentally reimagine the workplace as well. And this is something that's not been that obvious to people so far. Most people thought that AI would be largely, you know, conducted in the data centers but what you're now starting to see is the traffic patterns are going to start increasing in the workplace as well. So over the last year we've refreshed our entire campus branch network. This is the largest refresh we've had in Cisco's history. Everything from switches to routers to secure firewalls to Wi-Fi access points even industrial devices. We've made sure that we've kind of innovated across that entire portfolio of products. And what have we done? We've done a few things. One, we unified hardware. So, one of the things that was always asked was, you know, rather than having different brands that were actually hard to manage, what we now have is a unified brand of hardware between Catalyst and Meraki. We have unified software for managing these and we've got unified management.
And today, we're continuing that momentum. And so what you'll see over here is we will have new smart switches. We'll have new routers. We're going to have new premium outdoor Wi-Fi systems. But the one that I'm particularly excited about specifically in the era that we're going into is a Cisco 9550. This is the most powerful core switch we've ever created. Now, why is it important? Because there's two classes of switches and there's a lot of emphasis that's placed on access switches. And we've actually built a bunch of access switches. But the core becomes the backbone. And for agentic workloads, when you actually need to make sure that you have high volumes of workloads, if you don't have a strong backbone, then that becomes the bottleneck of your network. And the 9550 is one of those most powerful kind of switches that we've built for the core. And all of this is designed to be quantum-safe.
Now what we want to do is make sure that as we upgrade this infrastructure it's not just about upgrading the infrastructure we also want to keep innovating so that your infrastructure is protected in this new AI-powered world of attacks, right? Because Chuck talked about this, Mythos is a true phenomenon. In fact, we are living in what we like to think about as a post-Mythos world where the security must be actually fused and baked into the network. And models like Mythos really show how bad actors can actually compress the time from the time that you announce a vulnerability to the time that an exploit happens is now a matter of minutes if not seconds. So you have to make sure that there's a systematic way in society to address this challenge. And so how do we help address this challenge? There's two specific areas that I'm going to give you as examples of what we're doing to address the challenge. One of them is of course the basics are going to happen. We will make sure that if there's patching that needs to be done, we will provide, you know, kind of very regular patches so that you all can patch that. But that's a very tactical problem. What we need to also do is make sure that we can have some more strategic areas that we can work on with Mythos. So the first area is this notion of vulnerability shielding. And what that does is because the patch cycle is going to get more and more frequent, but sometimes you might not have the ability to patch as quickly as you'd like to, we need to make sure that we provide you mechanisms of what you do in between patch cycles. And so we now have this product called Live Protect where you can now have compensating controls that can be applied in between patch cycles. We'll talk a lot more about that tomorrow. And then the second area is this notion of exploit containment. And what I mean by exploit containment is you want to make sure that you can stop lateral movement through your hybrid mesh firewall within your estate because you have to assume at this point that the bad actors have already infiltrated your network. But what you want to do is prevent lateral movement and intercept it whenever there's actually behavior that's starting to seem anomalous. And hybrid mesh firewall gives you this kind of ability to sprinkle enforcement points for security across your entire fabric of the network. And you'll see, you know, demos of this later tomorrow.
So, what are we doing basically to help you scale and protect your networks with AI? The way to think about this is not only do you need to make sure that you have the scaling of the network with Agentic AI, but you also need to make sure that you secure your Agentic workforce. One of the biggest challenges people have with agents right now is they don't trust them. And if you don't trust them, you're not going to delegate work to them to use them. And so, what do we need to do to make sure that we feel good about trusting them? And the reality is these agents are going to be everywhere. And so there's three things that we feel need to happen to secure these agents. The first one is we have to make sure that we protect these agents from the outside world. There are going to be bad actors. There are going to be adversaries that'll try to have prompt injection attacks on agents that'll try to make sure that they can data poison the agents so that they can alter the agents' behavior. We want to make sure that we don't have that happen. That's number one. Number two, we want to make sure that we can protect the world from agents because these agents will oftentimes start behaving in ways that you don't want them to behave. And so, you want to make sure that when that happens, you can intercept and give them guardrails. The way I think about agents is they're kind of like teenagers. You know, they're supremely intelligent, but they have no fear of consequence. And sometimes they do stupid stuff and so you need to make sure that you can protect the world from them. And lastly, you want to make sure that you can detect and respond at machine speed.
So let's start talking about how do we start protecting agents from the world. So when we start talking about protecting agents from the world, think about this as a brand new attack vector. And so you have to make sure that there's both safety and security that's paramount. And about a year and a half ago, we launched a product called AI Defense. Now AI Defense was a product that was built so that you could actually make sure that these nondeterministic models that we all use these days, right? They aren't predictable. You ask the same question twice, you're going to get a slightly different answer. These non-deterministic models, how do you make sure that you have visibility into what data is going into those models? Is the model behaving the way that you want it to behave? And when you can get the model to not behave the way that you want it to behave, are you able to intercept that at runtime? And so we built this product called AI Defense. And it was largely built for models and apps.
Now, as we've moved into the agent world, we wanted to make sure that AI Defense could also be something that is used for agents. And so what we've done is we've added a whole basket of capabilities around adaptive testing and guardrails, around security for agentic supply chains, and around making sure that we have support for any agent platform whether it be Claude or Codex or OpenClaw so that your estate is safe and you can protect the world from the models, you can protect the world from agents. And so that's kind of what we actually started with by adding these capabilities. But here's the reality. We've actually made sure that these agents as they're being deployed in multiple different ways that we start to make sure that we can provide you with tooling so you can protect yourself and your organization from this. And so what we've done is we've developed capabilities to help builders everywhere in the world to safely create and deploy agents on any of these platforms, right? And so a couple months ago, what we did was we launched this product called Defense Claw. By the way, completely open-sourced. You can go out and download it today if you want. It has a bunch of services that can allow you to make sure that you can scan for your skills. You can scan your MCP server. It'll hook into either OpenShell, which runs OpenClaw, it'll hook into Claude Code, or it'll hook into Codex. And it fully integrates into Splunk Cloud and AI Defense.
And this was something that we announced at RSA a couple months ago. And what we've seen is a massive uptick within the market of the use of these services in open source. And so you're starting to see that the open-source tooling that we're giving out to our customers and to the builders and developers is starting to gain traction. So we are going to continue to keep investing in open source and you should rest assured that Cisco is very committed to the open source movement and we will continue to make sure that we keep investing in open source. Thank you.
And so we've made huge amount of investments in the open source community, but what we wanted to do was we also wanted to make it easy for you to use these technologies because not everyone might want to download open source versions within their organization. And so what we've also done is we're making sure that these capabilities will also be part of Cisco Secure Client and that's going to be available in July.
So now let's talk about protecting the world from agents. So essentially when you start thinking about protecting the world from agents, what you're trying to do here is make sure that when the agents start drifting, when they start behaving in ways that you don't want them to behave, you can figure out and detect that they're drifting and intercept right away. Right? And so this notion of zero trust which is least privileged access that you give to humans. Now what is zero trust? You give humans the least amount of permissioning so that they can get the job done. You don't over-permission users with rights that they can do whatever they want. Right? That zero trust for humans has always been focused on you have to prove your identity and we have to make sure that we can then verify that we trust you as a user and then allow you to do the things that we need you to do. That's the whole basis of zero trust which is access control. As you move into the agentic world, we have to move from just plain access control to action control. And that action control is actually something that is a very, very difficult thing to do because you have to make sure that you intercept every action and verify that the behavior that the agent is exuding is in fact aligned with the intent that you had of what that agent should be exuding as behavior. Right? And then you need to be able to control that agent. So what we've done with zero trust is we've moved from access control to action control. And the question is what needs to be done to make this happen? And there's three things that need to be done to make this happen. Number one, you have to know the identity of every agent within your organization that's running. Number two, you need to make sure that you can authorize every action of every agent that's running. And number three, we have to be able to adapt to risk that might emerge from the behavior of agents in near real time.
And so what we've done is we've actually focused an enormous amount over the course of the past year in ensuring that we've got all kinds of capabilities that we've made available to you for agent security and everything from discovering the agent, everything from being able to safely delegate work to an agent to having the right level of MCP gateways and LLM gateways so that you have some way to intercept a request before it goes to these different tools. All of those pieces are starting to come into play. Now, a big area for this that's foundational in agentic security is also the identity of the agent and specifically identity of the non-human identity. So, a few weeks ago, we made an announcement that we're going to acquire, we're going to have an intent to acquire a company called Asterisk. And Asterisk is a pioneer and invented non-human identity. And you should expect us to be very committed to making sure that every machine, every service, every agent within your organization will have an identity apparatus from Cisco that we will be able to provide to you. We're really excited about this acquisition. It's not closed yet, but we're very excited about the possibilities at the lab. So, with that, that's what we're doing basically to protect the world from agents.
Third is what do we do to make sure that we operate at machine speed and scale? And when you really start thinking about this deeply, here's the challenge at hand. You know, data is the key for defending at machine speed and scale. We have a lot of telemetry within our organizations. We have a lot of data and that's something that we have to understand well to be able to detect because what you need to do is you need to understand baseline behavior. You need to detect abnormalities and then you need to make sure you connect the dots. And that's why we recently announced Cisco Data Fabric. What is the Data Fabric? It's basically a common data platform that's powered by Splunk that handles data and signals at massive scale, petabyte scale. From networks and security as data sources, from applications as data sources, from third parties as data sources. And what we'll be able to do is drive action in the organization from this flow of telemetry that we have by correlating that telemetry. Now, one of the examples of how we're going to use this Cisco Data Fabric, it's going to be a foundation for several core capabilities. One of those is going to be the SOC.
A lot of you over here are security professionals. In fact, a lot of times people worry about, you know, in the security world, the short staffing that is there with security analysts. It gets to be very difficult sometimes for the security analysts to really figure out how they're going to be able to keep up with the inundated demand that they have because the signals are just increasing for them and they don't have enough capacity to go out and make sure that they can observe every single signal. And it's a state that's getting worse. There's about 4 million jobs in cyber security that go unfilled every year.
single year in the United States alone. And the reality is, you just don't have the luxury to investigate every single alert. So you have to make trade-offs. And think about how crazy that is that you have to make trade-offs on alerts that could be a breach within your organization because you can't address all of them.
The question we have to ask ourselves is: what if we could go out and investigate every alert that comes in? And the reality is, we're now starting to get to that place where we believe that we will have the ability to analyze every single alert. And that is going to be a reality that's within reach.
And the way that it's within reach is because your SOCs are going to get agentic. They're going to be powered by this data platform. And you're not going to need to have humans. You will have these agents that will autonomously be able to identify anomalies, that'll be able to identify breaches that are about to happen, and over time predict that something might happen and prevent it before it even starts. Right?
But security is not the only challenge when it comes to responding at machine scale. So you will hear a lot about the agentic SOC tomorrow. But in addition to security, you also need to have observability that needs to be done at machine scale. And when I talk about observability, you need to understand how agents are behaving. You need to understand not just the performance of the infrastructure and the resilience of it from an observability standpoint. You need to understand the quality of the outputs that these agents are producing.
And you need to have observability for the entire stack. And so what I mean by observability for the entire stack, and this is really important, it's making sure that you look at how is your infrastructure performing? Is the utilization of the GPUs at the right level? Is the infrastructure resilient? But you also need to see how your model is performing. You also need to see how your application runtime is performing. And then most importantly, are there agent evals? How are those performing as well? And what I mean by this is you will have to assess for resilience of your infrastructure, but you'll also have to assess for the behavior of agents. Are those agents behaving the way you want them to behave? And when they're not behaving the way you want them to behave, what do you do?
But there's a third problem that's starting to emerge right now. And the third problem is, I don't know how many of you have been told by your management right now, 'Hey, these tokens are getting too expensive, ease up a bit.' And the reason for that is because what's starting to happen is what you might have budgeted for your token capacity for a year, an agent going rogue could actually go out and consume that within a matter of a week.
And so you have to make sure that this notion of tokenomics, efficiently being able to go out and use tokens so that you can generate positive business outcome, is going to be superbly important for you. And so we've actually got this entire stack that we are building out for observability capabilities. And so essentially what you'll see over here is we're supercharging Splunk agent observability for that entire stack. So you will be able to go out and look at the resilience of your infrastructure, assess agent behavior, and make sure that you can gauge the tokenomics within your organization of whether or not you're using tokens in the right way. Is it too many tokens? Do you need to ease it up a bit? Do you need to kill some agents because they've started going nuts? You want to make sure that you kind of ease it up a bit. So those are the things that we'll be doing.
Now, in the quest of this kind of mission to make sure that we can provide full visibility and observability, we also acquired a company recently called Galileo. And Galileo essentially is an AI observability and evaluation company for a multi-agent system. In fact, we have Tin and Yash over here who are the founders of Galileo. The third founder, Vikram, stand up for a second. [applause] So these gentlemen and their co-founder Vikram, who couldn't make it here today, thank you, were actually working with Google and with DeepMind, and they worked with the DeepMind team to make sure that they could actually start thinking about how AI was going to be evolving. And so we're delighted to have them on our team, and Galileo is going to be at the core of our observability offering for agents and assessing agent behavior as we move forward. So folks, that wraps up what I wanted to cover from an agent security and observability standpoint.
Next, and probably the most important, is how do we make sure that we fundamentally reimagine operations? Because this is what you all live on every single day. And one of the areas that we've gotten a lot of feedback from you over time is, 'Hey, I need to make sure that you can fundamentally help me simplify operations.' Now, there were three design constructs that we had kept in mind for reimagining operations. The first one was we needed to make sure we gave you simplicity but without taking away the sophistication of the Cisco portfolio, because Cisco has a lot of richness in its capabilities. We didn't want to take that away, but we wanted to give you the simplicity for managing it effectively. Number two, we wanted to make sure that you could operate at agent speed, real time, and where you don't have to go out and linearly scale your capacity as your organization scales, because your budget is not growing proportionately with the growth of the company. And three, we need to make sure we leverage an open ecosystem, because there are many, many players in the market that you need to work with.
And so I'm here to tell you that this is all now a reality. And I'm delighted. This is probably one of the best pieces of work that I've seen a team do in a long, long time. And that is Cisco Cloud Control is now a reality. [applause and cheering] And for those of you that don't know, here's what Cloud Control is about. Cloud Control basically is at its core simplicity without losing the sophistication of Cisco. And so what we've tried to do is say all the products that you know from Cisco and love, literally all the products that you know from Cisco and love, will be managed from Cisco Cloud Control. I thought you guys would get wrapped up about this. This is pretty cool.
And so what this allows you to do is it basically allows you to operate at agent speed, right? This is going to be available, by the way, in controlled availability in the United States today. We've already got customers that are using this. You'll hear from a customer a little bit later, and we'll show you a demo. But it's now in the US, and we'll actually take it out globally to the rest of the world as well. So this one, it allows you to operate with simplicity without losing sophistication. Number two, it allows you to operate at agent speed. What does that mean? It means that you will be able to make sure that for things like troubleshooting, for things like debugging, you will be able to have cross-domain telemetry that can be correlated.
What I mean by cross-domain telemetry is if you've got a configuration issue, or let's say you have a user issue and you want to know where that issue is originating from, you will be able to correlate telemetry from the network and security and come up with an insight by looking at those two different domains. Second, we will have purpose-built models that we have built. We've built models for networking. We call it the Deep Network Model. We've built models for security. We call it the Foundation Security Model. And we've built models for time series data, which is the Time Series Model. And so we've built purpose-built models, and we are investing very heavily in AI research to make sure that networking becomes a fully automated, AI-based kind of system. And third, we're going to have these trusted agents. And the operative word is trusted agents that'll be able to go out, monitor the entire network, and be able to be working on your behalf.
But the reality is these agents are Cisco agents, and we know that you might want agents sometimes from other vendors. And one of the big pieces of feedback we got with Cloud Control was, 'Hey, this is great that you're doing it for your systems. What about other systems?' So here's what we did. We said, why not open up Cloud Control to the entire ecosystem? [applause] Right? And so we have on day one 52 partners, and I hope that this continues to keep growing very, very exponentially. And they are going to be building apps within Cisco Cloud Control. Right? You will have a marketplace, and you will be able to make sure that you can publish applications that are built within Cisco Cloud Control. And these are not just companies. You yourself might want to go out and build an app in Cisco Cloud Control, and you'll be able to do it.
So the best way to understand Cisco Cloud Control is not by me talking about it, but by me showing you a demo. And so what we thought we'd do is we'd show you a demo with five key capabilities. By the way, there's thousands of capabilities. I don't have enough time to show it to you both in today's and tomorrow's session, but there's one thing you should keep in mind. Every single demo you see at Cisco Live will originate from Cisco Cloud Control. How bonkers is that? The entire portfolio will start from Cisco Cloud Control, right? This would seem like fantasy if you just talked about it to someone just a few years ago. And by the way, here are the five capabilities we'll share with you. Number one, we're going to show you unified management across the entire estate. Number two, we'll show you how troubleshooting works across domains. Number three, we'll talk to you about connecting fabrics between data center and campus. Think about an EVPN and VXLAN fabric. We can actually, with one click, now configure that entire fabric. Number four, we have secure agents that'll be there within the enterprise that you'll be able to secure. You'll be able to secure these agents anywhere in the enterprise right from within Cloud Control. And number five, we'll be able to monitor and optimize token usage so that you don't get surprised with the budget, because you can always monitor what's going on. And so to show you this, I'm going to have my good friend and the person that actually runs the Cisco Cloud Control team and our platform and AI, DJ Sath, come up on stage, and he's going to give us a demo. Give him a big round of applause. [music]
All right, DJ. So, I've built you up over here, buddy. So, what are you going to show us?
