Morning, everybody. It's great to see everybody once again. It's always best to see everybody live and in person. Welcome to what is our biggest Fal.Con event ever, and we've never been more excited to host you with so many of our customers here, so many of our partners here. Hopefully, you've had a chance to talk to a few of them. Hopefully, you had a chance to catch some incredible keynotes from George, and Mike, and others. It's been a fantastic start. We've got a really, really exciting lineup for you today. I'm going to go over the quick agenda, and hopefully, you're able to leave here with the same feeling that I have when I came here and how I'm feeling today about the opportunity in front of us. George is going to come up first. He's going to talk about our next chapter, securing AI. We're going to then have a customer fireside chat with DB. We'll have a customer and partner fireside chat with DB. Then we're going to have a little break for lunch, quick break, maybe about 25 minutes. Then we're going to hear from Mike, the winning platform for the AI era. Then you'll hear from me. I'll come and finish it off. Then we'll have a quick interlude, and then we'll go right into Q&A. I'm excited for today. I hope you guys get a lot out of today. We've extended the Q&A session. That was one of the part of the feedback that we got from last year, where you guys wanted more time with us. Not only are we going to give you more time for Q&A, but George, Mike, and myself are going to stay extra after so you can have one-on-one conversations with us. All right, so enjoy the day. Let's kick it all off. Please welcome Chief Executive Officer and founder of CrowdStrike, George Kurtz. All right. Okay, well, I can't believe what an event that we've been able to pull off. I think one of the premier security conferences now, and obviously we get to spend some time with the financial community. It's important to be able to explain what we're building, the innovation that we're driving at CrowdStrike, and more importantly, how we're solving customer problems. With that, I will jump into the presentation. As Burt said, we're excited for the fact that we've got more time to answer questions, which I know many of you want to get through. What is the state of AI? Some of this should be self-explanatory, but I at least wanted to go through what we've seen and some of the numbers. This McKinsey stat. If you look at organizations that used AI in at least one business unit, it was 55% in 2023, probably like in ChatGPT, and today it's 89%. ChatGPT messages are up 8x, and reasoning tokens, 320x. That's where we are today. We started with chatbots, then we moved to reasoning, now we moved to agents. That is driving, of course, the spend. I always talk about following the money. There's always this question, I know, in the financial community, is it a boom? Is it a bubble? What's happening? Is it going to last? Because we've been through fits and starts in AI for a long time, if you remember the cycles of AI and machine learning and those sort of things. I think where we are, to me, it's very clear that this is a mega trend that we've never seen before. It's the fastest growing transformation of spend in the enterprise history. Just look at the stats, $2.67 trillion, with a T, in 2026, going to $5.95 trillion in 2030. Again, a Gartner stat. For me, if we follow the money, it's all going to make sense in what that means in terms of the opportunity from a security perspective. We've seen this movie before, and we've seen it with the cloud adoption and the hyperscalers. What we saw over a period of time is the adoption of hyperscalers. I remember when I started CrowdStrike, we were a very early customer of AWS. I think they had five services at the time. Compare that to today. You can see the slope of the curve over time and the adoption of the cloud. There was a point, if you remember, in the not so distant past, where you would walk into a financial services company, maybe like yourselves, and you would say, "Hey, let's talk about the cloud." They go, "We're never going to the cloud. That's risky. We're not going to do that." Think about where we are in 2026. If you fast-forward to looking at cloud security, and I say this all the time, security parallels the slope of the curve, the technology curve. If we just take cloud as, again, a tectonic shift in technology, you can see cloud security actually paralleling the slope of that. Now, if we plot frontier labs, it's a much different curve, and it's something that I've never seen before. I don't think any of us in our lifetimes have seen growth as what we're seeing now in the frontier models. You can just look at the slope of their revenue and what they put out publicly, if you will, even though many of them are private. In general, you can see how fast the adoption is and how fast the spend is. Certainly, I think the adoption now for security is going to parallel that. What's important to realize, there we go. What's important to realize is that's okay, just keep it over there. Security's role is different this time in that security is now the accelerator, not the brake. In many of the technology shifts, it was like, well, security's got to get in the way, and it's going to slow us down. Now, you have to have security in order to go fast, and that's one of the defining changes that I've seen in security and AI. Adoption is a necessity. It's larger net new opportunity, and it's immediate and rapid spin. What I would say specific to this is, again, just sizing the opportunity. If it's $2.6 today, going to $3.6, and you took a simple example, just a simple one, now you can put your own filter on this, but if it was 1% of the AI spend going to 8% of the AI spend, it's a $36 billion to $291 billion opportunity for security. Let that sink in. Again, apply whatever filter you want on it. What I can tell you, it's big. I'm a simple math guy, and I look at this and I look at AI, and my simple math is, do I think there is going to be more AI in a year, in 3 years, in 5 years than there is today? If you can say yes to that, then the natural corollary is then by definition, you're going to have more security for all of that AI that's created. Where we are today, I was talking to Jensen in the green room. We were just talking about the future and where things are going and what we're seeing today, we're probably going to laugh in 5 years when we come back. We're going to remember these kind of simple concepts and what it was able to do. The more complexity that you have, the more security challenges that we're going to have to face. Massive opportunity, it's a sustained opportunity, and it's certainly not just a peak of what we're seeing. I've said this, you've heard me on the conference calls, we've talked about it on many of the one-on-ones. I see AI security being bigger than EDR. Obviously, when we started the company, we redefined the market. There wasn't even a term called EDR, but we redefined it, and we knew we built a big business, and obviously now have built a big platform company on EDR, and then obviously in additional technologies. When we think about the velocity of the initial launch, just to put it in perspective, I'm not going to move around a lot because of my mic here, so I'll just kind of stay here. If we think about the velocity, when we launched EDR, we saw a 10x increase 3 quarters from launch. Now, different time, it was a lot earlier in the life cycle. What we have seen with AI DR is a 79x increase in uptake just after 3 months. Okay, massive adoption. Yes, we are a bigger company. It is a different period of time. I like 79, and I can do math pretty quick, and to me, that is escape velocity. What is driving this? Why is it happening? AI has evolved. If we think about the non-agentic adversary that we have been fighting for many years when I started the company, dwell times were in the months. Expertise, you needed a nation state. You needed to have high expertise. The cost was high. The success framework was a 1/10/60 in terms of minutes. Now, the agentic adversary has really changed everything. Talking about seconds, minimal expertise, cost is negligible, and it is in real-time. The speed of inference, if you saw my keynote yesterday. It just dramatically shifted in a very short period of time, and this is what customers are facing. From nation state to agent state. I went through the pyramid of pain, as I call it, which is hacktivists at the bottom, e-crime in the middle, nation state at the top, but now it is the agent state. That agent state now is powering and enabling everyone to act as if they were a nation state and really flattening this curve of expertise. Whether it is a hacktivist or an e-crime group or an agent, they are all in the same category. They all have the same level of expertise. Let us talk about cyber's next frontier. Where is it actually going? For us, we made an announcement yesterday. We talked about the Cyber Security Superintelligence Lab. Why did we create this? It is important to realize where the world is. You have frontier labs, which have done a tremendous job. We partner with them. They are changing the world. Again, the success criteria is a little bit different. If you look at what they are focused on, building incredible models, selling lots of tokens, and solving lots of broad problems. Fantastic for society. CrowdStrike's Frontier Lab is really solely focused on solving today and future cybersecurity problems with frontier caliber technology from a model and a harness. That was part of our SafeMind announcement, which we are going to get to. What is important to realize here is where the technology has evolved, what we have done with NVIDIA, what we have been able to reproduce in our labs, and we will go through some of the stats. The capabilities are now being put back in the hands of the defenders, which was, again, a big part of my theme. The adversaries have frontier caliber AI, but the defenders did not. Let us jump into this a little bit more. Superintelligence Lab is totally focused on frontier AI research, open weight, open source model training. We have been big proponents of these technologies. Again, open source and open weight are actually two different things. If you look at Nemotron from NVIDIA as an example, it is actually open source and open weight. Some of the other models are simply just open weight. Use a combination of these is fine, but a lot of it comes down to the harness engineering. What Jensen said, the exoskeleton of what makes the model work, and we have seen incredible advances in the harness engineering that we have been able to build in the reinforced learning based upon the massive amount of data that we have, the team, the patents, the partners, and then obviously, a national security element to this. Why are we trusted with this data? Why are customers looking at us to solve some of these problems at this scale, at frontier model scale? We generate 7 trillion security events per day, which is a massive amount of information. We are a net creator of security data. You heard me a couple of quarters ago in the SaaS apocalypse saying, if you are a net creator of data, you are going to be very sticky. The data that we have is not available. You cannot go to Reddit and go find CrowdStrike's telemetry and security data. When we think about the training of these models and the outcomes that customers are looking for, it does start with the data. Cyber expertise, just to give you a few stats, 270 PhDs, 300 AI researchers, and 500 threat researchers. That does not count a whole bunch of other areas that we are focused on. The R&D investment over $1 billion. You see the ecosystem. I am not going to go through all the names, but we are in the right place with the right people and the right technology at the right time. Why does cybersecurity need its own model and harness? I think Hugging Face was a great example of why the defenders were outmatched. When you look at that example and you go through lessons learned, we got lucky that the model and the agents were totally focused on passing the test rather than causing havoc and damage. If it was a nation-state, it might have been a different outcome. When we look at the frontier models, they are really good at general knowledge, finding all the possibilities, but they can be very expensive. This very specialized model that we built, we will go through the red and the blue, it is cyber-specific mastery. This is very important. Our models are not designed to solve the world's complex math problems. They are designed to solve the world's most complex security problems. Precise, deterministic outcomes. Deterministic outcomes. This is very important. Everyone in this room has used a model somewhere and asked a question three times and got three different answers. That is not great in security. You have to have deterministic outcomes. What customers are looking for is cost efficiency and data sovereignty. Where is the data that customers care about from a security perspective? Last I checked, it is in the Falcon platform. You know what they do not want to do? They do not necessarily want to ship it out and pay for that, and two, have it in some other place that maybe they are not as comfortable as having it in the Falcon platform. These are very important elements of why customers. They literally mobbed us after the announcement yesterday on the models and when they can get access to them. That is why we announced cybersecurity's first frontier model and harness family combined as an agentic system, which is SafeMind. It is a family of models, open source, open weight, CrowdStrike data, and the CrowdStrike harness. That, again, we've taken our expertise, we've worked as well with NVIDIA, you heard from Jensen yesterday, on what we've been able to do together. There's a massive focus from NVIDIA and us to be able to take these models and to be able to create them in a way that they can be very specific in solving security use cases. We did a lot of engineering with them. We did a lot of work in the training, did a lot of post-training. We provided feedback in what we want in future models, and I think we're in a great position to continue to deliver an at-scale model with incredible performance and cost characteristics. There's also a routing layer. I talked yesterday about model choice, right? When we think about the models, sure, we have models, but there are other open source models. Fantastic. You have frontier models, right? You heard from OpenAI yesterday. You're going to hear from Anthropic today. At the end of the day, we want to use the best models, the best tools to get the best outcome for customers at the best cost. So there's a model routing layer. By the way, this opens up an opportunity for us to be able to provide technologies like Frontier Labs to our customers and be able to monetize that through the platform, which is really interesting, right? Because we've become a distribution path for the Frontier Labs as well. Trusted partner, trusted data sovereignty, again, allowing customers choice to get the best outcome, which is stopping the breach and providing automation and capabilities that they couldn't get from any human on the planet. So let's talk about the model. You've got Red Tempest, which is our red model. Of course, this is the model focused on finding vulnerabilities, finding exploits, finding ways to attack. So you need a red model. Part of what I talked about in the keynote was there's been a lot of work around finding vulnerabilities. Hey, I'm going to basically run this model over my source code. I'm going to find a ton of vulnerabilities. I can create some patches, but then what do I do? To be fair, the industry has been focused on finding all these vulnerabilities. If you understand code, great, you can find a lot of vulnerabilities in source code, but then what do you do with all that? So you have to have a red model that's focused on delivering security offensive capabilities at nation state level and find ways to attack. But at the same time, you must have a blue model. This is the area that I highlighted in the keynote, which was really a missing piece to me. If you think about all these models, they've been focused on the attack piece, the red piece, not the blue piece. What we've been able to do is to create a blue model that actually learns from the red model, that detects threats, stop breaches, and uses cyber tooling to stop the attacks. This is one of the areas, just think about the Hugging Face incident. When I told the story, I mean, it's out there, it's their story, I kind of repeated it yesterday, which was they didn't know what was going on. They had some signals. They saw all volumes of attacks, 17,600 attacks that they saw looked like a nation state. They tried to piece it all together. They went to the Frontier model, and it refused. Refusal rates. This is a problem. When we think about Blue Solano, these are the kind of technologies that organizations need to be able to, in real time, provide a view of what's happening, provide defenses. But more importantly, when you look at putting red and blue together, it's this constant learning loop. This is the unique piece. You're constantly learning from the red team. Why is this important for us? Scale matters in AI. I think we all know that. Scale matters in AI. If you have a massive amount of data, which we do, and you have a massive amount of telemetry, which you're getting every day, and you understand the attacks because you're seeing them, you have threat intelligence, you have an MDR service like Falcon Complete, you have this constant loop where you're always learning from the adversary. Those get fed into the model, and then very quickly we can run attacks either in a real environment, or I show the digital twin, where we can actually simulate an environment so that we can understand what the exploitability is. This is a very important term that the industry is focused on, exploitability, because you can't patch everything. We hit the sound barrier on patching. You need to know where the priority is, you need to know what controls are in place, and what risk is out there if you don't take an action. This very quickly, this is one use case. I mean, it's unlimited number of use cases. It's models in a harness. One use case that our customers went nuts for, because the digital twin aspect of recreating the environment. We actually know the environment. We have the agents running. We have the identities. We have the infrastructure, basically map of what's running. We know every system. We know what's on it. We know the versions. We know the vulnerabilities. In our digital twin, we can recreate that very quickly and then begin to create these automated loops of red and blue testing and then go back to our customers and say, "These are the things that are really important." Again, I'm giving you one example. It's an unlimited number of examples of what we can do, but this is what we actually wanted to show. Very excited about SafeMind. What does it all mean? These are the benchmark testings that we did. Let me just take you through. We have Frontier Lab One and Frontier Lab Two, and then we have Blue Solano. You can see Blue Solano had 37% better detection rates than Frontier Lab One, and 29% better than Frontier Lab Two, because it's very specific in our domain. But what was it trained on? It wasn't trained on a bunch of stuff on the internet. It was actually trained on some of the most valuable security and threat data in the industry, which is what we've accumulated over the last 15 years. So detection rates better. Number 2 is the cost per task. What this is saying here is you got Frontier Lab 1 and 2, and that Red Tempest was actually 66% less or 78% less per task. Then when you look on the blue side, it was up to 99% more cost effective. I'm going to let the numbers sink in, because in your travels, I'm sure that you're talking to customers, you're here, you ask anybody that you bump into, they will say, "Man, the cost is out of control." I'm trying to use these models, and it's amazing how expensive it is. What they want is they want the best outcome, using the best model, with data that is protected in some fashion, sovereign to them or with a trusted partner, and they want the lowest cost. That's what we're delivering. Better value. Now we look at detection speed. Well, last I checked in security, speed matters, right? Matters in racing and it matters in security. 2 things that I like. And we are 6 times faster with SafeMind detection. What I want to reinforce here is that, yes, you have better detection rates, better value, faster speed, but it isn't just the models, right? It isn't just a harness. It's the combination of all these put together in a system, which is constantly learning from a loop and creating this kind of reinforced learning element to it, red and blue, and then being able to have the flexibility across the massive data set that we have. This is how you achieve Frontier-level performance at a fraction of the cost. So really excited about that. Now, I couldn't go through this without, I'm going to try to get to this to avoid the questions that I'm going to get anyway. How are you going to price all this? What does it mean? Now, this is very important. So there will be initial token packs. So there will be CrowdStrike tokens. That means a different way to monetize the platform. I'll say that again. There's now token pricing which will be a different monetization stream that will be open to CrowdStrike. This is exciting. There will be token expansion packs where you can buy more tokens, and it's all supported by Falcon Flex. I talked about SafeMind as being the harness for our cyber models, but the commercial harness is Falcon Flex, and they go together hand in glove. Now, when we think about the tokens and working with customers, I can tell you what customers don't like. They don't like variability. They don't like runaway spend. They don't like surprises. Nobody does. And I think they're getting a lot of that with some of the other technologies that are out there. So what we are doing is really looking across the entire organization, figuring out how many tokens they need as a company, and then we can set bands of token usage, and they certainly can burst up from that. But again, we want to put them in a position where they can understand and predict their spend at the lowest cost. Again, very important compared to how they are consuming tokens in other areas from some of the other model providers. So we are excited about this piece. Again, it opens up a whole another monetization opportunity for CrowdStrike, which is really, again, one of the most exciting pieces for me. So when will it be available? Well, let us talk about where it is going to live. It will live in two places. One, it will live in the Falcon platform. Blue and red will power the modules that we have, and the harness will be available to be able to take advantage of the platform data. The second piece, then, we have got tremendous amount of interest from customers and governments who want to be able to use the models on their own, and we will allow that to happen. The way we are going to go work through that is in the QuiltWorks partner program. So remember, we started QuiltWorks, we did a lot of work around AI readiness using our technologies, but it was the setup for the trusted partner and access program that we have created here. So we are going to first work, if you want to just get access to the models directly and outside of the Falcon platform, you are going to be able to do that, but you need to be part of the QuiltWorks program. This will be in preview, and we will be rolling it out over the back half of the year. Again, you got to make sure that you have got the right partners, you got to put the right controls around it, and we want to get the right feedback. So we are excited about this, but we want to be very thoughtful about how it gets rolled out, and the folks that we are working with. So that is availability and monetization. Okay. Meet SafeMind, an autonomous agentic system to redefine cybersecurity as we know it. SafeMind is a complete set of state-of-the-art harnesses and models that exceed frontier level capability for cybersecurity applications. SafeMind excels at delivering the prevention, detection, and response life cycle at machine speed. It goes beyond surfacing findings all the way to taking autonomous cyber defense action. In one use case, an analyst activates SafeMind as a skill in an AI tool. Powering SafeMind are two new classes of models, Red Tempest and Blue Solano. The analyst loads data about their enterprise. This data can be anything, asset inventories, identity stores, adversary intelligence, threat graphs, Falcon telemetry, everything needed for cyber operations. SafeMind's best-in-class models both build and deploy a complex clone of your enterprise, matching hosts, topology, operating systems, applications. Let us harden our defenses by loading an attacker scenario, unleashing Red Tempest. The model immediately assesses multiple paths to achieve its goal, autonomously succeeding in data exfiltration. SafeMind captures the traces from Red Tempest, and Blue Solano takes over. A defender model. Beyond frontier level, operating in a CrowdStrike harness, Blue Solano learns, identifies holes, validates new detections, and deploys them immediately. With the new detections in place, let's see how Red Tempest does. The new defenses written by Blue Solano stopped Red Tempest in its tracks, and it's forced to adapt and find a new path. The loop repeats again, and again, and again, until there isn't a path left for Red Tempest. The Blue Solano model outsmarted the attack, learning from Red Tempest while creating new remediations and defenses. Detection coverage grows, raising the cost for an actual adversary dramatically. But what you saw doesn't just happen once. SafeMind gives every defender beyond frontier capabilities, as Blue Solano is built and evolves with Red Tempest, ensuring the best defense is always aware of the best offense. Use Red Tempest for the best offensive security and to identify risk. Use Blue Solano for your defense. Use them independently or combine them to ensure your defense stays ahead of elite offense. SafeMind goes beyond generic frontier models for cybersecurity, turning findings into autonomous action. The defender advantage starts now. CrowdStrike, delivering the first frontier AI models purpose-built for cybersecurity. All right. Well, for me, again, seeing this technology in action, there's a lot more that we haven't exposed. It's been an incredible journey. Again, I think when we look at the next act of CrowdStrike, to actually be able to operate a frontier-type model specific to cybersecurity and the ability to actually monetize that through token flows is exciting for me. Again, I think the sky is the future when we think about how many agents are going to be available and how much the environment will actually change over time, and how important these models will be to our customers. We'll talk a little bit about securing AI and agents. As I mentioned in the keynote, AI is the new attack surface, right? We originally started the company protecting computers and people. Now it's about protecting agents. This is a stat where today, on average, or evolving, that they believe there will be 90 agents per person, per employee. The reality is that's today. We're in the near distant future. If we think about agents themselves, it's almost unlimited. It's almost unbounded. Why should you have 90 and not 1,000 or 10,000, right? Who knows what it's going to be? I don't know, but I sure know it's going to be more than 90. We have the ability then to tap into a TAM that really doesn't have an upper bound. You're not talking about just cloud workloads. You're not talking about just systems. You're not talking about just employees. You're talking about all the agents that get created from now in perpetuity that we have the ability to actually protect. For me, that massively expands our TAM and part of the reason why we pioneered AIDR. AIDR, AI detection response, is a new category. The product is Guardian, but the category is AIDR, just like EDR was a category. AIDR is focused on visibility discovery. Where are all my agents? One of the biggest challenges that you are going to find when you talk to customers is they do not know where AI is taking place. They do not know where it is actually being consumed. First you have to find it, understand what shadow AI is. Who is hiding Claude somewhere? Who is hiding this? Who is running that? We can do that. Then you have to have AI control and guardrails, and then certainly you have to be able to protect the prompts. Is it a malicious prompt? What is it doing? What is it sending out? Mike did a great demo this morning on that. Guardian AI Agent Security is something that we have worked on. It was around the clock that we worked on this with some of our biggest partners, some of our design partners. Our biggest design partner on this was Amazon. You heard from CJ this morning if you were in the keynotes. This is something we have been working on at scale to meet the most demanding enterprise, and we launched this during my keynote. We flipped it live, and it was operational, and the results that we have seen so far are incredible. Let us talk about Guardian and the agentic security lifecycle. An agent gets created, well, you need to know where it is. It accesses data, and it has permissions. It has an identity, and it has permissions and entitlements. It then assigns tasks, it executes, and at some point it gets decommissioned. It could be a short-running agent or it could be a long-running agent. But you have to be able to instrument all of this, and Guardian covers every stage of the agent lifecycle and provides visibility and traceability on exactly what it is doing. Think about a financial institution. You cannot go back to a regulator and go, "Well, I do not know what happened. It was some AI thing." You cannot do that, right? You have to be able to know exactly what happened, what it did, what it touched, what identity it was, and have full traceability. Obviously important from a security perspective, but it is going to be mandated from a compliance perspective. There is no way it is not going to be something that every company that is regulated, public company in a certain industry, is not going to have to have. It will be a must-have technology. Number 2, what does it actually do? It discovers the agent. It controls and observes the agent runtime. Runtime. This is something you have heard from us, what, 15 years? The agents run, and they are commanded by a prompt, and then it detects and stops threats. One of the cool things that we have done is we have now combined the prompt layer into the runtime layer. Mike's demo this morning talked about putting a prompt in that would have just gotten through sort of anything because it looked like a benign prompt. But unless you are actually looking at the runtime and understanding what is happening, you are going to miss it. We've taken our historical heritage and expertise over 15 years, and we've combined that with prompt visibility and essentially our agent, sometimes down to the chip layer, I talked about Intel, but our agent being able to have full visibility into the operating system, the prompt, the application, maybe a browser, maybe it's an Electron app, which is how AI applications are consumed like Claude or ChatGPT, all the way out using identity, to where that agent is actually talking to. What does the future look like? The future of endpoint, and I say endpoint and I'll say workloads here, is cloud and endpoint. I've talked a lot about endpoint, but the reality is we have a lot of customers actually running this in their cloud workloads. I don't want you to miss this point. It covers endpoints. You're on your Claude, you're on your Codex, covers you, no problem, on all the other applications. You do have customers that are now running these agents in cloud containers, ephemeral containers. We actually instrument those, too. We work in both environments. Oh, by the way, there's a third one called SaaS agents. We instrument those, too. You have three areas. My focus has been on endpoint, just to kind of make the point, but we're covering all three of those areas. That's a very important piece that I want to make sure that you don't miss. We can protect these agents wherever they run, endpoint, cloud, or SaaS environment. You combine that with identity, next-gen identity, which is our non-human identity for agents, zero standing privileges, only getting the entitlements you need at the time of execution, and then they're gone. This is critically important and creates a control plane from an identity perspective. Obviously, Guardian wraps this together with visibility, protection, control. This is really the future of what endpoint and cloud workload protection looks like. All right. Falcon Flex, we're going to go through these. Obviously, accelerating our adoption. Almost $2.3 billion in total Flex value, 101% growth. Last quarter, 935 new Flex opportunities added. You can see the growth there. It's been remarkable, the success we've had with it. Again, I call this the commercial harness. So much so that everyone in the industry is copying Flex. They don't even bother to change the name of it. Each Flex is not created equal, and I think we've done a good job of leveraging this with our customers. The other thing that I'll mention, too, is just by having something you call Flex doesn't mean it's operationalized. We invented it, and we've been at it the longest. What that means is we know how to sell it internally. We've enabled our partners, we've enabled the third-party ecosystems, like the marketplaces, to be able to deal with this, and you have to have the systems that actually work. There is a barrier to entry because you have to make all that work before you can just say you have Flex. This is why we've driven the friction out of this model. What's our next chapter? I think we're in an incredible time period in technology. I couldn't be more excited about just the technology changes that are coming, and we've got the right technology at the right time. Right? Cyber is at the top of the list of everything that we're doing in terms of AI, and you're not going to be able to accelerate the adoption of AI without cybersecurity. When Jensen and I had our first call on building out the models probably six months ago, he said, "George, this is the most important thing I'm working on with you. We've got to solve security." Because he knows if we don't solve security, it's just going to hold back AI adoption. This is very important. Right technology, right time, and the right execution. We've got an ecosystem. Look, I'm blown away when I come to this conference. This is now an industry conference. Look at all the partners that are here. You will talk to them. They will tell you this is the most valuable conference they come to, just from a customer perspective. By the way, we've opened it up. We have competitors here. It's been an open conference. It's been incredible. You have to have the right ecosystem, and you have to be around long enough to realize how important ecosystem is. If you're just Johnny-come-lately, you may not know how important an ecosystem is, and we've cultivated that over the last 15 years. Right go-to-market with Falcon Flex, as I said, the right time. Security is how AI scales, and this is one of the reasons why we've been able to leverage the right technology with the right commercial harness to solve problems that haven't been solved in the past. For me, I really think it's an inflection point for CrowdStrike. You can see the momentum, which we demonstrated last quarter. You can see the demand environment is not a spike. It's sustained, and I think that's important because a spike is just buying more of something. Sustained is solving a different problem. Solving a different problem, and the problem is AI security, and that's a long-tail problem that is going to be plenty of opportunity to solve. With that, I want to thank everybody here. Obviously, we're going to be up for questions, and I look forward to chatting with you soon. Thanks so much. Please welcome Chief Business Officer at CrowdStrike, Daniel Bernard, and enterprise go-to-market cybersecurity at Anthropic, Ash Alhashim. Ash, pleasure. Always a pleasure. to have you with us today, and good afternoon, everyone, both in the room and those watching. Wow, what a great conference, and how cool is it to hear what we just heard from George and to have the leading frontier lab, Anthropic, here on stage with us here at this conference? Everybody sit back and get ready for some interesting, good conversation. I think the feedback is consistent that everybody enjoys these sessions. We're going to make this interactive, and it's an honor to have Ash, who leads the cybersecurity go-to-market at Anthropic, with me up here today. Ash, I want to start our discussion first focused on Anthropic as a CrowdStrike customer. Congratulations. You guys won a big award today. Talk a little bit about that journey from when we started working with a company nobody had ever heard before, playing around with AI a number of years ago. Yeah. First of all, thank you for having me. It is a pleasure to be here. I joined Anthropic a little less than three years ago, and back then we had been working with CrowdStrike as long as I can remember. It was a very different kind of company. It was an AI research lab, very focused on safety. We did not have the household name that we do today. A big part of when you think about our mission, and you think about what we are trying to do and what is at stake, we want to usher in transformative AI in a way that benefits humanity, in a way that is safe and helps the world. In order to do that, you got to go fast. You have to have smart people, you have to have a lot of things, but you also got to just move incredibly fast. One of the principles I think that guides a lot of our decision-making is building with the best building blocks. Reinventing the wheel in an area like security that goes so deep just does not make sense. Partnering with the best partners does make sense. You guys are essentially the operating system for how we secure our company and what I can say is we use Falcon across the business. The benefits of the endpoint security, all the threat intelligence and the telemetry that comes off of that, allows us to be able to scale really, really effectively, while remaining incredibly secure. Cloud, next-gen SIEM, you guys really use the full stack. I think one of the questions that I heard a lot earlier in the year, and maybe it is one of the questions that is on everybody's minds here, is why did not you guys just Claude Code your way to CrowdStrike? When Frontier Labs really came out, and it came out with some really cool harnesses and the technology really evolved, that was one of the questions of the day. What better way to answer that question than actually ask the source? Yeah. First of all, I am legally obligated to say Claude Code is amazing and everyone should use it for everything. At the end of the day, in all seriousness, this is a very effective developer productivity tool. You can build really, really cool things and one-shot some really cool apps with Claude Security. Again, going back to we use CrowdStrike ourselves to secure our business. This is not something that a Claude Code or even a team of very talented engineers working at a company like Anthropic can just build overnight. This takes time, skill, distribution, network effects, a level of determinism that comes with experience. These are hard-fought things that you have learned over decades effectively. You cannot build a product like this. A lot of this is just inherent to the time it takes. So, when we think about what CrowdStrike is for us, it is the operating system for security. Claude Code, again, while useful in building some things, we are not there. The data is also something that I think we talk about, too. Maybe your perspectives on what we do over there. Absolutely. Again, going off the telemetry that you all have, the data that you have gotten. This is proprietary data. George was talking about how these models are trained. Obviously, there is a lot of inputs that go into these things. A lot of it is public knowledge and working on helping these models get better at reasoning over time. You cannot reason your way into building something like a CrowdStrike. This is something that, again, takes a lot of industry skill, expertise, and proprietary data that just cannot be amassed in other ways. Perfect. I think that is really a clear way to capture the creative question of can you Claude Code your way to CrowdStrike, and you heard the answer right there from Anthropic themselves. Maybe moving along a little bit into some of the last 12 or so months, and really it started a lot earlier than that, Anthropic and CrowdStrike working together as a partner. Let us maybe go there next. We will come to Mythos, but a lot of the things actually predate Absolutely Mythos. Why did Anthropic start working with CrowdStrike? Where did that journey start from your perspective? Yeah. Around the beginning of last year, our Frontier Red team and our post-training team started to realize these models were getting more and more capable at certain technical tasks. Coding was the first breakout use case that we had late 2024. Over time, in early 2025, we started to realize these models were starting to get, I do not know if I would use this term capable, but knowledgeable, sentient, and aware of how these things work. You could see where the curve was going. These models were eventually going to get to a moment where they could become very, very effective and highly capable and dangerous. They are dual use, right? There is offensive capabilities as well as defensive capabilities. In the wrong hands, these models can do a lot of damage. In the right hands, we can really protect and secure the world in a way and at a scale that we've never been able to before. That's where you all came in. You all were one of a tiny handful of companies where you had amazing researchers, industry expertise, and we needed to work with you because, again, going back to our mission, we want to make sure AI is safe and beneficial for humanity. We don't get there by going it alone and trying to do all this ourselves. We also just don't have the expertise that you all have in security. We worked with you very closely to help us train these models, tune these models to be even more effective at certain cybersecurity tasks with an eye in mind of making sure that we tip the scales as much as we can in favor of defensive use. Also to learn how to make these models generally more safe, like to build better safeguards for responsible deployment. When it comes to the other part of the ushering in transformative AI in a way that's beneficial to humanity, we knew we would need to lean on the distribution and the reputation, and the brand, and the trust that CrowdStrike has in the market. As a company, we talk about this a lot, we see ourselves as an intelligence hyperscaler. Just the same way 10 or so years ago, companies started moving everything to the cloud for compute. We think of ourselves as providing that same kind of layer of intelligence that other organizations can build on. That's our heritage. We build for builders. Working with CrowdStrike to help you all build solutions that are powered by our AI, which is where we are best in class, and marry that with what you're best in class at, is just the right way to get broad distribution in ways that benefit ourselves and our partners, as well as the world at large. Perfect. Let's go back to April of this year, where we all encountered the Claude Mythos moment, as we call it here. Inaugural founding members of Project Glasswing from the start, working very closely with our friends at Anthropic. What did that really change in your perspective and the team's perspective on cyber, and how did it also evolve your business? Yeah, the Claude Mythos moment was like an inflection point, where all of a sudden. Again, we knew the moment was coming. It was just a matter of how big it would be and when it would come was the outstanding question. When we were close to finishing training Claude Mythos, we realized, holy hell, this thing is here. There were a few things that we wanted to make sure we got right with launching Claude Mythos. Number one, again, our safeguards needed work. We needed time to catch up and make sure these models were generally made safe. We needed to get the word out to the world. We also needed to work with our closest defensive partners, to help you all harden your systems and prepare your customers and the world. It was very fun to work with you all in those early days. One of the reasons I think we also partnered with you was not just the business benefits and the distribution, but also just the quality of your engineering teams, the product teams, and researchers. This is not an industry that is known for moving slowly, right? We cannot afford to partner with companies that just cannot keep up. You guys have been tremendous at the pace that you have been able to operate. Well, thanks. You guys have also been super helpful in Project QuiltWorks, our program over there, where some of our trusted partners that are part of that program use our technology and your technology to really deliver frontier AI readiness at a new scale. That has been a great collaboration, and we appreciate that. Moving along, we talked about this past year. I want to now hit the partnership work we are doing and things we are doing in the market, specifically at the platform level. The vision here is to enable different module experiences within CrowdStrike to be powered by different Anthropic models, when a customer wants that to happen. Can you talk about that strategy on your side, and what does it mean from a token flow perspective? Yeah. I think we just announced the marketplace work we are doing together. Now, a big part of how we think about working with our partners, like CrowdStrike, is we want to, again, expand our reach, lean on the strengths of our partners, the network effects, the reputation, the trust. We also want to make it good for our customer or for the end mutual customers. Removing friction, aligning commercial and operational incentives. These things all matter a ton. What having CrowdStrike products available on the Anthropic marketplace allows for is a couple of things. Number one, no longer are our mutual buyers doing the fight of like, is this an AI line item or a security line item? Now these things are kind of merged into one. At the source. Exactly. If you have a commit with Anthropic and you have a commit with CrowdStrike, you can use a CrowdStrike product on the Anthropic marketplace and burn down your commit simultaneously. It is hugely beneficial to our customers and really unlocks a lot of speed. It also, both of us are thinking about net new logo growth. In particular now, we have the benefit of reaching into your distribution network and you have the benefit of reaching into ours. Companies that have come to lean on us for that premier frontier model intelligence and lean on you for their security needs, they get it all in one place. Yeah. We actually announced that today, the marketplace go live. Anthropic has done a fantastic job building the premier Frontier Lab business in the last couple of years. It opens all the investment that companies have made in Anthropic to utilize that investment on CrowdStrike purchases. We are really excited about that, and I think you guys are very forward-leaning in how you are engaging with the ecosystem, engaging with us. Even at a product level, the build-out that we are going on together in terms of powering certain modules to be Anthropic-powered modules. It all comes back to what George was talking about, customer choice, and getting to the point that a customer can bring a token, their API key, and actually use their Anthropic spend again to activate different experiences within the CrowdStrike platform with us being involved in the token flow. I don't know if there's anything else you want. I think that's sort of like the big journey that we're on. Yeah, encapsulating it well. That's exactly right. Perfect. Great. Lots of opportunity there. Again, Ash was here. We start our week with our partner summit. Partner summit now has over 2,500 folks there. It's the folks who are on the floor, the folks who bring us to market every day, the folks who implement our technology, and I think it was very powerful for them to hear from you as well there on how you want to work with the ecosystem. Maybe some thoughts there on how you want to work with the ecosystem. Yeah. Again, we talked about the commercial benefits of working with the ecosystem. It's good business for us, it's good business for you. As far as beyond that though, I think technically and operationally, again, I said this, I think, the other day, which is if you want to go far, you don't go it alone. You need to partner with companies that you trust, that have shared missions and congruent strengths that help you with your thing. So, building an ecosystem is a huge part of how we go to market, and it's going to become an increasingly important part. The Anthropic marketplace itself is very young, and you're one of our first. We might be the first security partner in there. We are the first security partner. There you go. in there. We are very happy to have you there because it is going to become such an important part of how we go to market. I think underneath this is the collaboration, and you were saying a bunch of nice things about the folks at CrowdStrike, and I would echo it right back to you. We've been on these paths before, and we listen to each other. What we've learned in succeeding with many other hyperscaler marketplaces, we're happy to share that intelligence with you so that it's very successful over here. It's good. That's all goodness, and it opens a lot of doors for you, and opens certainly a lot of doors for us. To kind of round out our chat today, what gets you the most excited about working with CrowdStrike? There's a lot of names, there's a lot of folks, there's a lot of places- Yeah there's a lot of choices. We have a very special relationship and what is it about that that kind of you'd want to share with this audience that makes CrowdStrike unique, special? Yeah. These are strange times. Early at Anthropic, we used to say things will never be chill again. It's definitely proven not to be true. In moments like this, where it's just uncharted territory, we're all figuring it out together, you have no chance at winning if you're not doing it with folks you enjoy working with, who are like-minded, who can operate at the same speed. I would break it down into the business parts and the engineering or technical parts. The engineering is the savvy and the speed and the scale that y'all can operate at. Gives us a lot of wind at our sails. On the business side, again, it's the fun, the network, the relationships. I'm a salesperson by training, so it's my favorite thing in the world to come to things like this and get to hang out with folks like you and George and team, and meet customers and partners alike. Yeah, a huge fan of- Yeah being able to do this together. Well, thanks. I think we certainly find ourselves in a lot of the same hallways and accounts, and it is good for us to be able to help your customers optimize and use their Anthropic spend and unlock more value. Certainly, it is something that we hear a lot from customers answering the question that George really talked about earlier, how do we secure agents? So, together, I think we really fulfill that vision, that mission. We had a great chat today about your thoughts on both CrowdStrike as a customer and what you hear from the team over there, on how we partner together at a technology level, on a go-to-market level, and ultimately, how you cannot cloud code your way to CrowdStrike. With that, thank you, Ash, for the time. Thank you so much. Appreciate it. Please welcome Vice President, Enterprise AI of NVIDIA, Justin Boitano, Vice President, Business Development of CoreWeave, Seann Gardiner and Chief AI and Autonomous Systems Officer of CrowdStrike, Bartley Richardson. Okay, we have got a good, exciting panel here, bringing another aspect of diving headfirst into AI to the forefront. A great crew. We are going to talk a lot about SafeMind here next, and this is really the crew from an ecosystem perspective, behind the announcement of SafeMind yesterday. Bartley, I am going to start with you. We have fantastic collaboration with these partners who are also customers and friends. I want to start the discussion today talking about the problem that we set out to solve. The problem when you came here and the opportunity, what was the thought process and the genesis really of SafeMind? Yeah. We took a very straightforward path. The premise and the setup is fairly simple. I think we all heard from George last year talk about cybersecurity AI and cybersecurity super intelligence, so we start from that premise. I think we have seen a lot in the news, we have seen about models, agents, harnesses, all the tech we can talk about being used for all these offensive capabilities. Like able to evade, able to attack. When we set out to think about SafeMind and what we wanted to add to the conversation was, well, how do we take best-in-class AI? How do we take best-in-class models? How do we take best-in-class harnesses and engineering and really advantage, and I would say disproportionately advantage the defender, versus the attacker. SafeMind is about harnessing all this power and everything that we hear about, which is valid, and these models have real capabilities to find vulnerabilities and exploit. It is taking that and harnessing it and turning it back towards the advantage of that defender. Justin, I am going to come to you next. NVIDIA is a key part of how we were able to productize and deliver SafeMind, specifically with Nemotron. Maybe walk us, from the audience perspective, back a little bit. Talk to us a little bit about why NVIDIA made Nemotron. What is Nemotron? Feel free to introduce it to the audience, and then we will come back to open source a little bit later. Take it away on Nemotron. Yeah. Our view, as AI has advanced, is there is a lot of great advancements happening at the frontier. But if you can build open models that are near frontier, you can unlock many new use cases across industries. The goal of Nemotron is to build an open foundation that every company can domain adapt into these new domains, and unlock new business. I think if you look at the open source software industry, it powers 80% of the digital economy today. In the long run, I think open models will power 80% of the intelligence out there in the world, and it gives really a cost advantage way to run AI in these specialty use cases like cybersecurity. Our approach is put the data out there, put the techniques out there openly, put the weights out there, and then work with experts in industry like CrowdStrike to do that domain adaptation so you can have always on lowest cost defensive AI to help defenders. Certainly open source is something that is near and dear to our hearts as well. You guys are active in that. I think there was a letter back in the summer, maybe a little summary on there. We were proud to be, I think, the first cyber phone call on that one. Yeah. Well, this all starts, it is funny, with the models. I think the conversation stops there too soon. The harness is really where the frontier is happening right now. We also, I think, created this program together we call the Open Secure AI Alliance. Exactly. That was really to teach the industry, when we are talking about auto-run on these models to do autonomous long-running work, there is a model and there is a harness. So there is a safety question that the industry has that if an agent that has the wrong understanding of its goal decides to break out of an environment, where does that probabilistic system meet deterministic controls? The only way that the industry can understand where these agents are going is if we all share traces from these agents. If there is a lab leak, you want to look at the trace because then you can know exactly the attack path and what can be exploited. That allows us to, I think, build the mitigating controls into the right security layers within an organization. So I think that openness and transparency across organizations is going to allow us to build the best defense and the best deterministic controls for these future frontier AI systems. That is certainly the journey that we went on and the value that we see in Nemotron. Now, all this needed a home, and Seann, I am coming to you next. We needed a performant AI Neocloud with the right GPU stack, with the right inference speed. Can you talk about the rise of CoreWeave, what you guys are up to, and a deep partnership that you have with NVIDIA as well that really brought us together? Yeah. CoreWeave is a purpose-built AI cloud. We have more than 50 data centers and growing. We offer the latest and greatest NVIDIA hardware. We're predominantly 100% NVIDIA. Then we build a very thoughtful software stack on top of that to offer an AI cloud to our customers. We've actually been a longtime partner of CrowdStrike. CrowdStrike, I would say, is our number one security partner. We work very closely with them to secure our infrastructure for our customers. And about a year ago, I think after a meeting with NVIDIA and CrowdStrike, NVIDIA said, "Hey, you should talk to CoreWeave as you start to build these agentic systems and offer them to customers." And I remember very late on Halloween, because I skipped taking my kids out, working with CoreWeave to get this done, or sorry, working with CrowdStrike to get this done, and they've moved incredibly fast. And it's amazing to see. We work with a lot of SaaS and security companies and just general sort of software companies, and I would say that CrowdStrike really looks more like an AI company to work with, the speed they operate at, and just sort of you can see with SafeMind and what's been produced and released over the past year, it's really incredible. So for us, we have the training and inference infrastructure. We've partnered with the team. We're looking forward to now going and seeing how this scales to all of our customers. Fantastic. So the building blocks, the ingredients, Nemotron needs a home. CoreWeave is the cloud. Bartley, right back over to you. Let's talk about the performance and what we've already seen so far with the design partners and in the lab on what SafeMind is capable of. Yeah. It is really encouraging, and like you said, there is a lot of moving pieces, and there is a lot of blocks and a lot of components that go into this. Certainly, it needs a home, certainly it needs models, it needs harnesses. SafeMind is actually comprised of 2 different classes of models. We call them Red Tempest and Blue Solano. The core premise there is that the best defense that you can make, it must be informed by the best and the most and as many permutations of offense that you have. That is driven and accelerated by also CrowdStrike's wealth of data in all of these areas, right? We have over a decade of experience in both red teaming and offensive capabilities on the counter-adversary side. But we also, on the defender side, we have an equal or even greater amount of data that we see when our OverWatch teams or our SOC teams are actually going in and providing the remediations. So we not only see the outcome, we see the whole path that they took, the reasoning path that they took.. So that allows us to post-train and train these Red Tempest and Blue Solano models that are highly performant. Dan, to your question, what we see is, one, it has to be, at first, you must be accurate, right? So if you are going in and you are using a model like Blue Solano to automate your defense and remediation, it has to be accurate, it has to do what you want it to do. We see comparatively with just if you were to go to the AI shop and pull a model off the shelf and pull a harness off the shelf, compared to that, we see a 70% increase in accuracy versus that, right? That is due to all the fine-tuning that we were able to do. Like Justin was saying, we are able to take a model near or at frontier capability, use the data from CrowdStrike, and blow past, right, like this. So a 70% increase. So at first it must be accurate, right? It must do what you want it to do. Then second, you really quickly realize, well, it must be cost-effective, right? If it is accurate, you also must be able to run it at scale, at speed, because these defense and remediations, you are not doing it once, you are not doing it twice. This is continuous. We are in the world of continuous defense, continuous remediation. On that front, again, versus you go out and go to the leading frontier capable model in the leading harness, we are at 99% cost reduction from that. So to give you an example of that in dollars per remediation, per detection that we go and remediate, that is $10 before, and that is down to $0.03 now. Fantastic results, and so much of it became possible through the deep collaboration over the past year with NVIDIA. That collaboration really started years ago, working with NVIDIA. But the last year has really been an accelerant in terms of the technical collaboration, the amount of time, and the folks and people that have been invested on it. Justin, I want to ask you, having been involved with us at a very deep level, can you talk about what this partnership means to NVIDIA and how you guys work with CrowdStrike? Well, I will just say it is a pleasure to work with you guys. You guys are so fun to work with because you move fast. I think, George and Jensen, you could see the relationship they have built, sitting down with each other, talking about where the industry is going, and talking about how we make the world a safer place. It takes both of us. We are really just an accelerated computing company. We like to make algorithms run faster and more efficiently. Jensen will talk about the five-layer cake. That is really because power is the limiter, and then you want to make sure that you can generate tokens at the lowest performance per watt. If we can do that, and we can work with you guys on cybersecurity, we can make these AI models run faster and more efficiently on partners like CoreWeave. We can do it at scale. It can run anywhere, any cloud, any data center, any infrastructure to deliver the lowest latency results. I think the shared mission across the teams is we want to put the defenders at a differential advantage. With that as the North Star, you guys have the data, you guys have the perception of every enterprise system and customer, all the attack insights of what is going on, both with human attackers and now agentic attackers. Working together, I think, can build the best, most cost-effective, most widely available AI system. Totally agree. for defense. It is really not like a, "Oh, here is our drop. Good luck, guys." Feel free, Bartley, to jump in, and even Seann, because you guys work so closely with NVIDIA as well. When you are at this level, it is not a, "Hey, we are going to meet next Tuesday," or, "We will see you in November." It is a full-contact sport. That is how Jensen is. So it is a full-contact sport. Maybe some insights on that collaboration. It is a mosh pit in all the great ways, right? We have all of the experts. I stole that from Justin, right? I stole that comment from Justin. But it really is that. That is the speed you need to move at, right? That is the speed we must move at. It is to Dan's point. It is not like, "Yeah, let us get on a call on Tuesday, and we will work through these types of things." Then it is not like, "Oh, let us get on a call, and then we will have our engineers on this call, and then we will have our product people on this particular call." No, we all get together, and we are all working the problem in real time, and solving it in real time. There has been many days where I have been on the phone with Justin, or I have been on the phone with CoreWeave, or our engineers have been talking in real time, or we will go to headquarters. I think that is the key. One is, it is having all the people there in the right place to do it, and that can be virtually in the right room, whiteboarding. But it is not a, "We do this at this cadence." This is the job, right? This is the job, and we are doing the job all the time in this very highly collaborative, in all the best ways, mosh pit, full team sport. Seann, let's cover our partnership together, what we're doing together. I think we pioneered a very new, exciting type of partnership in the Neocloud space. I don't think it had ever been done, a security company in a Neocloud. Why don't you fill everybody in on what we're doing together, both within CoreWeave and then in the market? Yeah. So, we've grown extremely fast. I actually came through an acquisition of a company called Weights & Biases to CoreWeave. It was incredible to see at the pace that we're adding infrastructure and that we're also adding customers, and then also just the general demand. One thing the team at CoreWeave saw super early is that security is very important to our customers, very important to us, and we've made a huge investment there. Deploying CrowdStrike across our estate was sort of the first step. I think, as we see some of the newer solutions that CrowdStrike's offering, from the Pangea acquisition, we did a very nice integration with the Pangea acquisition and some of our software tooling to help customers secure generative AI. Beyond that, now with SafeMind, we really see that as an opportunity to go together, both to the AI natives, but also the enterprise, and bring joint solutions to them. We're extremely excited to see this, one, out in the open, and two, how do our sales teams, and our go-to-market teams, talk to our customers and solve real problems for them? Yeah, I think of CoreWeave, and I think it's an important point to make here, as really a net new greenfield attack surface. A couple of years ago, nobody had ever heard of a Neocloud. There was no such thing. It's this whole new world that's emerged, and it's a whole new set of workloads that didn't exist before either that runs on the Neocloud. So it's really important, and none of it would be possible if it wasn't for the foundational layer that NVIDIA's brought to the world. But the attack surface has just expanded so much. We have a huge opportunity together, to go secure all the great work that your customers do. Of course, it's an honor to secure you, and we love your cloud, but there's a sizable opportunity in the market that we're after together. Yeah, this is all net new as well, right? All of our customers use hyperscalers. This is just an expansion of the market. We really see a huge opportunity. Justin, maybe rounding it out a little bit here, why do you guys work with CrowdStrike? Jensen had a lot to say on stage, but I would love to get your thoughts on why we are on this mission together, why our companies work together. There is, I do not know, 3,000-plus security companies in the world. There is a lot of great ones downstairs. Everyone should go check them out. But we put a lot of effort, focus, attention, in a concerted manner on you. You guys certainly do the same for us. Why? I smile, DB, because I can get a hold of you anytime I need to. It is true. You guys move fast. I think you guys have deep expertise in cybersecurity. We're a big customer of CrowdStrike. I think Jensen said it on stage, you're his number one cybersecurity partner because you guys have this vision for how to secure the world's infrastructure. We don't want to be a cybersecurity company. It's very clear. We know what we do. We can accelerate algorithms. We can make LLMs run fast. We can create a foundational digital intelligence for the world. But you guys have the domain expertise. You guys can really take it into every global enterprise, in every country to secure the digital infrastructure that's critical to our economy. Well, I think what our panel today really represents is, in essence, the power of the crowd. I mean that because there's a whole new world as AI came and collided with security over the last couple of years. There's a whole new world that we've embraced. A lot of great people and a lot of opportunity. Certainly, we haven't sat on the sidelines investing heavily in these partnerships, utilizing the tech, working very closely with the people to innovate and co-innovate and create new solutions for customers. I think SafeMind's a great example of that. It's a next chapter for CrowdStrike. The fact that now we're a frontier lab, cyber's first frontier lab, and the honor of building with and on NVIDIA and having the home for that be CoreWeave, and then the fantastic results that we're delivering in the market. This is the full circle right here, and we wanted to take some moments to share it all with you. So thank you to our panelists, and I believe next up is lunch. Thanks. Thanks. Thank you. Thank you. [Break]. Please welcome President of CrowdStrike, Michael Sentonas. Good afternoon. How are we all? Great to see everybody in person. We always do this every quarter over Zoom. It is much better, I have to say, doing it in person and getting an opportunity to see everybody and to connect. As you heard from Burt at the start, we will make sure that we have a lot of time for Q&A and get an opportunity to see everybody afterwards. So, I hope everybody got to see the keynotes over the last couple of days. I have spoken to a few of you, so I know some of you have, which is absolutely fantastic. What a Fal.Con. For those of you, and I know some of you were here for the very early days back in San Diego, great to see how big this has become and such a huge community event. It gives us a great opportunity to connect with you all, which is absolutely fantastic. Let me go back to last week. We shared a lot of numbers with you all. Fantastic results, talking about how we achieved $5.8 billion in ending ARR, which represents a 23% CAGR over the two-year period. This is incredible, and I just wanted to sort of unpack a little bit of this as we go through the numbers. It is about 30% of our way to our $20 billion ARR target we provided last year at the investor briefing. I am going to unpack some of this and dive into a little bit of this and talk a little bit about how we are going to get there, and just give you a view, a walk, to how we get to that $20 billion, the products, the investments, where we think we have an unfair opportunity, and I want to step you through that, which I think is going to be a lot of fun to go through. Those of you that were here last year, you probably remember this slide. It was in the deck that we shared with everybody. We talked about our rapidly emerging businesses. We talked about identity, next-gen SIEM, cloud, key drivers of our $10 billion- $20 billion ARR targets, which have performed incredibly well. I am really proud of the way that the team has performed, and we shared numbers across those areas last week. Just to take a little bit of a step forward as we talk about this path to $20 billion, let us just start off with the key information straight up. Today, I am excited to show you that we believe we can reach the $10 billion ARR target within FY 2030. Similarly, we believe we can reach the $20 billion target within FY 2035, and I am going to talk you through the path to get there. This is an important one, and I think if we look at the path to $20 billion from a market share perspective and what we have to do, the target addressable market, which I go through with you every year, continues to get bigger. It is a staggering $565 billion market by calendar year 2034 when you look at all of the pieces that go into the portfolio that we have. This is not all security. All security is obviously bigger. These are the areas that we play in and technology that we believe we have a right to win in, and I am going to talk through. Let me break this down a little bit for you, because when we think about that path to $20 billion, how do we get there? We talk about market share in core endpoint. If we look at how we are performing there, we talk about around modern endpoint, we have around 20% of the market. Close to 50% is still using a lot of the legacy. If you start to look at and you build that out, and I will go through a little bit of the product areas, and you start to think about the market share that we have and how we get to that $20 billion, start to unpack a little bit of this. From a market share perspective, if I take a step back, on average, we have captured roughly 4% share of the security market that we address for the last six years. Start to play some of that math forward on that path to $20 billion. What do we need to capture in calendar year 2034 on that path to $20 billion? If you start to think about that math, we only have to capture 3.5% of the market in calendar year 2034 to hit our $20 billion ARR target, despite the fact that we have averaged over 4% market share every year over the course of the last 6 years. We obviously strive for a lot more, but I am just talking you through an illustrative path to that calendar year 2034 timeframe. Let me jump into the product, and let me unpack some of the opportunity by product. Jumping into this, let us start with agentic security. We have got to start there today. We estimate this market will reach $115 billion in calendar year 2034. This is where we look at technology like SafeMind and Falcon Guardian, which we announced yesterday, Charlotte AI Gateway, new announcement that I made this morning. If we capture 3.5% of that market, we can add $4 billion in ARR. As you heard from George, we launched a new AI model called Models and Harnesses called SafeMind. Incredibly excited about this when you think about the applications and the problems that we can solve. This is very, very significant. We announced Guardian this week. Today, I announced the AI Gateway and additional innovations that we are working on. Hopefully, you saw some of the demos where Guardian is working with other products. You are going to get the opportunity to build on the platform and see how all of the components can come together. So let me jump into Guardian and unpack a little bit of this. Hopefully, again, everyone saw the demo. This is incredible innovation that we built in-house. We took the Pangea acquisition, we took some of the capabilities that Pangea brings, but we also took the sensor and the platform, which is the incredible point to touch on for a little bit. We give customers the ability now to see what agents are doing. We give customers the ability to control those agents. We start to have the ability to understand what you are doing in Claude, in Codex, to have a look at Kiro, to have a look at the open weight models, and start to make sure that anything that is going on inside the organization that is happening at machine speed, Guardian has the ability to either give you visibility, to give you some of that control, but importantly, to stop attacks. I talked about supply chain and the threats of supply chain, and agents having the ability to pull down packages and compromise an organization at a speed that we have never seen before, all part of the same capabilities with Guardian. So that visibility across the entire estate, this is one of the biggest things that CISOs are asking us for, CIOs are asking us for. Tell us what is in our network, tell us who is using it, tell us what they are doing with it, tell us what tokens they are spending, and help us get control. We have answered that question, which is absolutely phenomenal. The architecture that I stepped through as well today, and I am just going to touch on this a little bit here to make sure that everyone caught that or if people were not in the session, is really important, because we think about AIDR as a category. When we have been talking about AI/DR, we think about the data, the identity, the applications, the infrastructure, the models. All of this has to be protected. This is what you need to protect yourself in an AI world today. It requires a comprehensive solution. It requires a platform. You are not going to stitch this together with 10 different products. It is going to be very hard to be able to do that. That is why we have been making sure that we bring Guardian together with a sensor, with the platform, and then it becomes an off-ramp to all of the other products across the CrowdStrike platform. That is really, really important. So adding capabilities around discovery, posture, hooks into identity. Data protection will have a reconnaissance as a technology category, because wait until you see an agent that starts to exfiltrate data, and every organization on the planet, if I can speak English this afternoon, will start to want to bring in data protection solutions into their organizations. Runtime guardrails, response, SaaS, endpoint, cloud, everything needs to be part of the capability here that we talk about. This is important for me because when we think about the journey that CrowdStrike has been on, EDR defined how we secure the endpoint. AI/DR will define how we secure the estate, and Guardian is how we deliver it. Hopefully, that is super clear. I want to roll the demo again because this is a pretty cool demo and I am super proud of this one. So roll the demo. Introducing Falcon Guardian, our new solution for securing AI agents where they execute on the endpoint at runtime. The Falcon sensor now extends to AI agents. No hooks, no SDK. Guardian fuses AI and OS telemetry from prompt to process across your entire AI estate. Let us see this in action. Last month, a Fortune 500 company found 18,000 AI agents running across their endpoints. They approved 300. Every dot you are looking at is an AI agent running right now discovered by Guardian on day one. From OpenAI Codex, Claude Code, Cursor, and Kiro, to unclassified shadow AI. Active, dormant, or hiding in plain sight, they are detected by sensor, scheduled scans, and DNS queries. You see them, now you can control them. Start with the simplest case. Ken, in customer service, is working on a billing dispute. He reaches for Gemini out of habit. It is what he uses at home. Guardian blocks it and points him to Claude, the approved app. So he moves to Claude and pastes in the customer record, including a full credit card number. Guardian masks it at the prompt. The number never reaches the model. Ken gets a policy notice, the security team gets the event. Ken was typing. But agents do not type, they act, and they can act on text that is invisible to humans. Marcus in engineering prompts Claude Code to debug a Lambda deployment and walks away. The agent gets to work. Unaware, Marcus just put his whole company at risk. Falcon Guardian brings these risks into sharp focus. In a cluster of high-risk Claude Code agents, one lights up as critical with a threat score of 90. Guardian identifies it as Marcus’ agent and resolves it to his AD identity. Unknowingly, his agent introduced three critical risks, a malicious skill, an attempt to transmit data to an external destination, and AWS credentials were accessed outside of their expected scope. His agent footprint shows every skill, tool, MCP server, and connection, feeding one composite threat score. The sensor builds a single causal timeline, prompts, skills, and tool calls from the agent fused with process spawns, file reads, and network egress from the operating system. One story end to end. This is the agent graph. Marcus' agent follows a link in the repo docs to a GitHub issue thread. Buried in that thread, a hidden instruction, an indirect prompt injection. It tells the agent to load a skill and exfiltrate Marcus' credentials. Marcus never sees it. Guardian stops it. The credential exfiltration is blocked before the keys leave the machine. Where else is the skill running? In an AI-first world, security teams can now use their agents to directly investigate these events with the Falcon MCP server. In Claude Code, an analyst queries the Falcon platform for agents that have previously used the skill. 12 found, every attempt blocked, no credentials left any machine. The hosts are quarantined for investigation. That was the prompt layer. Here's an attack that never touches it. On another workstation, Claude Code installs a new plugin from a public repo. It registers a local MCP server to assist with project indexing. The prompt layer looks clean. The model isn't being manipulated. But that MCP server quietly performs credential theft and exfiltration on each tool call to index project. The sensor layer catches the exfiltration. Prompt-only security tools see a clean plugin. The sensor sees the attack. That's layered protection. Now the operational side. Guardian tracks token usage in real time, per user, per agent, and per model. Each session attributed, each model accounted for. Finance has expense visibility. Security sees anomalies. Both get the data from the same source. Two attacks, two layers, the prompt and the process. One sensor, one graph, one truth. This is Guardian. This is security for AI. Now, a lot in that demo, and the coolest thing for me, when George announced Guardian yesterday on stage, the product became generally available for every one of our customers, literally in the 30 seconds, the first 30 seconds of that announcement, which is fantastic. We got a flood of customers last night saying, "How do we get it?" Falcon Flex, call your account rep, ask for the module to be flexed in, and you're using Guardian. It is a licensed entitlement. The sensor was updated some weeks ago for all of our customers. So today, we have customers that are running Falcon Guardian. So incredibly proud of that project. Let's talk a little bit about how we win in this market, which I think is very clear to everybody. AI has to be secured at runtime. We already have the sensor on the endpoint where agents run, where the models and applications run. Even if you're using frontier AI in the cloud, you are connecting from the user, from their endpoint, you're leveraging identity, you're connecting to those models. We give visibility and control into that whole thing. We have the data and the context that no other vendor has because of that visibility. And finally, you need to have protection across the entire agentic attack surface. That includes, as I've said, models, identities, applications, infrastructure, all of it. The Falcon platform covers all of it, which is incredibly powerful. Now, another topic. We've been talking about this for the last couple of years. Let's talk about the agentic SOC market and opportunity. We estimate this will be a $52 billion market opportunity by calendar year 2034. Let's unpack a little bit of this. Again, let's use that same math. If we capture 3.5% of this market, of course we want to do more, that would equate to $1.8 billion in ARR. So let me get that clicking. Now, here's a different perspective, a different way to look at this. Unlike Falcon Guardian, which is a new product, a new category, we're not starting at zero in this market. We have nearly $700 million in ARR from our Next-Gen SIEM business. So to capture 1.8 of that ARR in calendar year 2034, we only have to grow our annual recurring revenue at a 13% CAGR, which is very achievable, I would argue. As you can see from this slide, we're currently growing our ARR by 60%. So again, just playing that out so you can see an illustrative example up through to calendar year 2034. So I talked a lot about this this morning. Even with an army of agents, even with a lot of solutions, defenders still have one problem, and that is time. I believe, and George talked about it in his keynote, the concept of breakout time that you've heard from us for so long now is probably disappearing because things are happening so fast that you need to operate at machine speed. So we talk about agents and adversaries, more importantly, having the ability to reason, to adapt, to act at machine speed. That breakout time as a concept is basically at zero today. The reason why we talk about this is attacks don't wait for humans, which means the solutions that we use, that organizations use to protect themselves need to discover, decide, and move in real-time speeds. If the breakout time does get to zero, if it's happening that quickly, the time to investigate will also go to zero. That's why the old SOC model is broken, and we need to think a little bit differently. Attacks that move at that real-time speed need solutions that can bridge that gap. So why don't I show you another demo and jump into the agentic SOC solution built on Next-Gen SIEM. Roll the demo. This is CrowdStrike's Agentic SOC. You can see attacks assessed, detections triaged, meantime to resolve, and autonomous resolution rate. Detections are being triaged, investigated, and resolved continuously without anyone grinding through a queue to make it happen. This is the operation at full scale, thousands a day, around the clock, at machine speed, with humans on the loop wherever you want them and on every critical decision. Let's go inside one of these investigations, start to finish. The investigation opens and runs from the top. It's an AI abuse case. An attacker tricks the IT help desk copilot into issuing a rogue credential, then uses it to reroute payments. When a verdict comes out of a system like this, the first question is always the same: How do I trust it? How do I know it's right? The Agentic SOC earns that trust by showing the work done, dispatching expert agents in parallel, each specialized in its own area. Underneath, they run an analysis of competing hypotheses, the same tradecraft analysts have used for decades, weighing the evidence and ruling out the benign explanations. Together, the collective agents land the verdict. True positive, malicious, response staged and ready, but nothing irreversible fires until approved. Work that once took hours or days, done in minutes. Your people stop doing the grind and start making the call. That's what runs out of the box. But this is your SOC, with your use cases and your data sources. Build your own agents and wire them in. Describe what you want. An insider risk enrichment agent that understands your business and pulls in HR data, badge access, and prior incidents. That data typically lives in other systems, like your ERP or ITSM platform. Connect it to your Workday and ServiceNow MCP servers and your Next-Gen SIEM data for correlation. A few clicks, and Agentic SOAR handles the rest. Before rolling it out, test it. Pull a recent detection and see what the new agent flags. Once it looks right, publish it. The agent gets to work, enriching insider cases with your data alongside our built-in experts. You saw the SOC dashboard, attacks assessed, investigations resolved. But you're also managing the automation behind the operation. Dozens of agents and workflows across multiple projects. You can't check each one by hand, and you don't want to hear something's wrong from an analyst. This is your view. One dashboard for every project, every agent, and every workflow. You see the work each is doing, what each is costing, and what's failing. Let's drill into the SOC automation project. Now we see each resource inside, agents and workflows all in one place. The SOC gets the automation benefits while you maintain governance and control. This is your Agentic SOC, powered by CrowdStrike, triaging, investigating, and responding on its own. Nearly $700 million in ARR, as I mentioned, on our next-gen SIEM technology. Why do we believe we will win in this market? We are faster, we are natively available to all of our customers, and it has a massive disruptive price structure. Faster in terms of the way that the technology is architectured, which is critical to winning in this space, available to all of our customers. We did the work several years ago. Every one of our customers is next-gen SIEM enabled. It is a license entitlement to turn it on if they want to use that. Importantly, we do not charge them for the data that is already native to CrowdStrike. We also give them capabilities like federated search, leave data where it lies, and give them the ability to search across data source. Architecturally better, smarter, faster, and a much more disruptive price opportunity. Let me sort of touch on the Agentic Identity market. We estimate this will be a $48 billion market in calendar year 2034. Again, playing out the math. If we capture 3.5% of this market, this will equate to about $1.7 billion in ARR across identity, privilege access management, and so on. As I said with next-gen SIEM, this is not a market that we are starting at zero. We have nearly $585 million in ARR from our next-gen identity solutions. Today, we launched Agentic Identity Provider, which means we are well-positioned to grow our ARR at a 14% CAGR to reach $1.7 billion in calendar year 2034. As you can see by the chart here, we are actually growing at 33%. You can do the math across this. I want to walk through a little bit about how the solution works, how it is a little bit different, and then you can see how the solution is going to win in the marketplace. The reason why this is so critically important, today with identity, people get trusted when they log in. They get trusted when they authenticate, access that they keep and is persistent in their session. That is the wrong way of looking at identity. We need to flip the model, remove standing privilege, and give access only when needed. Give access that is tightly scoped to the task. Give tightly scoped access that is short-lived, which reduces risk, and then when the task and the work is completed, access is gone. It is a different way of thinking. When agents hand off between each other, you have the ability, again, to give agent to an access, to remove it as the handoff completes, and make sure that you remove all the risk inside the organization. That is why today we announced CrowdStrike's Agentic Identity Provider solution, which extends continuous identity discovery, enrichment, and zero-standing access to all agents. I hope the demo works this time. Roll the demo. Emily, a DevOps engineer, wants to use Claude with an AWS MCP server. A CrowdStrike policy secures this end-to-end. It defines the required risk level, trusted device posture, ServiceNow ticket context, and identity verification with MFA needed to grant AWS access with the appropriate roles. Emily must authenticate first. CrowdStrike checks policy conditions are true and requires MFA. In the background, Falcon Guardian discovered the agent and automatically registered it with CrowdStrike's agentic identity provider. Here you can see the agent record has been enriched with information about Emily, business context, such as the department she works in, and security information such as risk score, privileges, and account type. This also includes an assigned SPIFFE ID, a stable, standards-based identity that tells us exactly what software is running. With the agent registered and its identity known, we can now enforce the right authorization policies. Emily queries all ECS services for changes in the last 24 hours. Claude communicates with the AWS MCP server and returns the data. Next, she attempts to access production customer data in S3. The policy assigned an AWS role that excludes S3 permissions. They aren't required for this task, and the request is blocked. That blocked query triggers an anomalous sensitive data access detection, a medium-severity alert flagging an attempt to reach a sensitive resource outside Emily's behavioral baseline. The policy evaluates risk holistically. She is already flagged as a departing employee, an insider risk signal. Together, these push the risk score to high. Because of this change in risk, AWS access is revoked. The policy also extends to processes on the endpoint itself, preventing agents from executing under privileged, high-risk, or non-human accounts. Emily pivots to a service account with on-prem access and AWS deployment rights. The moment she tries to run Claude via that service account, the process is blocked from starting, and the breach is stopped before it ever begins. All of this is part of our vision for securing AI. CrowdStrike's agentic identity provider will govern what resources agents can invoke, enforcing real-time, risk-based policies, registering trusted identities, and ensuring end-to-end accountability for every action taken. In this example, we show how Claude, Gemini, and OpenAI agents are granted access to Salesforce and GitHub with vaulted credentials and only when operated by authenticated users. This is just the beginning of our agentic identity provider journey. Agentic identity security demands continuous access and evaluation. This is how you solve identity problems today. Solutions for both humans and agents, which you saw, continuous access, which you saw, the addition of SGNL AI acquisition that we made at the start of the year gives us incredible power and opportunity to make sure that we enforce zero standing privilege. This is mandatory to solve identity problems today. Let's wrap this up and look at everything that we've talked about. We know that adversaries are going to use autonomous systems to find weaknesses, to adapt, to attack. Our defenders are going to need the same solutions to be able to respond. They are going to need to use AI to defend against AI. We have the single AI native platform that can do that for our customers with simple deployment with the one agent, one sensor solution. So just jumping through this a little bit, when you start to sum up the total AI security opportunity in front of us, we estimate that that will be $215 billion by calendar 2034. If you recall from George's presentation where he tied AI security spending to the overall AI spend, he showed a slide range of 1% of the spend at the low end and 8% at the high end, which equated to a range of $36 billion -$291 billion. So this market estimate sits towards the high end of that range. If we can capture 3.5% of that market, that would equate to $7.5 billion of ARR in calendar 2034. That $7.5 billion is roughly 38% of the $20 billion target that we are chasing. So we're incredibly excited about the AI opportunity that's in front of us. That $7.5 billion that we can potentially add to our ARR in calendar 2034 is incredibly exciting for us, which is why we are so focused on this as we start to build out our platforms and work with the team. So we announced an incredible amount of innovations this week. We've still got a day to go. Platform enhancements, new features, new releases, and I can't say this enough, I'm so proud of what the engineering team has done, the speed that they're working at to get customers the right solutions that they need to protect themselves. This is the summary. We'll make sure that you all have the slide. I'll be back for Q&A. Thank you. Please welcome back Burt Podbere. All right. I heard a lot of cool things today. I heard one of the Frontier Model guys say, "Reasoning can't build another CrowdStrike. They don't have the data, they don't have the context of the data, and they don't have the expertise." I heard that over and over. I mean, that was pretty cool. I knew that, but it's something else when somebody else, a frontier model guy says that. We heard from our CoreWeave friends and NVIDIA friends, and you combine that with what George had talked about with Jensen yesterday, it all is packaging up for us to be in this unique opportunity, in this time, in this world where something has come to us and we were ready for it. And that's the important thing to take away. CrowdStrike's ready for what's coming, right? All this amazing innovation that we are seeing in the world today needs to be secured, right? CIOs are panicking because their CEO is saying, "We are AI." 89% of companies that George talked about are touching AI, right? They need to be protected. Guess what? We have Guardian. Let us do a quick recap of what was talked about today. There is a lot of opportunities that we talked about. AI adoption is soaring. We talked about the stat. You guys see it. We see it every day. We see it every day in the news. It is here, it is now, it is happening at AI speed. It is just incredible what we are seeing with respect to AI and adoption of AI. George talked about building cybersecurity's first model purpose-built for defenders. He was on stage with Jensen, and the two of them were going back and forth about what this could mean. Jensen is excited about it, let alone George. I think everybody in that audience who was there was kind of going, "Wow, that is something unique." That is actually kind of the DNA of CrowdStrike, being such an ahead of the curve type of company. Innovative, thoughtful, knowing where the puck is going, if you are a hockey fan. That was incredible what I saw, right? Securing agents. George and Mike talked about securing agents. George talked about, hey, for every single person, there is potentially 90 agents. George talked about, well, could be more. Could be 1,000, 10,000, whatever the number is. I agree. I spoke to one investor after our Q2 earnings call, and we went through that, and we said, "Hey, one human to 90 agents." There was a pause. This was a very big investor. There was a pause. He goes, "Can't that be 1,000 or more?" We are like, "Yeah. Potentially could." So his mind was thinking kind of like what your mind is thinking right now. This is a massive opportunity for us. Today, we now have 34 modules. We have a cyber frontier model. We have the technology that is required in today's world. We are showcasing it here at Fal.Con in front of 10,000 of our partners and customers. If you were in that auditorium when we were going through all this stuff, you saw the energy, you saw the excitement. What I saw, I saw people going on their phones and texting their folks, "We need this. We need this. We need this." That is what I saw. We talked about the endpoint is the control plane. Look, AI is consumed at the endpoint. We have shown you for the past several quarters that our endpoint business is growing, accelerating. Why? Because of what I just said, AI is consumed at the endpoint, right? So there is nobody in the world today who does endpoint better than us. Now, again, something has come to us that requires the endpoint to be able to digest all this activity that is taking place in this new AI world. Mike just walked us through this incredible TAM, right? He walked us through on a byproduct, how we think about getting to the numbers that we talked about, the $20 billion and the market in front of us, the $563 billion. That is a big number, right? We have the tools, we have the technology, we have the people, we have the AI to get there. The TAM expansion is going to come from inorganic and, of course, organic as well. We have both of those avenues to go after that big TAM. I think we are in the pole, George likes racing and we all like racing in CrowdStrike, pole position. We are in the pole position to go get that TAM. Oops, sorry, guys. When you put this all together, all the things that I just talked about, gives me conviction in our ability to scale this company to new heights. Each one of those things, to me, screams this conviction towards our goals. I will share some more of those goals with you a little later. It is probably worth a minute just to take a look at the Q2 highlights. It was $333 million in net new ARR, 51% year-over-year, an all-time record for us. Non-GAAP operating income up $372 million, an all-time record for us. You have got top line growth an all-time record. You have got profitability at an all-time record. For me, this was the greatest quarter in CrowdStrike's history. For me to see it all come together was just super exciting, and we paused when we read all of your notes. You wrote brief, thoughtful notes about the quarter, and not just about the quarter, but the outlook. I read them all. We all read them all. We want to just thank you guys for hitting on point, because you did capture that, and that was, I think, well received by all your readers. Let us talk a little bit about Fal.Con Flex. I have said it a zillion times, but it is worth repeating. It is a commitment model, not a consumption model. George talked a little bit about it. We worked diligently with our auditors and everybody else and obviously our customers, to be able to get what they wanted. I would love to be able to take credit for Flex, but it is the customers. The customers deserve the credit. Those are the folks that told us what they wanted. They wanted to digest all of the things that we had, make it simple for us, make it easy for us, and then we put our thinking caps on. George was adamant that we got to make this thing friction-free. We got to make sure that this thing can be super well-absorbed, super fast, and super simple to be able to get through the end of the contract. You have heard it many times about why it is successful. You have heard many times about you contract once and then it is another PO just on the reflexes. I have talked to each of you after our earnings calls to talk about the fact that, hey, look, if this is being consumed faster and you have seen the data then how fast it is being reflexed, all you need to do with the reflex is get to the CFO and say, "Hey, I just need a little more for CrowdStrike. By the way, I am going to give back a bunch of dough that I am spending on somebody else." Any CFO worth their weight in gold is going to take that deal every single day. The other thing I would like to talk about is that Flex has the same kind of recognition that non-Flex does. ARR and revenue are recognized ratably. There is no change. There are no mixed models. There is no confusion over how we are doing this thing. It is very simple. There is no change. And that is part of the beauty and the elegance of this model. It does a lot of other really good things for us. It allows us to contract faster, we get bigger deals, we get longer deals. You are able to kind of reflex really quickly because they have taken the friction out of the sales process. So there is a lot of goodness in Flex. As George has talked about many times, it is not just, here is the license and everybody can use it. They are stealing the name, but you have to have the right technology. You have to have the best technology out there. You have to have many modules where you can reflex to. You also need to be able to flick on those modules straight away. You need all those things to be able to have proper Flex and see the results that we are putting up on the board. George has shown you these numbers just before, you have seen them all, but I like to go to the one all the way on the right. The average ending ARR uplift from a non-Flex deal to a Flex deal. We talked about the 40%, but last year it was 34%. That is a big jump, 34%- 40%. When we go to the sales team, you have folks that they like the old ways or they are used to selling the old ways. When you put that stat in front of them, they go, "Oh, yeah, I am going to get them on Flex." Hey, these folks are coin operated, right. So it works really well for the customer, really works well for us, it works well for our sales team. It all kind of comes together. Let us talk a little bit about what we are seeing with the power of Flex first go to market. You can see here, the Flex logos as of 2Q26, we had 61% enterprise and 39% non-enterprise. Now it is 50/50. So why is that happening, right. Flex first. Our sales team is going out there and selling Flex first because they know that if they go to a non-Flex deal, they got to get Mike to go approve it. That is additional friction from the sales perspective. The sales rep does not want to go to Mike for an exception, right. So we are really leaning forward on Flex first. There is friction if you do not go there, and we have already seen all the benefits why you need to go there. I envision a period of time, some point in the future, where there is no such thing as a non-Flex deal for CrowdStrike. George covered the tectonic shifts, what we are seeing in cybersecurity or in the world. I have talked to, I think it was Michael, we were talking about the fact that this is a generational shift that we are seeing in technology. You can call it whatever you want. Generational, there is an inflection point. For me, it is transformational, what we are seeing in the world today. Mike talked about our path to $20 billion in ARR, from a market share perspective. I am going to now talk about the $20 billion from a customer and geographic perspective. For me, I like looking at things on different planes. You can do market share. I like customers, I like geos. I want to see it all, and that's how we build our internal models. Mike talked about the 3.5%. You saw the numbers, you can all do the math. Given the track record that we've had of capturing roughly 4% over the last six years of market share, we feel that we can be able to reach the 3.5% and then some. You got to execute, you got to do all those things, but it's there for us for the taking. 3% isn't a huge number against the $565. 3.5%, $565. For me, I think we can achieve that. We just got to execute. We got to do all the things we've been doing in the past to be able to get there, but really achievable from, as I think from a TAM perspective. Look, if you want to look at our customers and you look at new logos, today we've talked about the fact that we've got 100,000 organizations as of Q1, and that includes SMB, it includes MSSP, it includes everything. Then you look at these numbers, and if you start from the group that has the smallest, so the global enterprise of 13,000 plus in that category, all the way up to 50 million in the SMB space, 100,000 organizations is a drop in the ocean in terms of logo runway to get to our TAM. I've talked to all of you over the years about how we have so much headroom in both new logos and our base, and I'll talk about our base in a minute. But you can see the stats here. We have just begun, really. We went public in 2019. It's only 2026, right? We're in the early innings of our journey in terms of going after customers. We feel we have a tremendous amount of headroom to go with respect to going after a new logo. The geographic opportunity. Look where we were in FY 2022. 72% in the U.S., and the rest of the world was 28%. Five years later, 65% in the U.S. and 35% rest of world. I'm excited about our international opportunity. You can see CY 2034 international market opportunity of $244 billion. It's great. I have a goal in my mind, I think I've talked about it with all of you in the past, I'd like to be 50/50. It's hard to get to 50/50 when the U.S. is doing so well. But the international has done well, too. We're spending more money in S&M in the international markets. We're spending money on sovereign cloud to go after the international markets. Again, I think we're in the pole position to be able to capture a whole bunch of that $244 billion by CY 2034. The expansion opportunity. As I talked about, we have a lot of headroom and new logos. Tons of headroom. We also have a tremendous opportunity with our current base. We were at $5.8 billion, 2Q 2027 ending ARR. This is existing modules within the existing customer base in terms of the white space. We showed you the white space opportunity 2Q 2026, of $21 billion. We're going to add another $8 billion based on our customer base today. Total white space opportunity for 2Q 2027, $29 billion. That's pretty good. I have this tremendous opportunity and headroom in new logos. I have this tremendous opportunity to sell more to my existing customers who love us already, who want to do more with us, who want to swap out other technologies. They want to have the best outcomes, and we help them lower their TCO. This is why we have been winning. I love this slide. My good friend DB came to the company four years ago, and I will be honest with you, transformed our whole partner go-to market. We kind of looked like everybody else before. It was there. We were not investing in it enough. DB came along and really transformed, for what I saw, the partner community. All of our partners, whether it is GSIs, whether it is the partners we had on stage. You can see the numbers. We had eight partners, each with a lifetime TCV of $1 billion. That is impressive. 19 each with lifetime TCV of $500 million, and 65 each with lifetime TCV of $100 million. 65 each with a lifetime TCV of $100 million. I remember, George, when we hit $100 million TCV as a company. Here I have got 65 partners each with $100 million plus. Incredible. 890 of our partners doubled their business with CrowdStrike year-over-year. 890. Part of our success is nourishing and growing with our partner community. What I have seen in the last four years has shown me that if you can do it right, you can hit the numbers that we have been hitting. As you saw, you heard earlier from Mike, we pulled in some of the dates on when we are hitting the $10 billion and the $20 billion. For the $10 billion, we said we were going to do it at the end of FY 2031, now we are doing it within FY 2030. Similar theme for $20 billion. We said we were going to be able to hit that by the end of FY 2036, now we are going to do it within FY 2035. Why are we saying that? We are seeing significant momentum in the business. The business is firing. There is a huge opportunity. We have got both of those dynamics happening. For us, when you see that momentum, I can tell you right now that my friend George Kurtz, he is not letting that momentum slide away. Foot on the accelerator, right, George? Okay, let us talk about the target model. Here is a summary of the target model in full year FY 2029. The ones in green, S&M, R&D, G&A, we are already in the band. Subscription gross margin, 81% to our target model of 82%-85%, so we are right on the cusp. Operating margin, we have seen incredible increases in that. We are now at 24%, this is all of one half of 2027, to our target model of 28%-32%. I have talked about free cash flow for the first half, and we talked about ending the year at 30% plus to our target model of 34%-38%. I think we are in earshot of all these things. I feel that this is the same three years ago when we said it, same now. Not changing it. I am not extending it out, keeping it as it is. There is still room for efficiencies across a lot of this stuff. AI is helping with that. So for me, you got a lot of things that are making this model work, from all of the products that Mike has talked about, to Flex that we've come in, to the partnerships. I feel really good about this model. I will go into a little more detail, especially about gross margin expansion. Many of you have met me for the first time back in 2015 when I started. When I started, gross margin was in the 30s. I looked at George and I said, "We got to do some different things to be able to go public." He goes, "Yeah, I know. Let's figure it out." We did. We figured out what we needed to do with respect to increasing our product portfolio, moving to our own private cloud. All those things needed to come together in a way that we are going to increase our gross margin and still be able to deliver the products that are best in breed, and then a platform that is best in class in the world. We did. We went from the 30s to 81. This is a story that is most near and dear to my heart. George and Mike have their products. They love all that stuff. I love gross margin. I worked hard with the teams on gross margin. We figured out ways to be able to deliver the best-in-class products to the world and still be able to deliver a great gross margin, which then I can take and reinvest into our business. Last year, we put $1 billion into R&D. A billion. Gross margin expansion matters. How are we doing it? We have our public cloud partners and customers, and we did a couple things with them. One is we looked at where it was more economical to put the compute and storage. For example, Amazon, we went more to West 2 versus West 1. We also have scale, so we are able to have volume discounts with our public cloud providers. That is going to help us more in the future as well. As we get bigger and everything else, we are able to enjoy better discounts. On the private side, we are able to look and say, "Hey, look, where are the more expensive areas that we can bring into our private cloud, and let's do some more migrations." We continue to look for areas to continue to migrate for those more expensive areas. We also are using AI to find multiple data stores and reduce that, reduce our costs. I feel really good about where we are going with gross margin. Obviously, a point at 81%, going from 81 to 82 is a lot harder than going from 30 to 31. But I feel really good about it. For over 10 years that I have been at the company, we have never taken our eye off of gross margin, ever. All right. Let's talk a little bit about free cash flow margin expansion and some CapEx. As you all know, we talked about FY 2027 having a 30% plus free cash margin. Now I am talking about FY 2028, and I am looking at a 32.5% plus free cash flow margin. I feel really good about that. I also look at my CapEx. We have a slight increase in CapEx, going to 11%-12% next year. Even with that, I still feel really good about the free cash flow margin. They go hand in hand. You got to talk about both, right? Obviously the prices of storage and everything else have gone up. We are aware of it. We have done a great job in terms of procuring and great job with our vendors and our strategic vendors to lock in certain pricing. I feel really good about, at the end of the day, the 32.5% plus free cash flow margin. We got a lot of growth opportunities. Mike touched on them earlier, George touched them on earlier. I believe these, capturing a percentage of the TAMs that I talked about is very achievable. It is a large, it is an expanding TAM. Large and expanding. Those are really good words, and I think we are right in the epicenter of being able to get it done. Conviction in FY 2027 net new ARR acceleration. Let me walk you through some of the history. When I was here last year at this time, I said that for this year, we are going to do 20% in terms of growth on net new ARR. I think you were all here for that. Then at the 3Q 2026 earnings call, I said I am going to do 20%, but on a bigger number. You all remember that. Then after 4Q, I said I am going to take that 20% and raise it to 22.5%. I felt pretty good about that. Then we had the big jump in 1Q, right? 520 basis points. I was really happy about that. I said I thought it was a prudent guide when I did it. We are looking at all the factors. You still have to execute against it, but we did. Then, of course, last quarter, 34%. Overall, total raise to date, $220 million FY 2027. I am really proud of that, and a lot of things needed to happen for that to work. Now as I think about the future and all the things we talked about here today, here is what I think about. I think about this. I think about 20% plus year-over-year net new ARR growth for next year. That is what I think about. We told you that the midpoint, we are going to be at $1,355 million at the end of this year, and the next year, $1,626 million equal to or greater than. I think that summarizes our conviction in where we think this business could go. It is pretty exciting times for us, and I am willing to go out again at Fal.Con, which is well before end of Q4, to talk about next year. With that, I hope that this has been enlightening for everybody. I hope that it has been transparent with everybody. I hope that it has been consistent with what you have seen in the past. With that, we are just going to get ready for set up for Q&A. Stay tuned. You can take a bio break or whatever, and come back in about five minutes, and then we will start Q&A. Does that make sense? Work for everybody? Awesome. Thanks, everybody. Thanks for your time. [Break]. Please welcome Vice President, Investor Relations and Strategic Finance, Andy Nowinski. Joined by Chief Executive Officer and Founder, George Kurtz, President Michael Sentonas, Chief Financial Officer Burt Podbere, and Chief Business Officer Daniel Bernard of CrowdStrike. They crossed themselves right now. Oh, look at the hands. All right. There you go. Thanks, everyone. This is my favorite part of the day. We have 45 minutes for the Q&A. If you could please raise your hand, I'll get you in. Then wait for the microphone, because we have a lot of viewers on the webcast that also want to hear your question. Then if you could, please state your name and your firm for the webcast viewers, that'd also be helpful. Why don't we get started? Sure. Let's go first here with Brian. Hey, good afternoon. Brian Essex from JP Morgan. Thanks for doing this again this year. I'll say, you have to be here to appreciate the scale. I don't think I've had to sit in the top tier of an arena to watch a keynote before for a security software company, so that was pretty impressive. Thank you. But George, the question on why you can't Claude code your way to CrowdStrike, I think it was great to hear Anthropic's answer. Operationally, I think, we've all been Yep reiterating why you can't do that. I would love to, from a technical perspective, if you could just put a bow on that and describe for maybe generalists that are listening, if you take a model and add a harness to get an agent, how your engineering of your own proprietary agent protects your IP, your data, your processes, your context from access by some of the foundation models and limits their ability to maybe distill what you're doing on your platform. Well, a lot of it starts, again, with the data itself, but the data is really captured because of the architecture. If you think about the scale that we operate in, we have a single agent that operates in multiple operating systems, if you will. Then you have to collect this data en masse, which is really hard to do in a performant way. Then you have to have a cloud at scale that understands how to process all this. Then you have to make sense with all the algorithms, and then you have to have deep security domain knowledge to be able to understand what's good, what's bad, and you have to be in line. This is really important. There's no LLM that's in line. It doesn't stop anything. It'll give you a nice text explanation of something, but it doesn't stop anything. In security, you have to be right the first time, first and final, as we talk about, when you make a conviction on something. So it's a much different model, if you will, operating model, than just say, "I have an LLM and a harness." That's not going to do it. We talked about how to leverage models which we're building. We talked about the frontier models which we're partnering with. Again, it's about customer choice. But as a net data creator, the 7 trillion events, this is what our system was trained on. There is no Reddit for what we actually created. You're starting to see the frontier labs. They're buying old books so they can scan all the out-of-print books because they need more data. The fact that we create data, have it, and we have an architectural system gives us a unique advantage. Two is we're totally focused on security. This is a big deal. We have a company and a sales team and a customer success team that's focused on making sure that companies don't get breached. That's our mission. We're not doing other things. This focus is very important. Then when you put it all together, we have distribution. I think sometimes people forget in 2026 what it means to have distribution. They think all of a sudden you whip out Claude and you're going to dominate the world. Yeah, it's great technology, but all the fundamental things that we all learn in business school, you still actually need. So that's the reason why we're partnering with Anthropic, with OpenAI, and others. You've seen the ecosystem. Because customers want a trusted security partner. They want their data to remain with the trusted security partner, but they still want to have access to all these different models, ours and others, in a way that gives them sovereignty and the results at the right cost. Thanks for the question, Brian. Please be patient. We have plenty of time. We got 40 minutes here. Next question we'll take from Saket. Thanks, folks. Saket Kalia at Barclays. Thanks for a great couple of days here. George, I want to ask a little bit about product. There were so many great announcements. The one that I thought was super interesting was Guardian, and that too with OverWatch. Yep. As I compare this to cloud security, it felt like cloud security in the past lacked cloud spending more than we expected. That market also had some big players, like Wiz, for example. Maybe the question is, because you had a great comparison of hyperscalers to the frontier models. Maybe the question is: Do you think agentic security will lag as much as cloud security did? Is there another Wiz out there in this space that's on your radar? Well, I think if you go back to the graph that I created, which basically had the cloud adoption and the security adoption. It was a longer time horizon. I think right now, as I said, you can't roll out AI without security. When every customer is literally begging you for a product like Guardian, I don't think that's going to take long. It's just a different point in time. It's a different type of technology. What you have to realize is, from a cloud perspective, when cloud first came out, and we were one of the early adopters, if you remember when I started the company, not everyone needed cloud. You could wait a few years. If you go to the board and you say, "Hey, we're going to wait the next three or four years until technology shifts a little bit more, and we kind of get around to it to implement AI," you're not going to be in the seat for long, right? It's not going to happen. The fact that it's a mandate, the fact that the benefits are incredible from a cost perspective, it isn't like companies are going to wait to migrate to AI. They're there. They have to be there. That's number one. That's going to force it. I think, when you look at where we are today and sort of other companies, I think we're in pole position. We already have the agent. We delivered prompt visibility. What we were able to build and the unique capabilities to fuse prompt visibility with the runtime heritage that we have is incredible. I got off the keynote and was like, "When can we get this thing?" Even though I said it was available, like, "Where? When? I got to have it." My phone blew up. Your phone blew up. I haven't been really involved in something like that, really since when we first launched EDR. When people saw EDR, they're like, "Okay. Nobody has that. We want that." It's the same thing now. I think we're in a great position to be the dominant player in that space because the flywheel that we've built, the heritage around endpoint. Again, I want to be clear, we've been talking sort of endpoint. We have many, many customers with workloads, and this works in a workload. If you're running Claude in a workload or opening Codex in a workload, in a container, we run there and do the same thing. It's very important. Thanks, Saket. We're going to take a few from this side of the room now. Let's go to Gabriela. Hey, good afternoon. Thanks so much. Gabriela Borges from Goldman Sachs. My favorite product announcement was actually SafeMind. Maybe for George and Mike, talk a little bit about how customers have been saying for years, "The odds are against us as defenders." What do you think the paradigm shift is going to look like to go from all of the vulnerability management issues, the pen testing, the patching, the patching Tuesdays becoming continuous, versus just adopting something like SafeMind? Burt, for you, how do you model something like this? Because there are all of these pieces between frontier model distribution and token consumption, and there seems to be a lot of moving pieces to it. Would love your thoughts collectively. Thank you. Well, when you look at SafeMind, what we've found from customers is they want something that allows them to operate at the same frontier caliber, and they want specific to security use cases, right? We're not solving the world's problems. We're solving the world's securities problem, and that's a big difference. We have the data. We've got the technology. You saw Bartley. He came from NVIDIA, and this is what he was doing there. We are in a unique position, again, to build not only a lab for security, but to actually deliver on these frontier caliber models. Tomorrow, you'll even hear more stats on what we were able to do. We're excited about that piece. I think data sovereignty, what I mean by that is not just by geography. The fact that the data is being kept within CrowdStrike is a huge opportunity for our customers because they can use our models or they can use a frontier model. They can use all of it. Doesn't matter. We can do the model routing, right? That is a massive opportunity. But I think maybe to sort of the heart of your point around vulnerabilities and exposures and things of that nature, in a post-Mythos world, it's all about exploitability, and there isn't enough time to get the patches out. There just isn't. You think about your organizations. If you're a cloud organization and you want to put something out in a template, great. But not everybody operates that way. We have to be able to, like in this first incarnation, this is one example. The thing is unlimited in what it can do, but customers love the fact that we can create a digital twin of their environment and then very rapidly understand where those exposures are and then rapidly allow the system to create the mitigations. That's going to drive a lot of adoption because they can't do that on their own, and it's sometimes too risky to try to figure it all out. It's a great first use case. Anything else you want to add to that? Just that it's an iterative cycle that you can run every day, all day, every weekend, every night, and just keep getting better as you go. Yeah. The bad guys are getting better, right, with reinforced learning, so we might as well do this. This is one of the most exciting projects I've been involved in. I can tell you, I had standing calls 3 times a week at 9:00. I was in there in the UI, change this, do that. Here's what we're building. We really galvanized the whole company, would you say- Yeah around this, and we handpicked the best engineers in a relatively small team to be able to put this together and deliver something that I think is just going to be extraordinary. Yeah. On the modeling, it's no different than anything else that we do with new products, right? I look to George and Mike and all the customers that they talk to, get the inputs, get kind of a flow of what we can do, and then really bake it on assumptions and on the top line to figure out where it could go. Then, you bake in some of the assumptions on the cost side, right? It's similar to the 40-page deck I did to gross margin when we took it from 30%- 80%. So it's no different. There's no change in terms of the methodology. I will say, when you think about cost, it just kind of jogged my mind. If you remember the demo that Mike did, we actually include the ability to understand the token usage and provide visibility to cost. This is huge. There isn't an IT organization that isn't asking for this because they have no idea what's going on. In fact, we had one customer that basically came back and said, "My EA was spending $10,000 a month on tokens." And he's like, "How did that happen?" So I'll tell you the story. She was very diligent, and she was taking all these emails, and she was putting them all together and what's going to happen for the week, putting it into Claude. But Claude didn't really understand how to read the EML format. So it built a parser every time she put a new email in. It literally built a parser. All these new emails get dumped in, and the token usage goes nuts. Not only do we have the ability to actually provide protection, but we are selling now into the financial world. I mean, the IT world that is concerned about cost, right? The CFO is not necessarily our customer, but when you think about cost and AI, it is going to be a huge benefit to be able to provide that visibility to the entire organization. I mean, you want this 100%. Right? Okay. All right. Thanks, Gabriela. Great question. Next one, let us go to, and let us stay on this side, let us go to Mita. Maybe just a question on, you noted greater efficacy than the frontier models that you guys have from the Red Tempest and Blue Solano and the lower cost. A couple of questions there. Just one, what were instances where they did find vulnerabilities that maybe your models did not find originally? Then second, just where do you envision, I understand that customers want choice, but where do you envision that they will use choice kind of in OpenMind? Yeah, you have to look at, it depends on the model, depends on the scenario. At any point, some model could be better. Obviously, these evolved new models are coming out, right? I think when you look at how the models work, you have to include the harness as a needed element, right? I do think this gets overlooked. You can have a couple of models, and they can be good, but when you actually apply the harness to it, they get really good, and you get to reduce the false positives. To be fair, as we went through this and we have trained our system, it kept getting better and better and better until we got to a point of better performance in the areas that we were focused on, which are generally what people care about. The other thing to realize, too, when you think about cost, there's this concept of a turn in a model. If it took you 20 turns in a model, versus 16 in our model, just as an example, it might take less. If it took 16 in a frontier model, it's so cheap to the extra four turns that it doesn't matter in terms of cost. If it took you four extra times, you'd get there. From that standpoint, you can get essentially the same or better results at a cheaper cost, and it's just kind of the engineering that goes into it. A lot of it comes down to the training and the harness and then obviously the post-training around how you keep getting better and better, which are going to give you the great results. We only have one thing to focus on is security, right? We're not worried about all the other things in the model, and that's really what allowed us to get these sort of results. Look, it's going to go up and down over time. New models are going to come out, and I think maybe the last part of your question is, we can provide the routing so the customer can pick what do they. They want to use a frontier model, great. They want ours, they want the best answer, they want a composite view, use all of it, right? The whole idea is that we want to get paid through the whole token flow process. Okay. Thanks for the question. Let's keep going down the line here. John? Thank you. John DiFucci from Guggenheim Securities. Listen, there's a lot here, and I appreciate it, because it was almost like a, I don't know, it was like an AI 101, or actually, AI 401. What you're doing looks really different from what we're seeing from a lot of others. I'm not quite sure, because admittedly, the others are really vague, which makes me think they either don't have much or they're just farther behind probably. We can think about, and you showed those graphs up there about the potential for growth, and you gave some numbers, but the potential looks pretty impressive. One of the things I was thinking about when you were up there, and I guess this question's for Burt. I know you gave the cost per task, which is a lot lower than the LLMs. We do not even know what the profit is on the LLMs because they are private companies. Maybe some people in this room do, but I do not. How should we think about margins? If this really takes off, and I know you gave some 28, the free cash flow, that is great. How should we be thinking about this? Is this going to weigh a little bit on margins? You have done a great job. I remember talking to you guys when you were private, when it was 30%, and I said, "Do not tell anybody that until you get it above 60." How should we think about that right now? Because it seems like they could weigh on margins a little bit, this business. Yeah. It comes down to what George talked about, tokens. At the end of the day, we are going to price tokens appropriately to be able to have the cost baked into those tokens so they can make sense for everybody. It goes back to the regional theme about margin. We talked about it on stage, and I talked about it with Gabriela's question. There is going to be no change in how we think about putting that model together, but it is going to come through in tokens. Yeah, and what I would say is, again, as you build these, because they are so specific to security, there are less things that we need to do in all the other areas. When you are a frontier model, you have got to account for anything that thing could possibly do. So our scope is more focused on security, which again, if you are just in that area and you are training in that area and you are doing alignment in that area, yes, it takes money, but it is focused in an area. Again, I think the great thing that we talked about was partnering with NVIDIA to be able to leverage some of their technologies to do that. I think we put together very cost-effective way, which gives us frontier caliber capabilities with unique data sets and partners that are helping us and working with us to get the best performance at the lowest cost. Yeah, thanks, John. Start over on this side. Fatima. Good afternoon. Thank you so much for taking my question. Fatima Boolani from Citi. George, I wanted to ask you a question, with respect to a slide that you shared. It was a triangle slide with cloud and endpoint, AI DR, the network. My question for you is how much of the traditional, conventional alphabet soup that is cybersecurity for the average buyer, how much of that is an opportunity for you, and how much of that is a limitation or a challenge to you in articulating your vision? What I am getting at is, eventually, does the distinction even matter because your core principles are grounded in the one sensor approach? How is that resonating and is that something that buyers are saying that, "I am ready to just sensorize everything to help solve for these multitudinous use cases in cyber." Just your thoughts on these false dichotomies and alphabet soup, and if that is a hindrance. Well, there is an alphabet soup, obviously, in the market, and you heard me say this probably in the past, I have never seen a PowerPoint that was wrong. It all looks good. It all has same message, different colors. But the proof is in the pudding. When you come to an event like this in a stadium that you have never seen before with so many customers. With partners that you have never seen, that Jensen shows up to, you know you have got something special. I think what customers have recognized is that the point. You have the AI creation, we are going to build models and those sort of things. Got it. That happens in the data center. We are working with all those partners. But for the companies that are actually consuming AI, it is what you are doing right now, because probably each one of you has some level of AI on your desktop. If not, there are long-running agents in the cloud. That is where it is going. Or it is a SaaS agent from any of the platform players. But if you take 2 of the 3 of them, we are world-class experts in that. We have more agents deployed than any private security, not private, but any standalone security company. So that really is a unique benefit and that sort of flywheel, the Crowd and the CrowdStrike, and the platform piece that we have been talking about. No one wants yet another agent. No one yet wants all these other vendors. And when customers stop their buying and they basically say, "What is CrowdStrike doing? Are you solving this problem? Because if you are solving this problem, we are going to buy it from you using Falcon Flex and using the single agent." And that is a huge differentiation between us and just about all the other companies that are out there. Thanks, Fatima. Let us go next to Gregg. Great, thanks. Gregg Moskowitz from Mizuho. Also have a question on SafeMind, and just have to say, running countless iterations of Red Tempest and Blue Solano in a closed loop, it just really shows incredible ingenuity. But as it relates to the token aspect of the equation, naturally, SafeMind is going to be part of Flex, but do customers need to procure initial token packs, or will they be included with Guardian or some other module to whet their appetite? And secondly, how are you guys planning to approach pricing of Guardian? Because to my mind, the functionality is surely even more powerful than your AI DR module. And as we have seen, demand for AI DR out of the gate, has been pretty remarkable. Thanks. Yeah. I'll try to make sure I cover a few of those. I'll let Mike jump in if I forgot anything. Pricing will go up on Guardian. The capability is unbelievable. Compared to what we would call AI DR, that's going to be retired. The category will be AI DR, and the product will be Guardian, and the price goes up. From a token perspective, companies, I'll start with, you really won't be able to consume. From a SafeMind perspective, you won't be able to consume SafeMind unless you're a Falcon Flex customer. Let me just say that piece. Put that aside. But for the purposes of what we're talking about here, you can buy the tokens, and we'll look at the estate that you have, and then generally what we try to do, and this worked out well with customers is, we'll basically come up with a number of tokens for your entire estate. Then, if Burt doesn't use a lot, but I use a lot, it sort of all equals out, but we give some level of visibility to customers. So it isn't sort of up and down and, Burt didn't use any. I crushed it. I've got to pay more money. We basically look at the pool of tokens, and then if they exceed that pool for that month, then they can certainly buy more from the Falcon Flex packs. Customers really like that. It's predictable, but there is obviously a step function if they're using more. We want customers to be able to have access to Red Tempest. We want customers to have access to Blue Solano, the harness, all three. Think of all the areas where we can use this in the platform. Having SafeMind, working together with Guardian. Think of the power of SafeMind with exposure management. Think of SafeMind working with Next-Gen SIEM. There's going to be many different ways that people can access the technology. Some will include with tokens, some you buy outright. As George said, Falcon Flex is the easiest way to get this. Same thing that I said earlier with Guardian. Everyone that was texting us, flex it in, go for it. You're done. That needs to be easy as well, which I think is important. People want predictability. Finance teams want predictability when they think about the usage of it. Thanks, Gregg. Let's go to the next question up here. Let's go to Patrick. Thank you. Patrick Colville from Scotiabank. I am actually going to stick to SafeMind. I think one of the reasons why is we have just seen the power of proprietary harnesses. I mean, tonight, Snowflake just printed this most extraordinary result, and part of the reason is because they have a proprietary harness, and here today I am like, "This is potentially going to be explosive for CrowdStrike." This is really exciting. When I speak to CISOs, some are already using these models in Cybersec, and what they have been telling me is that it is really expensive in terms of the commits that OpenAI, Anthropic are demanding. I can see how this will be really compelling to kind of lower that cost. My question is this, what proprietary is CrowdStrike adding? You touched on this in your presentation, George, but is it the threat intel? Is it the kind of red teaming for years? What value add are you guys providing to make that harness so robust? Just to be clear, SafeMind is included in the gross margin, free cash flow margin, and ARR guides you put out today. Well, let him go. Yeah. So if you think about it's a system, right? You've got the harness and the models, and I think you really bring up an important point because the models are going to get to a point where there's just this diminishing returns as they keep getting better and better, right? They'll keep getting better and better, for sure. But when you think about what makes the difference, a lot of it is going to be this exoskeleton, the harness. And there is a lot of know-how that goes into our software, as an example, right, that people haven't replicated. And there's a lot of know-how that actually goes into something like the harness, which essentially is software. So we've taken all the years of experience in seeing the attacks, understand how they work, understand how to protect against those, and then we've built a harness to be able to get the best outcome with the lowest false positives. And there's also a level of cost engineering that goes into it. You can do a lot of dumb things and potentially get the right answer, but it might be really expensive. It's just like cloud. When you think about how we've optimized our cloud to get 81% gross margin. You can't just go, I have a cloud and I have an agent and have at it. Things have to be optimized. Those become barriers to entry, right? The 15 years of data that we've used to train the models, the experience that we have in Falcon Complete, in our services team, we understand how these attacks work, we understand how to recover from them, and it's all that intellectual property that goes into building proprietary software, AKA a harness. It's no different than the software that we've built for the last 15 years. I think you will see that across different industries. You pointed to one here, in your example, where the harness is going to make the difference because the models are all going to be really close. If you look at the open weight models now, they are very close to the frontier models. Close enough to get good results, right? The software layer that goes on top of it, I think is really going to be the X factor that gets you the best results for the lowest cost. That harness engineering is important. We saw this early on. We couldn't have delivered a harness now if we didn't start it yesterday. We started it long ago because we saw what was happening in preparation for this point in time. We've seen examples where you can grab a model and a harness and get 80% false positives. If you're doing red teaming and you're looking for issues inside your environment, you got 80% false positives, you're burning tokens and you're getting nothing. Then you have to keep tuning and tuning and tuning, and you're burning tokens the whole time. By the time you get to a point where you're getting an effective outcome, you've got no money left. To your question about what's different about us, we provide the models, we provide the harness, we provide the training, we provide the threat intelligence. The goal is that the customer uses it, and they get the result when they first run. That's hugely different. One thing to add to Mike's comment, I think there's a long history of security buyers and security practitioners wanting security products, and I think this speaks to that very trend. You guys have tracked this for a long time. A lot of generic technology companies have played in cyber. There's a reason that there's a CrowdStrike. To answer your second part of the question, it hasn't been released yet, right? I haven't baked it in this year. It's contemplated for next year, but just remember, it's early days with this thing. Thanks for the question, Patrick. Ittai? Thanks. Ittai Kidron from Oppenheimer, and thanks for the day. Super informative. Burt, great guiding 28%, and you are out for 34%, another 34% year-over-year growth. Appreciate that. George- You said that, I didn't say it. Yeah. George, I want to go back to your early presentation. You made a comparison of the cloud adoption and AI adoption, but one comment that you said immediately following that was that you don't expect a spike in security spend as it relates to this. You expected, I think it was a consistently strong growth. Maybe I'm paraphrasing there, but you certainly try to remove the possibility of a spike. My question is, why? If AI spend is growing vertically up and enterprises need to operationalize it here and now, and they can't do it without security. Why shouldn't the growth of your business not decelerate into a fiscal 2035 target, but rather accelerate not for a quarter or two, but accelerate for two, three, four years heading into this? Especially given what we're seeing from the frontier models. Well, I want to be specific what I said, which was a spike is just buying more of the same. Sustained momentum and tailwinds are buying something different to solve a different problem. So when you look at the growth, the growth can be explosive, for sure. What I was saying is, the Mythos moment isn't like a one-quarter trend. To be very specific to answer your question. When you see the curve and how that curve is going up, these are sustained tailwinds because you're not just buying more of the same, you're buying something to solve a different problem, the problem that we see today and the problem in the future. That is a massive opportunity. That's what I was really saying. It isn't a one-quarter phenomenon of like, "Hey, you've seen a spike, it's Mythos moment." This is going to be a long-term trend. Getting back to my point, if you think there's going to be more AI in a year, 3 years, or 5 years from now, security's going to parallel the slope of that curve, for sure. Thanks, Ittai. Let's take our next question over on this side. Let's go to Michael. Thanks very much. It's Michael Turrin with Wells Fargo Securities. Appreciate all the content over the past couple of days. I just wanted to spend some time getting your perspective on the tone of customer conversations you're having at the event. I'm sure you've been busy. I'm curious specifically on the sense of urgency around AI this year versus last year when we were talking about some of these same things, but you didn't have the fully featured product vision behind it. We didn't have the Mythos moment and some of the things that are out there today. So curious to pick your brain on that. Then George, with enterprise AI spend, you mentioned it's net new. Are you finding that true for cybersecurity currently? Or if not, where does that come from? Sure. I'll let you take the first part. Yeah, look, I think as George said, we got hammered with questions as soon as George presented. It was SafeMind, it was Guardian. Obviously, today, people are asking about Agentic Identity Provider because they are trying to wrap their head around like this is going to solve a problem in security that we have had for a long time. Last night, I was walking back to my room, and I walked past the lobby of the hotel, and there was a group of customers, and they called me over. They were saying, "You guys are operating at a different level. We are just not hearing this from other people." It was, "When can we get it? Is it part of Flex? Where does it go in the products? What comes next?" Ultimately, everybody said, "We have comfort. We came here worried about what we are doing with AI because we have to go back to our business and tell them how we are securing it. You have given us the answer." I have run into the sales reps, and people are like, "Can you come and see this customer? Can you come to that event?" I have had about 12 EBC requests in the last six hours. Everyone has had the same sort of thing. So it is definitely hitting the challenges that people are having, and we are giving them the answers, which I think is critically important. I think George said it well. We have got a stadium full of people, and that live stream went to an incredible amount of people around the world. So it is like coming in from everywhere. It is not just the people at the event. I think when you look at the. I may reframe the question. When you look at the budget, where is AI security budget coming from? It is coming from all over. You could have a CTO basically saying, "Hey, we are rolling out AI agents in this part of the business, and we cannot roll it out unless we have the security piece." Boom. You are getting security budget from the larger budget of spend. So that happens. You have budget coming from, "Hey, we are going to get rid of a few products, and we are going to take that, and we are going to burn down more Flex." Here you go. We have it there. Then you have budget that there is always corporate money somewhere. I always use the example, if a customer has a breach, are they not going to find money to go remediate it? Yeah, of course, they are. So there is always this sort of money that is hanging out there where somebody has to do something very quickly, and it is CEO money and CFO money somewhere where they go, "Hey, we have got to get this done because we are holding back the adoption." When we met with one of our big design partners, huge, I said, "What is the number one success criteria?" It was not like this feature or that feature. The number one success criteria was we need to roll out AI faster from the CEO down. If that is the mandate, this is not the security CISO, this is the CEO going, "Roll it out next week," you are going to find money for it. So that is what we have seen. Thanks, Michael. Let's take our next question over here from Brad. Thanks. Hey, guys. Brad Zelnick with Deutsche Bank. Thanks so much for having us. Excuse me. I think what my friend John DiFucci was trying to say is that this has actually been a master class this week. I think of all the vendors and luminaries that we listen to, this is the strongest, clearest vision for the future of where AI and cyber are going. Thank you. Hats off to you. A lot of questions to ask. I want to touch on next-gen identity, which is a cornerstone of that pyramid that you showed us, really important, and how differentiated the approach is that you're taking. There's a well-established PAM market that's been around for years that others are saying is the answer. You, after a couple of very key acquisitions, are now just getting up and running with something that potentially seems very different, very disruptive. If you can just explain to us the why, and why it's the right solution would be great. Then from Burt, if you can maybe even frame it for us in the context as we saw of like AIDR, how quickly it's ramped in a matter of months, what the milestones and markers are for this next-gen identity solution that you're thinking about ahead. Thanks, guys. Yeah. So look, as you know, we've been talking about identity threat detection response for years now through the Preempt acquisition. We're very unique in the way that we approach that. We've not stopped there. For us, we were thinking about just-in-time access. We were thinking about PAM use case. There's not a customer that I've spoken to that said to me, "I love my PAM." They kind of feel it's a bit bloated, it's a bit archaic, but they use what they have. They're sweating their assets, but they're asking us for a much more efficient way to do just-in-time access. There's some session recording and some vaulting things that people ask about, but we added the just-in-time access piece. The biggest thing that we've thought about is, how do you solve the problem for identity tomorrow? Solving identity is solving it for the human, it's solving it for the non-human, it's solving it for the agent. This concept of giving somebody trust once so that they can carry out their activity, making somebody an administrator, you forget they're an admin, you don't take it away, they get breached, there's a huge drama that happens with it. That model is fundamentally broken. Without going sort of too deep in the topic, it's the same thing with things like session cookies. Somebody gets a session cookie, they can do things because they've got privilege. Even if an employee leaves, getting that cookie back is really hard. That's why adversaries go for it. There has to be a better way, and that's why we bought SGNL. SGNL has thought about that problem and built a very elegant way to give access when you need it, access on demand, based on business logic, based on taking in threat identity information. So you give somebody access to do the task that they need, and then you take it away as soon as they finish. This is the dream of security professionals. We've talked about this at Identiverse, last quarter. Several organizations stood on stage and said, "Hey, this is how identity needs to work today, and the one company that can do it is CrowdStrike." That wasn't a CrowdStrike event. It wasn't a security event. It was an identity event that has all the identity players. So we're really excited about that. The identity provider release includes that capability, and we're going to just keep adding releases. So we're going to be talking about identity provider at Fal.Con Europe. We're going to save something for the next event, and we're going to keep building on the way that that works. But you saw all the pieces on the screen. You can use it for authentication. You've got the Falcon Identity manager that you can use there. You bring in vaulting, you do the agents, you do the humans. It's all included. So just to follow on that, I just want to make an example. If we think about next-gen identity versus the older technology, like a PAM technology, right? As Mike said, we are solving for the problems of the future, which is, I should be able to give you my username and my password and have nothing bad happen. Because ultimately you have my username and password, but you are not entitled to do anything. You do not meet any of the criteria to do what I want to do. So you have a username and password, nothing happens. If you think about all these attacks that you read about, it is mostly a user and a password and an MFA that is compromised, and then all of a sudden everything falls apart because of those entitlements that are attached to it. You know why? Because they pulled it out of some vault that had standing privileges. That is the difference. You are protecting the credential which has all the entitlements attached to it, and you pull it out of the vault and it works everywhere. And our model is, I will just give you the username and credential, but it does not have any of the entitlements. Does not matter, right? So you do not get entitled until the system understands what you want to do, and then you only get those entitlements for that micro slice in time. It is a much different model, and it is much more conducive to what the agentic world needs. And this is why, again, we got ahead of the curve in buying something like SGNL. And now you can see the fruits of what it is going to bear. I think the ramp is going to be just as fast as securing agents. I think people are going to realize really fast, "I need both. Yep. So. Yeah. Yep. We have spent a lot of time on the integration. The big thing acquiring it was making sure that it is part of the identity suite, making sure that in the back end it is all integrated. Same thing that we are doing with Seraphic Security. I was sitting at a customer dinner last night where a customer was talking about SGNL technology that is in this release. Obviously, they did not use the name because they did not know the new name last night, but they were talking about Seraphic Security and the two. And they basically said, "Hey, we are turning off SASE products." There was a whole conversation, like, "You guys have really done it." And we are using SGNL, we are using Seraphic Security. It then became a conversation, "Well, what can we do in the identity space?" It was just fascinating to see people saying, "Hey, this just does it a much more elegant way and actually solves the problem for tomorrow. Thanks, Brad. Let's go to our next question here. Let's take one from Matt. Thanks, guys, for doing this. This is great. George and Mike, I know you've talked about this before, and Burt, you mentioned it today. You guys fundamentally believe that the endpoint is the key control plane for the AI era, and it's clear in your numbers. Burt, you always say the scorecard is the financials, and the results and the guidance certainly illustrate that's the case. I still get the question from investors, like, why not a network vendor? Why not a SASE vendor? I guess, we get your perspective, but I guess I'm curious, why is that such an important place versus a network, a SASE-based vendor? I am going to say it in a different way, because you said endpoint, which is true, we talked about endpoint, but let me say it in a different way and I think it will make sense. Runtime is the control point. You never get to generate any network traffic until you actually run it and execute it. The agent has to be spawned, it has to have an identity, and then it has to do something, and then it has to reach across a network. Right? You are never going to have any network traffic unless you run something. This is why when we think about runtime as being one of the most critical control points, along with identity, right? Because you have got to run it and you have to have the right identity. We are actually seeing this. We actually have a transparent proxy. By the way, all the network traffic we see, because we have a transparent proxy built into it. So we see it before it hits the network. So we are actually in front of the network seeing it. Then, by the way, if there is some AI that is taking place that doesn't have our agent on it, Mike talked about the gateway. We are actually launching a gateway. So, by the way, if you are not using our gateway, fine, there is a whole bunch of other gateways. We integrate AI DR into all the other gateways to get visibility and enforcement. But it starts at runtime. You will never have network traffic unless you actually execute something. By the way, the network traffic goes through our proxy, so we actually see it before it hits the gateway. Pre-encryption. Thanks, Matt. We will take our last question here from Joe. That is both endpoint and cloud workload, just to be clear. Yep. Joe Gallo, Jefferies. Thanks for the question and thanks for all this. George, you laid out the AI security TAM as a percentage of AI spend. You are not sure if it is 1%, you are not sure if it is 8%. You answered Michael's earlier question just like, "Hey, budgets are fluid. People just need security." So, as you are thinking through your pricing, how are you approaching that? Because on one hand, you want to make sure you set the appropriate price and drive long-term value. On the other hand, you want to drive adoption. So, how do you even approach that with this never-ending stream of new products? If people don't really know if I am allocating 1% or 8%, what are customers thinking about and what are you guys thinking about as you set prices? Well, look, I wish I had a crystal ball on pricing. If you figure it out, let me know. I am just trying to give you one representative example of what has happened in security in terms of percentages. You can take one, you could take eight, you could take somewhere in between, but I know it is going to be meaningful, and I think it is going to be different in each customer. If you are in an environment that is highly regulated and you are driving AI adoption, it is coming down from the CEO, and they are trying to keep head count down, there is going to be a lot more AI security, right, that might have a higher budget. Think financial services. If you are in other areas, maybe it is a little bit less. But I think on average, I believe it is actually going to be a greater spend than what we saw in cloud security. If you use that as a benchmark, I think it is going to be a greater spend there. It will evolve over time, but at the end of the day, what is different in this cycle is when the cloud came out, not everyone needed the cloud. You operate in your data center, and you are like, "Oh, if I get to the cloud, I will get to the cloud." In today's environment, everyone needs AI, which means that security for the first time is the gas pedal, not the brake pedal. That, I think, gives us opportunity to take more of that budget. Given the fact that we are a platform, we can come in and we can consolidate. By the way, we can come in and be aggressive in some of the newer products just to get the adoption. Then you turn it on with Flex. It is a single agent, it is there, it is an entitlement. The ability to actually scale this quickly is there. It is there in the commercial harness, and it is there in the technology platform piece. All right. Thanks, everyone. All right. Are we wrapped up? We're wrap. Yeah. You have any final words or Well, yeah, thanks for that. I guess my final thoughts are, we know that you can be a lot of different places, and to come out, to spend a little bit of time in Vegas with us, to hear what we have to say, but probably more importantly, hear what our customers and partners have to say means a lot. That's the best use cases that we have. They're walking around with our technology in hand. So we appreciate it. We'll be around. We can take some more questions and do our little fireside chats over there. I hope you guys enjoyed the conference, and we'll see you soon. See you soon. Thanks, everybody.