CrowdStrike Holdings, Inc. (CRWD) Earnings Call Transcript & Summary

Good morning, everybody. It's great to see everybody once again. best to see everybody live and in person, and welcome to what is our biggest Falcon 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. And I'm going to go over to the quick agenda and hopefully, you're able to leave here with the same feeling that I have when I came here and I'm feeling today about the opportunity in front of us. So 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 filed a fireside chat with DB. Then we're going to have a little break for lunch, a 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. And then we'll have a quick interlude and then we'll go right into Q&A. So I'm excited for today. I hope you guys get a lot out of today. We've extended the Q&A session well was 1 of the part of the feedback that we got from last year, you guys wanted more time with us. So 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.

Well, I can't believe with an event that we've been able to pull off. I think one of the premier security conference is now, and obviously, we get to spend some time with the financial community, and it's important to be able to what we're building, the innovation that we're driving at CrowdStrike and more importantly, how we're solving customer problems. So with that, I will jump into the presentation. And 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. So 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 in some of the numbers this McKinsey stat. If you look at organizations that used at least one AI in at least one business unit, it was 55% 2, probably like a PT and today is 89%. ChatGPT messages are up 8x and reasoning tokens, 320x that's where we are today, right? So we started with chatbots, then we moved to reasoning. Now we move to agents. That is driving, of course, the spend. And 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 megatrend that we've never seen before, and it's the fastest-growing transformation of spend in the enterprise history. And just look at the stats, $2.67 trillion with a T in 2026, going to $5.95 trillion in 2030. Again, a Gartner stat. So for me, if we follow the money, it's all going to make sense and what that means in terms of the opportunity from a security perspective. Now we've seen this movie before. And we've seen it with the cloud adoption and the hyperscalers. And 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 like 5 services at the time, compare that to today. But you can see the slope of the curve over time and the adoption of the cloud. And 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 and they go we're never going to the cloud. That's a risky we're not going to do that. Think about where we are in 2026. So 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. So if we just take cloud as a, 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. And you can just look at the slope of their revenue and sort of what they put out publicly, if you will, even though many of them are private. But in general, you can see how fast the adoption is and how fast the spend is. Well, certainly, I think the adoption now for security is going to parallel that. And what's important to realize there we go. What's important to realize is that -- that's okay. Just keep it over there -- is that securities role is different this time in that security is now the accelerator not to break. And in many of the technology shifts, it was like, well, security 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. So adoption is a necessity. It's larger net new opportunity and it's immediate and rapid spin. And 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 to $36 billion to $291 billion opportunity for security. Let that sink in. And again, apply whatever filter you want on it, what I can tell you, it's big. And 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 and 5 years than there is today? And 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. And where we are today, I was talking to Jensen in the green room we're 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. And the more complexity that you have, the more security challenges that we're going to have to face. So massive opportunity. It's a sustained opportunity and it's certainly not just kind of a peak of what we're seeing. Now 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 built a big platform company on EDR and then obviously in additional technologies. But 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. But if we think about the velocity, when we launched EDR we saw a 10x increase 3 quarters from launch. Now a different time, it was a lot earlier in the life cycle. But what we've seen with AIDR is a 79X increase in uptake just after 3 months. A massive adoption. Yes, we're a bigger company. It's a different period of time. But I like 79 and I can do math pretty quick. And to me, that is a ski velocity. So what's driving this? Why is it happening? Well, AI has evolved -- if we think about the non-agentic adversary that we've been fighting for many years when I started the company, dwell times, we're in the months. expertise, you need nations that need to have high expertise -- the cost was high. The success framework was a 1060 in terms of minutes. Now the agentic adversary has really changed everything, talking about seconds minimal expertise cost is negligible and it's in real time, right? The speed of inference, you saw my keynote yesterday. So 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's the agent state. And 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's a activist or an e-crime group or an agent, they are all in the same category. They all have the same level of expertise. So let's talk about Cyber's next frontier. Where is it actually going? For us, we made an announcement yesterday. We talked about the cyber security Super Intelligence Lab. And why did we create this Well, it's important to realize like where the world is. You have Frontier labs, which have done a tremendous job. We partnered with them. They're changing the world. But again, the success criteria is a little bit different. If you look at what they're focused on building incredible models, selling lots of tokens and solving lots of broad problems, fantastic for society. Crores Frontier lab is really focused, solely focused on solving today and future cybersecurity problems with Frontier caliber technology from a model and a harness -- and that was part of our Safe Mine announcement, which we're going to get to. So what's important to realize here is where the technologies evolve what we've done with NVIDIA, what we've been able to reproduce in our labs, and we'll 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 didn't. So let's jump into this. a little bit more. So Super Intelligence Lab is totally focused on Frontier AI research, open weight, open source model training. We've been big proponents of these technologies. And again, open source and open weight are actually 2 different things. If you look at Nemoto from NVIDIA as an example, it's actually open source and Open weight. Some of the other models are simply just open weight. So use a combination of these is fine, but a lot of it comes down to the harness engineering, right? What Jensen said, the exoskeleton of what makes the model work. And we've seen incredible advances in the harness engineering that we've 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. So why are we trusted with this data? Why are customers looking at us to solve some of these problems at the scale at model scale? We generate 7 trillion security events per day, which is a massive amount of information. So we are a net creator of security data. And you heard me a couple of quarters ago in the SaaS apocalypse, saying, like if you are a net creator of data, you're going to be very sticky, right? The data that we have is not available. You can't go to Red it and go find Crown Strikes telemetry and security data. So 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, right? And that doesn't count a whole bunch of other areas that we're focused on. And the R&D investment over $1 billion, you see the ecosystem. I'm not going to go through all the names. But we're in the right place with the right people and the right technology at the right time. All right. So why does cybersecurity need its own model and harness. I think hugging face was a great example of why the defenders were out matched. And 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 nascent state, it might have been a different outcome. But when we look at the frontier models, they're really good at general knowledge, finding all the possibilities, but they can be very expensive. And this very specialized model that we built, we'll go through the red and the blue. It's cyber specific mastery. This is very important. Our models are not designed to solve the world's complex math problems. They're 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 the question 3 times and got 3 different answers. That's not great in security. So you have to have deterministic outcomes and 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's in the Falcon platform. And you know what they don't want to do, they don't necessarily want to ship it out and pay for that; and two, have it in some other place that maybe they're not as comfortable as having it in the Falcon platform. So these are very important elements of why customers -- they literally moved us after the announcement yesterday on the models and when they can get access to them. So that's why we announced cybersecurity front, first frontier model and harness family combined as an [ Ingenix ] system, which is safe mine. It's 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. And 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 and solving security use cases. And we did a lot of engineering with them. We did a lot of work in the training. We 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? So 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. And 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 become a distribution path for the Frontier labs as well. Trusted partner, trusted data, data sovereignty and 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 and finding ways to attack. So you need a red model. And 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? So to be fair, the industry has been focused on finding all these vulnerabilities. And 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 sort of security offensive capabilities at nation state level. and find ways to attack. But at the same time then, you must have a blue model. And 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. And what we've been able to do is to create a blue model that actually learns from the red model that detects threat stop breaches and user cyber tooling to stop the attacks. And this is one of the areas, just think about the hugging face incident. And 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 attack 17,600 attacks that they saw, looked like in nationstate. They try to piece it all together. They went to the frontier model and it refused. Refusal rates. This is a problem, okay? So 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. Now why is this important for us? Well, 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 showed 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 Barry on patching. So 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 just as one use case. I mean it's an unlimited number of use cases, 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 math of what's running. We know every system, we know what's on it. right? We know the versions, we know the vulnerabilities. So 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. So very excited about SafeLine. Now what does it all mean? And these are the benchmark testings that we did. So let me just take you through. We have Frontier Lab 1 and Frontier Lab II and then we have Blue Solano. You can see Blue Solano had detection rates, 37% better detection rates than Frontier Lab 1 and 29% better than Frontier Lab II, okay? 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. So 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. And then when you look on the blue side, it was up to 99% more cost effective. So I'm going to let the numbers sink in because in your travels, I'm sure that you're talking to customers, you hear -- 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 like amazingly how amazing how expensive it is. So 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. And that's what we're delivering. Better value. Now we look at detection speed. But last I checked in security speed matters, right? Matters in racing and it matters in security, 2 things that I like. And we are 6x faster with safe mine detection. Now 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. So 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. So 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 CrowdStrike. This is exciting. There will be token expansion packs where you can buy more tokens and it's all supported by Falcon Flex. So I talked about Safe Mine 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're 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, but remember, we started [ QuoteWorks ]. 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've created here. So we're going to first work -- if you want to just get access to the models directly and outside of the Falcon platform, you're going to be able to do that, but you need to be part of the Quote Works program. So this will be in review, and we'll be rolling it out over the back half of the year. So again, you got to make sure that you've got the right partners, you got to put the right controls around it, and we want to get the right feedback. So we're excited about this, but we want to be very thoughtful about how it gets rolled out and the folks that we're working with. So that's availability and monetization. Okay. [Presentation]

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. And again, I think when we look at the next active 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. And again, I think the sky is the future when we think about how many agents are going to be available and how much environment will actually change over time and how important these models will be to our customers. So 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. And the reality is that's today or 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 and 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. So for me, that massively expands our TAM and part of the reason why we pioneered AI DR, right? AI DR AI detection response is a new category, right? The product is Guardian, but the category is AI DR, just like EDR was a category. And AI DR is focused on visibility, discovery, where are all my agents. One of the biggest challenges that you're going to find when you talk to customers, they don't know where AI is taking place. They don't know where it's actually being consumed. So first, you have to find it, understand what shadow AI is, who's hiding cloud somewhere, who's hiding this? Who's 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. So Guardian AI agent security is something that we've worked on. I mean it was around the clock that we worked on this with some of our biggest partners. Right? Some of our design partners. So we worked -- our biggest design partner on this was Amazon. You heard from CJ this morning if you're in the keynote. So this is something we've 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've seen so far are incredible. So let's talk about Guardian and the Agentic security life cycle. So an agent gets created, well, you need to know where it is. It accesses data, and it has permissions, it has an identity, right, 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 life cycle and provides visibility and traceability on exactly what it's doing. So think about a financial institution. You can't go back to a regulator and go, well, I don't know what happened, like it was some AI thing. You can't 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 -- so obviously important from a security perspective, but it's going to be mandated from a compliance perspective. There's 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. It will be a must-have technology. Number two, what does it actually do? It discovers the agent, it controls and observes the agent run time, run time. This is something you've heard for us, what, 15 years, the agents run and they're commanded by a prompt. And then it detects and stops threats. So one of the cool things that we've done is we've now combined the prompt layer into the run time layer. And 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're actually looking at the run time and understanding what's happening, you're going to miss it. So 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 may be a browser. Maybe it's a electron app, which is how AI agents -- sorry, how AI applications are consumed by cloud or chat PT all the way out using identity to where that agent is actually talking to. So what does the future look like? The future of endpoint, and I say endpoint, and I'll say workloads here is cloud and endpoint. And I've talked a lot about endpoint. But the reality is we have a lot of customers actually running this in their cloud workloads. So I don't want you to miss this point. It covers endpoints. You're on your cloud, you're on your codecs covers you, no problem on all the other applications. But you do have customers that are now running these agents in cloud containers, a femoral containers, where we actually instrument those too. So we work in both environments. And oh, by the way, there's a third one called SaaS agents. We instrument those 2. So you have 3 areas, my focus has been on endpoint just to kind of make the point, but we're covering all 3 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. So you combine that with identity, next-gen identity, which is our nonhuman identity for agents, 0 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. And then obviously, Guardian wraps this together with visibility protection control. And this is really the future of what endpoints 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 Flex -- total Flex value, 101% growth. Last quarter, 935 new Flex opportunities added you can see the growth there. It's been something that quite -- it's been remarkable, the success we've had with it. And again, I call this the commercial harness. So much so that everyone in the industry is coping flex, they don't even bother to change the name of it. But 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 ecosystem like the marketplace is to be able to deal with this and you have to have the systems that actually work. So there is a barrier to entry because you have to make all that work before you can just say you have Flex. And this is why we've driven the friction out of this model. So what's our next chapter? Well, I think we're in an incredible time period in technology. I mean 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 6 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. So this is very important. So 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. And by the way, we've opened it up, we have competitors here. It's been an open conference. It's been incredible, but you have to have the right ecosystem, and you have to be around long enough to realize how important the ecosystem is. If you just -- [indiscernible] 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 and as I said, at 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 and have them been solved in the past. So 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 spikes is 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 there's going to be plenty of opportunity to solve. So 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.

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 and 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. So we're going to make this interactive. And it's an honor to have Ash, who leads the cybersecurity go-to-market anatropic 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. But 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.

Yes. First of all, thank you for having me. It's a pleasure to be here. I joined the therapy a little less than 3 years ago. And back then, we've 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 didn't have the household name that we do today. And a big part of when you think about our mission and you think about what we're trying to do and what's at stake, we want to usher in transformative AI in a way that benefits humanity in the 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. And one of the kind of principles, I think that guides a lot of our decision-making is like building with the best building blocks. So reinventing the wheel in an area like security that goes so deep and just doesn't make sense, partnering with the best partners does make sense. I mean you guys are essentially the operating system for how we run our how we secure our company. And without -- what I can say is like we use Falcon across the business. I mean, indeed, the benefits of like the in security, all the threat intelligence and telemetry that comes off of that just makes us -- allows us to be able to scale really, really effectively while remaining incredibly secure.

Cloud next-gen, you guys really -- you really use the full stack I think one of the questions that I heard a lot earlier in the year and maybe one of the questions that's on everybody's minds here is like, why didn't you guys just Cloud code your way to CrowdStrike. That was when Frontier labs really came out and came out with some really cool harnesses and I think the technology really evolved. That was sort of one of the questions of the day. And so what better way to answer that question than actually ask the source.

Well, I mean, first of all, I'm like legally obligated to say clock code is amazing and very or everything. But at the end of the day, I mean, in all seriousness, it's -- this is a very effective development productivity tool. You can build really, really cool things and 1 shows some real apps with cloud security. But again, going back to -- we use CrowdStrike ourselves to secure our business. This isn't something that a [ cladcode ] or even a team of very talented engineers working company like Anthropic can just build overnight. This takes time, skill, distribution network effects, a level of determinism that comes with experience. I mean these are hard fought things that you've learned over decades effectively. And you can't build a product like this. A lot of this is just inherent to like the time it takes. So when we think about what CrowdStrike is for us, it's like it's the operating system for security and cloud code, again, while useful in building some things, it's just not -- it's not going to -- we're not there.

And the data is also something that I think we talk about too, maybe your perspectives on what we do over there.

Absolutely. Yes. Again, going off like the telemetry and the -- that you all have the data that you've gotten. I mean, this is proprietary data. This is -- George was talking about how these models are trained. Obviously, there's a lot of inputs that go into these things. But a lot of it is public knowledge and working on helping these models get better at reasoning over time. You can't reason your way to building something like a crowd strike. This is something that, again, takes a lot of like industry skill expertise and the proprietary data that just can't be amassed in other ways.

Perfect. Well, I think that's really a clear way to capture the question, the creative question of, can you cloud code your way to CrowdStrike and you heard the answer right there from Anthropic themselves. So 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's maybe go there next. And we'll come to meet those, but like a lot of the things actually predate dose. So why did Anthropic start working with CrowdStrike? Where did that journey start from your perspective?

Yes. I mean around the beginning of last year, our Frontier Red team and our post-trading team started to realize these models were getting more and more capable or technical tasks, coding was like the first breakout use case that we had late 2024. And over time, in early 2025, we started to realize these models are starting to get be I don't know if I would use this term capable, but knowledgeable, Sentient and aware of how these things work. And we just -- 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're dual use, right? There's offensive capabilities as well as defensive capabilities. So in the wrong hands, these models could 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. I mean, you were one of tiny handful of companies where we 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 and try to do all this ourselves. We also don't have the expertise that you all have in security. So we work 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 like making sure that these -- we tip the scales as much as we can in favor of defensive use. And 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 like the ushering in transformative AI in a way it's beneficial to humanity, we need to -- 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 work with -- we build for builders. So working with CrowdStrike to build -- help you all build solutions that are powered by 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 a lot.

Perfect. Let's go back to April of this year where we all encountered the Methos 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?

Yes. I mean the methos moment was kind of like an inflection point where all of the sudden again, we knew the moment was coming. It was just a matter of how big it would be and when it would come as the outstanding question. And when we're close to finishing training meet those, we realized wholly held. This is here. So there were a few things that we wanted to make sure we got right with launching ethos. Number one, again, like our safeguards needed work. We needed time to catch up and these models were generally made safe. And we needed to get the word out to the world, and we also needed to work with our closest defensive partners to help you all hard indoor systems and prepare your customers so it was very fun to work with you all in those early days. I mean one of the reasons I think we also partnered with you was not just like the business benefits and the distribution but also just the quality of your engineering teams, the product team researchers. This is not an industry that's known for moving slowly, right? So we can't afford to partner with companies that just can't keep up, and you guys have been tremendous at the pace that you've been able to offer.

And you guys have also been super helpful and Quit works our program over there, where our -- 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. So that's 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're doing and things we're 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?

Yes. I mean I think we just announced the marketplace work we're doing together. So 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 strength of our partners, the network effects, the reputation of the trust. But we also want to make it good for our customers -- or for the end of mutual customers. Removing friction, aligning commercial and operational incentives. So these things all matter a ton. And what having CrowdStrike products available in the marketplace allows for on the entropic marketplace allows for is a couple of things. Number one, no longer are our mutual buyers kind of doing the fight of like, is this an AI line item or a security line item. Now these things are kind of merged in source Exactly. You can -- if you have a commit with Anthropic and you have a commit with CrowdStrike, you can use a CrowdStrike product on the therapic marketplace and burn down our simultaneously. So hugely beneficial to our customers and really unlocks a lot of speed. It also -- I mean both of us are thinking about net new logo growth. And in particular, now we have the benefit like 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 like frontier model intelligence and lean on you for their security needs. They get it all in one place.

Yes. Today, we actually announced that today the marketplace go live, which opens -- I mean, Anthropic has done a fantastic job building the premier Frontier lab business in the last couple of years. It opens all the investments that companies have made in anthropic to utilize that investment on CrowdStrike purchases. So we're really excited about that. And I think you guys are very forward leaning in how you're engaging with the ecosystem, engaging with us. And then even at a product level, the build-out that's going to -- that were 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 AP IT 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 absolutely do well. That's exactly right.

Perfect. Great. Lots of opportunity there. And again, ASH was here. We started our week with our Partner Summit. Partner Summit now has over 2,500 folks there. It's the folks who are on the floor or 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.

Yes. I mean it's -- again, we talked about like the commercial benefits of working with the ecosystem. I mean it's a good business for us. It's a good business for you. As far as Beyond that though, I mean, I think technically and operationally -- again, like I said this, I think the other day, which is like if you want to go far, you don't go it alone, right? You need to partner with companies that you trust to have shared admissions and rent kind of 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. I mean the Anthropic marketplace itself is very young. You're one of our. .

It might be the first security partner.

The first security partners in there. So very happy to have you there because it's going to become such an important part of how we go to.

And I think underneath this 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. And so 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. And it's good. That's all goodness and opens a lot of doors for you and opens certainly a lot of doors for us. So 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. There's a lot of choices. We have a very special relationship. And what is it about that, that kind of you want to share with this audience that makes CrowdStrike unique and special?

Yes. I mean these are strange times, right? Like the early Anthropic, we used to say things will never be killed again. It's definitely proven out to be true. And in moments like this, where it's like uncharted territory, we're all figuring it out together. You really just -- you have no chance of winning if you're not doing it with folks you enjoy working with who are like-minded, who can operate at the same speed. So I would break it down in like the business parts and the engineering or technical parts, the engineering, like the savvy and the speed and the scale that you all can operate at gives us a lot of wind at our sails. And then on the business side, again, it's like the fun, the network, the relationships. I mean, I'm a salesperson by training. So it's my favorite thing of the world to come to things like this and hang out with folks like you and George and team and meet customers and alike. So yes, hugely -- a huge fan of being able to do this together.

Well, thanks. And I think we certainly find ourselves in a lot of the same hallways in accounts. And it's good for us to be able to help your customers optimize and use their Anthropic spend and unlock more value. And certainly, it's 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 can't cloud code your way CrowdStrike. And with that, thank you, Ash, for the time. Thank you so much.

Appreciate it.

Please welcome Vice President Enterprise AI of NVIDIA [indiscernible]; Vice President Business Development of [indiscernible]; and Chief AI and [indiscernible] Officer of CrowdStrike [ Barley Richardson ].

Okay. We've got a good exciting panel here. bringing another aspect of diving head first into AI to the forefront, a great crew. And we're going to talk a lot about Safe Mind here next. And this is really the crew behind from an ecosystem perspective behind the announcement of Safe Mind yesterday. Barley, I'm going to start with you. So 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 Safe Mind?

Yes. It's -- and we kind of took a very straightforward path, like the premise and the setup is fairly simple, right? I think we all heard from George last year talk about cybersecurity, AI and cybersecurity super intelligence. And so we start from that premise, right? And I think we've seen a lot in the news we've seen about models, agents, harnesses, all the tech we can talk about being used for all these offensive capabilities, right, like able to evade, able to attack. And so when we set out to think about Safe Mind 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 the Defender, right, like versus the attacker. So Safe Mind is about harnessing all this power and everything that you -- that we hear about, which is valid and these models have real capabilities, right, like to find vulnerabilities and exploit. But it's taking that and harnessing it and turning it back towards the advantage of that Defender.

Justin, I'm going to come to you next. NVIDIA is a key part of how we were able to productize and deliver Safe Mind, specifically with Nemotron. So maybe walk us from the audience perspective back a little bit, like 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'll come back to open source a little bit later, but take it away on Nemotron.

Yes. So our view as AI has advanced is there's 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. And so 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-advantaged way to run AI in these specialty use cases like cybersecurity. And 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.

And certainly, open source is something that's near and dear to our hearts as well. You guys are active in that active. 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.

Yes. Well, this all starts, it's funny with the models. And I think the conversation stops there too soon. The harness is really where the frontier is happening right now. And we also, I think, created this program together, we call it the Open Secure AI Alliance. And that was really to teach the industry when we're talking about auto run on these models to do autonomous long-running work, there's a model and there's a harness. And -- and so there's 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, like where does that probabilistic system meet deterministic controls. And the only way that the industry can understand where these agents are going is if we all share traces from these agents. And if there's a lab leak, you'd 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. And 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.

And that's certainly the journey that we went on and the value that we see in Nemotron. Now all this needed a home. And Sean, I'm coming to you next. We needed a performant AI Neo cloud with the right GPU stack with the right inference speed. Can you talk about the rise of CoreWave, what you guys are up to and a deep partnership that you have with NVIDIA as well that really brought us together?

Yes. So CoreWeave is a purpose-built AI cloud. So we have more than 50 data centers and growing. We offer the latest and greatest NVIDIA hardware. We're predominantly 100% NVIDIA. And 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 long-time partner of CrowdStrike, where CrowdStrike, I would say, is our #1 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. And it's -- 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 Safe Mind and what's been produced and released over the past year is 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, need the home, CoreWeave as the Cloud, partly 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 Safe Mind is capable of.

Yes. It's really encouraging. And like you said, there's a lot of moving pieces and there's 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. And the core premise there is that the best defense that you can make must -- it must be informed by the best and the most and many permutations of offense that you have. That's driven and accelerated by also CrowdStrike's wealth of data in all of these areas, right? Like we have over a decade of experience in both red teaming and offensive capabilities on 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 remediation -- 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. And Dan, to your question, what we see is, one, it has to be that first, you must be accurate, right? Like so if you're going in and you're using a model like Blue Solano to automate your defense and remediation, it has to be accurate. It has to do what you wanted to do. And 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? Like -- and that's due to all the fine-tuning that we were able to do, like Justin was saying, we're able to take a model near or at Frontier capability, use the data from CrowdStrike and blow a path right like this. So a 70% increase. So at first, it must be accurate, right? Like it must do what you want it to do. And then second, you really quickly realize, well, it must be cost effective, right? Like if it's accurate, you also must be able to run it at scale, at speed because these defense and remediations, you're not doing it once, you're not doing it twice. This is continuous. We're in the world of continuous defense, continuous remediation. And on that front, again, versus you go out and go to the leading frontier -- go to the leading frontier capable model in the leading harness, we are 99% cost reduction from that. So to give you an example of that in dollars per remediation, right, per detection that we go and remediate, that's $10 before, and that's 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'll just say it's 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 kind of the relationship they've built sitting down with each other, talking about where the industry is going and talking about how we make the world a safer place. And 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 5 layer cake. That's really because power is the limiter. And then you want to make sure that you can generate tokens at the lowest performance per watt. And 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 Core Weave -- we can do it at scale, it can run anywhere, any cloud, any data center, any infrastructure to deliver like the lowest latency results. And I think the shared mission across the teams is we want to put the defenders at a differential advantage. And so kind of with that is 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's going on, both with human attackers and now agentic attackers. And so we working together, I think, can build the best, most cost-effective, most widely available system. Totally agree.

And it's really not like a, oh, here's our drop, -- good luck, guys. Feel free, Barley to jump in and even Sean, because you guys work so closely with NVIDIA as well. when you're at this level, it's not like, hey, we're going to meet next Tuesday or we'll see in November. It's a full contact sport. I mean that's how Jensen is. So it's a full contact sport, maybe some insights on that collaboration. It's a mosh pit in all the great ways, right? Like we have all of the experts, I sold that from Justin, right? Like I told that comment from Justin. But it really is that. Like that's the speed you need to move at, right? Like that's the speed we must move at. It's to Dan's point, it's not like, yes, it's got on a call on Tuesday, and we'll work through these types of things. And then it's not like, oh, let's get on a call and then we'll have our engineers on this call, and then we'll have our product people on this particular call. No, we all get together all working the problem like in real time, right, like in solving it in real -- there's been many days where I've been on the phone with Justin or I've been on the phone with or we or like our engineers have been talking in real time and we'll go to headquarters. And I think that's the key. One is like having all the people there, like in the right place to do it, and that can virtually in the right room, white boarding. But it's not a, we do this at this cadence. This is the job, right? Like this is the job, and we're doing the job all the time in this very highly collaborated all the best ways of mash pit full team sport. Sean, let's cover our partnership together. What we're doing together I think we pioneered a very new exciting type of partnership in the neo cloud 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 Core weave and then in the market.

Yes. So we've grown extremely fast. I actually came through an acquisition of a company called [ Waybiases ] to CoreWeave. And 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. And deploying CrowdStrike across our estate was sort of the first step. And I think as we see some of the newer solutions that CrowdStrike is 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. And then beyond that, now with Safe Minds, 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. And so 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.

Yes. I think of CoreWeave, I think it's an important point to make here as really a net new greenfield attack surface. I mean a couple of years ago, there weren't -- nobody had ever heard of a Neo cloud. There was no such thing. And 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 Neo cloud. So it's really important. I mean, none of it would be possible if it wasn't for the foundational layer that NVIDIA has 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.

Yes. This is all net new as well, right? Like all of our customers use hyperscalers. This is just an expansion of the market. And so we really see a huge opportunity.

So 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'd love to get your thoughts on why we're on this mission together, why our companies work together. There's I don't know, 3,000-plus security companies in the world. There's 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 [indiscernible], because I can get a hold of you. Anytime I to. 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 #1 cybersecurity partner because you guys have this vision for how to secure the world's infrastructure. And 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. And 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 whole -- a lot of great people and a lot of opportunity. And 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. And I think SafeMind is a great example of that. It's the next chapter for CrowdStrike. The fact that now we're a frontier lab, cybers first frontier lab, and the honor of building with and on NVIDIA and having the home for that be Core weave 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. [Break]

Welcome President of CrowdStrike. Michael Sentonas.

Good afternoon. 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 Bert at the start, we'll 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've spoken to a few of you. So I know some of you have, which is absolutely fantastic. What Falcon for those of you -- and I know some of you are here for the very early days back in San Diego. Great to see how this -- how big this has become in such a huge community event. And 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 2-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's about 30% of our weight to our $20 billion ARR target we provided last year at the Investor so I'm going to unpack some of this and dive into a little bit of this and talk a little bit about how we're going to get there and just give you a view of walk to how we get to that $20 billion 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. Now 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 team cloud. Key drivers of our $10 billion to $20 billion ARR targets, which have performed incredibly well. I'm really proud of the way that the team has performed and we shared numbers across those areas last week. Just to kind of take a little bit of a step forward as we talk about this path to $20 billion. Let's just start off with the key information straight up. Today, I'm excited to show you that we believe we can reach the $10 billion ARR target within FY '30. Similarly, we believe we can reach the $20 billion target within FY '35 and I'm going to talk you through the path to get there. So 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 go through with you every year continues to get bigger. It is a staggering $565 billion market by calendar year 34. 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 big. And these are the areas that we play in and technology that we believe we have a right to win in, and I'm going to talk through. Let me break this down a little bit for you because when we think about that path to 20, how do we get there? We talk about market share in core endpoint. If we look at how we're 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. But then 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've covered -- we've captured roughly 4% share of the security market that we address for the last 6 years, start to play some of that math forward on that path to $20 billion. What do we need to capture in calendar year 34 on that path to $20 billion. So if you start to think about that math, we have to -- we only have to capture 3.5% of the market in calendar year 34 to hit our $20 billion ARR target despite the fact that we've averaged over 4% market share every year over the course of the last 6 years. We obviously strive for a lot more. But I'm just talking you through an illustrative path to that calendar year 34 time frame. Let me jump into the product and let me unpack some of the opportunity by product. So jumping into this. Let's start with the genetic security. We're going to start there today. We estimate this market will reach $115 billion in calendar year 34. 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. And 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're working on. Hopefully, you saw some of the demos where Guardian is working with other products, you're 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 sort of 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're doing in cloud in codecs have a look at [ Kyro ] to have a look at the open weight models and start to make sure that anything that's going on inside the organization that's happening at machine speed, Guardian has the ability to either give you visibility to give you some 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've 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's in our network, tell us who's using it, tell us what they're doing with it, tell us what tokens their spending and help us get control. And we've answered that question, which is absolutely phenomenal. The architecture that I stepped through as well today, and I'm just going to touch on this a little bit here to make sure that everyone caught that or if people weren't in the session is really important because we think about AIDR as a category when we've been talking about AIDR, 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're not going to stitch this together with 10 different products. It's going to be very hard to be able to do that. That's why we've 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's 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 platform on the planet, I should say, if I can speak English this afternoon. We'll start to want to bring in data protection solutions into their organizations. Run time guardrails, response, SaaS, endpoint cloud, everything needs to be part of the capability here that we talk about. So 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's super clear. I want to roll the demo again because this is a pretty cool demo and I'm super proud of this one. So roll the demo. [Presentation]

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've got a flood of customers last night saying, how do we get it? Falcon Flex, call your count rep asked for the module to be flexed in and you're using Guardian. It is a license 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 run time. 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 the endpoint, you're leveraging identity, you're connecting to those models. We give visibility and control into that whole thing. We have the data in 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 '34 I'd like to 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 0 in this market, we have nearly $700 million in ARR from our Next Gen C business. So to capture 1.8 of that ARR in calendar year 34, we only have to grow our annual recurring revenue at a 13% CAGR, which is very achievable, I would argue. So as you can see from this slide, we're currently growing that we're currently growing our ARR by 60%. So again, just playing that out, so you can see an illustrative example up through the calendar year 34. So I talked a lot about 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 0 today. So the reason why we talk about this is attacks that way for humans, which means the solutions that we knew that we use, that organizations use to protect themselves need to discover this decide and move in real-time speeds. And if the breakout time does get to 0, if it's happening that quickly, the time to investigate will also go to 0. That's why the old stock 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 AgenticSOX solution built on next-gen same. Roll the demo. [Presentation]

We're faster. We're 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 architecture, 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-gens enabled. License entitlement to turn it on if they want to use that. And importantly, we do not charge them for the data that's already native to CrowdStrike. We also given capabilities like federated search leave data where it lies and give them the ability to search across data source. So architecturally better, smarter, faster and a much more disruptive price opportunity. So let me sort of touch on the identity market. We estimate this will be a $48 billion market in calendar year 4. Again, playing out of the math. If we capture 3.5% of this market, this will equate to about $1.7 billion in ARR across an entity, privileged access management and so on. And as I said with NextGen [indiscernible] this is not a market that we're starting at 0. With nearly $585 million in ARR from our next-gen identity solutions. And today, we launched Agentic identity provider, which means we're well positioned to grow our ARR at a 14% CAGR to reach $1.7 billion in calendar year '34. As you can see by the chart here, we're actually growing at 33%. So you can do the math across this. So I want to walk through a little bit about how the solution works, how it's a little bit different. And then you can see how the solution is going to win in the marketplace. And 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's 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's 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's 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 hand-off complaints and make sure that you remove all the risk inside the organization. That's why today, we announced CrowdStrike Agentic identity provider solution, which extends continuous identity discovery enrichment and z-standing access to all agents. I hope the demo works this time. Roll the demo. [Presentation]

So 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 single AI acquisition that we made at the start of the year gives us incredible power and opportunity to make sure that we enforce 0 standing a privilege. This is mandatory to solve identity problems today. So 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 '34. If you recall from George's presentation, we 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 million to $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 '34. That $7.5 billion is roughly 38% on 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 in calendar 34 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 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 out 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.

I heard a lot of things today. I heard one of the front tier mall 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 mean I knew that, but something else when somebody else frontier guy says that. We heard from our core -- we 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 its 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 is ready for what's coming, right? All these amazing innovation that we're seeing in the world today needs to be secured, right? CIOs are panicking because -- their CEO is saying, we're AI, 89% of companies that George talked about are touching AI, right? They need to be protected. And guess what? We have Guardian. All right. Let's do a quick recap of what was talked about today. There's a lot of opportunities that we talked about AI adoption is storing. We talked about the stat. You guys see it. We see it every day. We see it every day in the news. The tier, it's now it's happening at AI speed -- it's just incredible what we're seeing with respect to AI and adoption of AI. George talked about building cybersecurities. First, model, purpose built for defenders. I mean he was on stage with Jensen and the 2 of them were going back and forth about what this could mean. I mean Jensen's 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 a ahead of the curve type of company, innovative, thoughtful knowing where the puck is going, if you're a hockey fan, I mean that was incredible what I saw, right? Securing agents, George and Mike to talk about securing agents. We talked about -- George talked about, hey, for every person there's potentially 90 agents. Then George talked about, well, could be more. could be 1,000, 10,000, whatever number is. And I agree. I spoke to one investor after our Q2 earnings call, and we went through that and we said, hey, 1 human to 90 agents. And there was a pause. I as a very big investor. There was a pause. And he goes, "Can't be 1,000 or more?" And we're like, yes, 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's required in today's world. And we're showcasing it here at Falcon in front of 10,000 of our partners and customers. And 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's what I saw. And we talked about the endpoint is the control plane. Look, AI is consumed at the endpoint. We've shown you over 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's no way in the world today who does [indiscernible] better than us. And now again, something has come to us that requires the end point to be able to digest all this activity that's taking place in this new AI world. And then Mike just walked us through this incredible TAM, right? You walked us through on a market -- by product, 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's a big number, right? And we have the tools, we have the technology, we have the people, we have the AI to get there. The time expansion is going to come from inorganic and of course, organic as well. So we have both of those avenues go after that big TAM. And I think we're in the pull George likes racing in we all like racing in CrowdStrike, full position. We're in a full position to go get that TAM. Sorry, guys. So 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. I mean each one of those things to me, screens this conviction towards our goals. And I'll share some more of those goals with you a little later. All right. Probably worth a minute just to take a look at the Q2 highlights. It's the $333 million in net new ARR, 51% year-over-year, an all-time record for us. Non-GAAP operating income of $372 million, an all-time record for us. So you've got top line growth and all time record. You've got profitability at an all-time record. For me, this was the greatest quarter in CrowdStrike's history. And 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 so I read them all. We all read them all. And 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's talk a little bit about Fal.Con Flex. I said it at zillion times, but it's worth repeating. It's a commitment model, not a consumption model. George talked a little bit about it. We worked diligently with our auditors and everybody else and how we're -- and obviously, our customers to be able to get what they wanted, right? I'd love to be able to take credit for Flex, but it's 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 [indiscernible], 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. And you've heard many times about why it's successful -- you've heard many times about you can track once and then it's another PO just on the reflexes. And I've talked to each of you after our earnings call, talk about the fact that, hey, look, if this is being consumed faster and you've seen the data then how fast it's 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." And by the way, I'm going to give back a bunch of though that I'm spending on somebody else. So any CFO [indiscernible] their waiting gold is going to take that deal every single day, right. The other thing I'd like to talk about is that Flex has the same -- as the same kind of retention that non-flex does. So ARR and revenue are recognized ratably -- there's no change. There's no mixed models. There's no confusion over how we're doing this thing. It's very simple. There's no change, right? And that's part of the beauty of the elegance of this model. It does a lot of other really good things for us. It allows us to contract faster. They can get a bigger -- we get bigger deals, we get longer deals, able to kind of reflect really quickly because they've taken the friction out of the sales process. So there's a lot of goodness in Flex. And as George has talked about many times, it's not just here's the license and everybody can use it. They're stealing the name, but you actually need to have the right -- 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 click on those modules straight away. You need all those things to be able to have proper Flex and see the results that we're putting up on the board. George has shown you these numbers just before you've seen them all, but I'd 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 to 34%. That's a big jump, 34% to 40%. And so when we go to the sales team, you have folks that they're kind of like they like the old ways or they're used to selling new ways. Would you put that stat in front of them they go, yes, yes, I'm going to get them on Flex. Hey, these folks are quite operated, right? So it works really well for the customer. It really works well for us. It works well for our sales team. It all kind of comes together. Let's talk a little bit about what we're seeing with the power of Flex first go-to-market. You can see here the Flex logos as of 2Q '26. We had 61% enterprise and 39% nonenterprise. Now it's 50-50 so why is that happening, right? So 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. And that's additional friction from a sales perspective the sales rep doesn't want to go to Mike for an exception, right? So we are really leaning forward on Flex First. There's friction if you don't go there. And we've already seen all the benefits why you need to go there. So I envision a period of time at 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're seeing in cybersecurity in the world, right? I've talked to -- I think it was Michael, we were talking about the fact that this is a generational shift that we're seeing in technology. You can call whatever you want, generational, there's an inflection point, for me, it's transformational, right? What we're seeing in the world today. And Mike talked about our path to $20 billion in ARR from a market share perspective. I'm going to now talk about the $20 billion from a customer and geographic perspective. And for me, I like looking at things on there are different planes, right? You can do market share. I like customers, I like geos. I want to see how -- I want to see it all. And that's how we build our internal models. Mike talked about the 3.5%. You saw all the numbers, you can all do the math. Given the track record that we've had of capturing roughly 4% over the last 6 years of market share, we feel that we can we can be able to reach the 3.5% and then some. You go to execute, you got to do all those things, but it's there for us for the taking. [ 30% ] isn't a huge, 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 it 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. And then you look at these numbers. And if you start from the group that has a small global [indiscernible] 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. And 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 sorry, 2026, right? We're in the early innings of our journey in terms of going after customers. So we feel we have a tremendous amount of headroom to go with respect to going after new logo. The geographic opportunity. Look where we were in FY '22. 72% in the U.S. and the rest of the world was 28%. 5 years later, 65% in the U.S. and 35% rest of world. I'm excited about our international opportunity. You can see CY '34 international market community 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 capture a whole bunch of that $244 billion by CY 34. The expansion opportunity. So as I talked about we have a lot of headroom in new logos, tons ahead. We also have a tremendous opportunity in the -- with our current base. We were at $5.8 billion in 2Q of '27 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 '26 of $21 billion. We're going to add another $8 billion based on our customer base today. Total white space opportunity for 2Q '27 $29 billion. That's pretty good. So I have this tremendous opportunity in headroom in new logos. I have this tremendous opportunity to sell more to my existing customers who love us already. We 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've been winning. I love this slide. So my good friend, DB came to the company 4 years ago, and I'll be honest with you, transformed our whole partner kind of go-to market, right? We kind of look like everybody else before it was there. We weren't investing in it enough. And DB came along and really transformed for what I saw the partner community, all of our partners, whether it's GSIs, whether it's the partners we had on stage, you can see the numbers. We had 8 partners, each with a lifetime TCV of $1 billion. That's impressive. 19 each with lifetime TCV at $500 million and 65 each with lifetime TCV of $100 million. 65 each with a lifetime TCV value of $100 million. I remember, George, when we hit $100 million TCV as a company. And here, I've got 65 partners each with $100 million plus, incredible. 890 of our partners doubled their business with CrowdStrike year-over-year, 890. So part of our success is nourishing and growing with our partner community. And what I've seen in the last 4 years has shown me that if you can do it right, you can hit the numbers that we've been hitting. As you saw, we -- earlier for Mike, we pulled in some of the dates, some were 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 '31. Now we're doing it within FY '30, similar theme for $20 billion. We said we were going to be able to hit that by the end of FY '36. Now we're going to do it within FY '35. And why are we saying that -- we're seeing significant momentum in the business. The business is firing and there's a huge opportunity. We've got both of those dynamics happening. And for us, when you see that momentum, I can tell you right now that my friend, George he's not laying that momentum slide away. Put on the accelerator, right, George? Okay. Let's talk about the target model. So here's a summary of the target model in full year FY '29. So the ones in green, S&M, R&D G&A, we're already in the band. Subscription gross margin, 81% to our target model of 82%, that's 85%. So we're right on the cusp. Operating margin, we've seen incredible increases in that. We're now at 24%. This is all of 1/2 of '27 to our target model of [ 20% ] to 32%. And I've 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% to 38%. I think we're in year sort of all these things. I feel that this is the same 3 years ago when we said it, same now, not changing it. I'm not extending it out, keeping it as it is. And there's still room for efficiencies across a lot of this stuff, right? AI is helping me 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 into the partnerships. I feel really good about this model. I'll go into a little more detail, especially about gross margin expansion. So many of you have met me for the first time back in 2015 when I started. And when I started, gross margin was in the 30s. And I looked at George and I said we got to do some different things to be able to go public and he goes, "Yes, I know, let's figure it out." So we did. We figured out what we need to do -- 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're going to increase our gross margin and still be able to get the liver the products that are best in breed and then a platform that's best-in-class in the world. And we did. And we went from the 30s [indiscernible]. This is a story that is most near and dear to my heart, right? 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 take and reinvest into our business. Last year, we put $1 billion in R&D, $1 billion. So gross margin expansion matters. So how are we doing it? So we've got our public cloud partners and customers and we did a couple of things with them. One is we looked at where it was more economical to put the compute and the storage. For example, Amazon, we went more to the West 2, there's West 1. We also have scale. So we're able to have volume discounts with our public cloud providers. So that's going to help us more in the future as well. As we get bigger and everyone else, we're able to enjoy better discounts. On the private side, we're 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." So 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. So I feel really good about where we're going with gross margin. Obviously, a point at 81% going from 81% to 82% is a lot harder than going from [ 30, 31 ]. But I feel really good about it. And for over 10 years that I've been at the company, we've 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. So as you all know, we talked about FY '27 having a 30%-plus free cash margin. So now I'm talking about FY '28. And I'm looking at a 32.5% plus free cash flow margin. I feel really good about that. I also look at my CapEx. So we have a slight increase in CapEx going to 11% to 12% next year. And even with that, I still feel really good about the free cash flow margin. So they go hand and hand, you got to talk about both right? And obviously, the prices have storage and everything else have gone up. We're aware of it. We've done a great job in terms of procuring and great job with our vendors and our strategic vendors to lock in certain pricing. So I feel really good about at the end of the day, the 32.5% plus free cash flow margin. Look, 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 TAM that I talked about is very achievable. It's a large, it's an expanding TAM. Large and expanding. Those are really good words. And I think we're right in the epicenter being able to get it done. Conviction in FY '27 net new ARR acceleration. Let me walk you through kind of some of the history. When I was here last year at this time, I said that for this year, we're going to do 20% in terms of growth on net new ARR. I think you were all here for that. Then at the 3Q '26 earnings call, I said I'm going to do 20% but on a bigger number. You all remember that. Then after 4Q, I said I'm going to take that 20% then raised it to 22.5%. I felt pretty good about that. And then we had a big jump in 1Q, right? 520 basis points. I was really happy about that. And I said I thought it was a prudent guide when I did it. We're looking at all the factors. It's not to execute incident, but we did. And then, of course, last quarter, 34%. Overall, total raise to date, $220 million FY '27. I'm really proud of that. And a lot of things needed to happen for that to work. And now as I think about the future and all the things we talked about here today here's what we think about, I think about this. I think about 20% plus year-over-year net new ARR growth for next year. That's what I think about. We told you that at the midpoint, we're going to be at [ 1.355 ] in -- at the end of this year and the next year, [ $1626 million ] [indiscernible] great at that. So that I think that summarizes our conviction in where we think this business could go. Pretty exciting times for us and I'm willing to go out again at Falcon, which is well before end of Q4 to talk about next year. So with that, I hope that this has been enlightening for everybody. I hope that it's been transparent with everybody. I hope that it's been consistent with what you've seen in the past. So with that, we're just going to get ready for setup for Q&A. So stay tuned. You can take a bio-break or whatever and come back in about 5 minutes and then we 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; Financial Officer, Burt Podbere; and Chief Business Officer; Daniel Bernard of CrowdStrike.

All right. Thanks, everyone. This is my favorite part of the day. So we have 45 minutes for the Q&A. If you could please raise your hand, I'll get you in and then wait for the microphone because we have a lot of viewers on the webcast that also want to hear your question. And then if you could please state your name and your firm for the webcast viewers. That also be helpful. So why don't we get started? Let's go first here with Brian.

Good afternoon. Brian Essex from JPMorgan. Thanks for doing this again this year. I'll say you have to be here to appreciate the scale. I don't think I have to sit in the top tier of arena to watch a keynote before for a security software company. That's pretty impressive. But George, I'd love to -- the question on why Anthropic Cloud why you can't cloud code, you waited CrowdStrike strike. I think it was great to hear Anthropic so operationally, I think we've all been reiterating why you can't do that. But 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 you're 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 to 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. I mean 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 in mass, 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, didn't stop anything. Will give you a text explanation of something. It doesn't stop anything. And 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, I mean, operating model than just say, I have an LLM and a harness, right? That's not going to do it. Now 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 edit 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. And then when you put it all together, we have distribution. Like I think sometimes people forget in 2026, what it means to have distribution. I think all of a sudden, now you whip out Claude and you're going to dominate the world. Like, yes, it's 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 obvious 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've got 40 minutes here. So next question we'll take from Saket.