Palo Alto Networks, Inc. (PANW) Earnings Call Transcript & Summary

Good morning. What a great way to start our conference with Nikesh. I know he's going to kill me now.

Why would I do that?

From you, I know after that it is easy.

I would not kill you.

So I'm very happy to host Nikesh and I prepared -- we have about 30 minutes. I prepared a list of questions. And we're going to go -- I wanted to start with something more recent rather than the regular kind of tell me what you do. And if you have any questions at the end, we have the microphone, as always, like every year, we have the microphone around. So thank you, first of all. Good morning.

I want to start -- so I want to start with the word platformization. And I want to understand what is platformization for you? And how is it different than, let's say, Microsoft using the word platform, like the word platform. What does it mean for you? And what does it mean for others?

You should bring the guy before me back, Matt, he works at Microsoft, right? Maybe he can tell us. I should ask him about his?

I'll ask you about other companies.

So look, if you go back historically and think about any functional space, whether it's CRM or you think about ITSM or workflow management, over time, what has happened is multiple applications converged into a set of applications that share data and talk to each other. And I worked at Fidelity when I started my career, I used to be in IT. And we had 26 apps, which we had master tables and secondary tables and everybody had their own tables. We all had our own software, and we try to get stuff to come together. Today, you don't buy 4 CRMs. You buy Salesforce, there's one system of record in the back. You have a bunch of applications sitting on top because that's what you need to run an efficient organization. I think if you look at cybersecurity, it's the newest area of technology. It came about after connectivity from a software infrastructure perspective. So in that context, the industry lacks this backbone system of record integrated converged "platform." So our view is instead of going and selling individual products, these products which need to work together had to be put in a platform together. Now we can talk about various examples, but we have 2 fundamental platforms, one, which we call our network security platform, which effectively secures a perimeter, whether it's sort of hardware firewalls, software firewalls, now AI firewalls or SASE or browsers. And the other one we call is our Cortex platform, which is effectively the data-oriented cybersecurity platform, which allows you to use all of the data in the enterprise and be able to protect against bad actors. I don't know what Microsoft's platform means.

But the question is, you made a lot of acquisitions in the space. Is it about, is platformization at the end of the day today, is it -- the way you convince customers to consume more than one platform...

Yes. One product.

Is it about technology synergy between the products? Or is it more about pricing, bundling and pricing mechanism?

It's the former, not the latter. The latter is a consequence of it perhaps. So take the instance, like let's take a company which started off with the data center. You can buy a hardware firewall. You can buy it from Ken while you just saw in the hallway. You can buy from us, you can buy from Cisco, Check Point. And you decide you're going to go to Google Cloud. You can buy a software problem. You can buy it from Google, you can buy it from Palo Alto, you can buy from everyone. Then you say I want to do SASE, get [ J Choi ] to give you SASE or can get Shlomo to give you SASE on Cato or [indiscernible] They got SASE, you got hardware problems, you got software problems. I need browser. Let me go to Island. What happens, let's assume that's your enterprise. You know this as well as I do, pal. Then you'd have a set of policies you have to set up for the browser with Island. You will have set of policies that have to set up with Fortinet and a set of policies that you will have to set up with Zscaler and a set of policy that you have to set up with Google Cloud. Imagine we set the policy once and have it perpetrate all 4 use cases. A platform will allow you to do that. A set of point products that have been buying from 4 different companies. Now you have one policy playing, you have less errors. Not just that. When you put one policy playing together, your data is talking to each other, which means you ever want to use this wonderful thing I've learned presumably AI, it needs some consistency in the underlying data. You have to have that data be consistent across all those 4 capabilities. So to me, platformization is the convergence of those point products into a single set of solutions over time.

Where are you on the integration of all the acquisitions into this kind of journey?

So there is a fallacy that we bought 20 companies, we must be a raggedy muffin collection of products, which is kind of not true because our hardware firewalls, software firewalls, SASE is pretty much devoid of most acquisitions. That entire network security platform has about 3 acquisitions in there, which we have integrated over the last many years. A lot of our acquisitions are concentrated on the cloud side, which we have just rewritten, which we talked about in the last earnings call. We used to have a bunch of raggedy products that are connected together. We have rewritten the entire Prisma Cloud back end on Cortex and our data platform, which is about 9 acquisitions being collected. So 9 there, 3 here, Xpanse is different. Unit 42 is different. So a bunch of stuff are still stand-alone, which work independent of these platforms. So we're pretty close to being connected. The only one we haven't fully -- we're like 85% there on the browser, and we are 0 on Protect.AI, which we haven't closed yet.

Yes. And -- you changed the name. I mean, there is a strategy behind it, but you changed the name from Prisma Cloud to Cortex, right. Why -- philosophically, why was it important to take Cortex, which was more towards the endpoint and take cloud security that was more towards the cloud and put them together?

So what we've discovered is either your perimeter products in the network security space or your products that leverage a lot of data to solve the security problem. Take -- I'll give you an example. Take e-mail security, right? Email security has traditionally been something that's in line and checks your e-mail, looks at your e-mail and says, "Am I being sent to the wrong Internet address or the wrong URL" It stop you. But if you go past it, if you click on that URL, your e-mail security product is done. You're gone. You have no visibility as to where you clicked, where you went to, whether it's stored your credentials or it allowed you connect to your next AWS server. Now funny enough, we collect all that data in our SIEM in Cortex. So we can actually look at after you went through the e-mail security, you sort of missed the bad URL, you clicked on it. We know you hit the firewall. We know from the firewall, you went somewhere else. We know from there where you went because we have all the data. So we just launched a new e-mail security product, which does all the hard problems that e-mail security products don't solve. So realize once you collect all this enterprise data, they are going to do a lot more security, which over time lends itself to real time because they have all the data in the same place. So in that context, it was important that all the cloud data be resident in the same place as the enterprise data was. So we had to rewrite the whole thing. And it's very exciting. It's exciting because I think it allows us to converge our real-time security capabilities, not just for the enterprise but also for the cloud.

Did you have to change your go-to-market? Or how did you have to evolve your go-to-market when you started implementing platformization? Is it the same buyer, for example, for various...

Yes. Yes. Look, I have many examples where when a customer is a Palo Alto hardware firewall customer or hardware Palo Alto software firewall customer that they see the natural gravitation pull to buy SASE from us or vice versa. They land at SASE, 1/3 of our net new customers at network security are SASE customers. Over time, we get them to evolve into hardware and software. So we are seeing the benefits of a land with a certain part of the platform and expand the rest of the platform. And I think it's hard work and grinding right now. But to me, it's abundantly obvious that in 5 years from now, when you are sitting here or somebody else is sitting here, you'll look back and say, that made a lot of sense.

5 years, I'm still going to be here.

I'm working a long time. Got it.

But I ask you, is it, by the way, the same buyer, for example, for Cortex and for network security? Or are you going after...

It depends. Eventually, the CEO is the buyer or the CIO is the buyer or the CTO is the buyer or the CSO is the buyer. So yes, as deals get larger, I'll give you an example, last quarter, we announced a $90 million deal, which we did across both platforms. The CIO is involved. There's no way you're doing a $90 million deal with a company if the CIO is not paying attention and even the CEO knows the deals getting done. So as deal sizes increase, everybody knows in the organization where it's happening.

Got it. You also started to cooperate more with system integrators. What's behind it?

Well, look, when -- if you look at what's going on right now, many companies finally enough are accelerating their cloud migration because they've discovered if you're not in the cloud, it's harder to deploy OpenAI or Gemini or Llama or whatever have you. So they're all busy trying to chase the cloud transformation. The cloud transformation, the AI transformation, the network transformation, a lot of these are complex transformations, which are now being frontier led by system integrators. You hire an Accenture, Deloitte, Pricewaterhouse, British Telecom. All these people are actually sitting with the CIOs planning how to take them from a point where they are today to where they need to be in the future. So it's very important that the systems integrators understand our strategy, understand the capabilities, understand how much we can offload from what they have to do if they actually use a consistent set of products. So as you would expect, we partner with all of them and are working hard to make sure they understand our portfolio so they can help us implement it.

Got it. Does it have any margin implications for you, by the way? Or no, it's only labor work is offloaded basically to systems integrators?

Look, for the most part, we are a product company primarily. We have a small services business like many of our peers in the industry because that allows us to make sure we're doing it right. We're learning from it. So a small services business, we'll maintain the services business. The idea is to make sure these third-party partners, the system integrators or MSSPs or telecom companies, they do a good job implementing with us. So they get the service business, they get the consulting business, we get the product.

Got it. Okay. I want to talk to you about like one question that we ask all the companies that are here about current spending environment. You made some comments that April was tougher and then things went back to normal.

It depends what normal is here. Yes.

Yes. So that's exactly.

Every week, the normal changes. We're going to put tariffs in Europe. We're not going to put tariffs in Europe. We are going to put tariffs in China.

Where are we in terms of spending environment?

When you see something out of the normal happen, everybody is sort of deer in the headlight. Holy s***, what do I do now? So when the original tariff conversation came out, we all went deer in the headlights. Holy s*** what happens now and the stock market moved and everything moved. Now it's like every morning, you go s***, there's another piece of news about tariffs. The market keeps doing this thing. And don't underestimate how many CEOs, how many CFOs get knee-jerk by the market, right? You see the market go down 10%. You see that, oh my God, this must be true. If the tariffs come, the market is down 10%, I'm going to have to pay $20 for my thing, I was going to pay $5 at Walmart for. So everyone goes into this mode of holy s*** what's going to happen. And honestly, at that point in time, I was with a bunch of retail CEOs the week after the tariffs, and I said, we got a big deal with you guys. It's like, well, actually, we're trying to figure out how many Christmas trees to order because the price of Christmas trees is determined this week and then amount of Christmas trees going to ships, we have to order now 4 months in advance. They're more busy figuring out what the price of Christmas Ties is going to be in October, November than buying Palo Alto cybersecurity. So obviously, the bogey changed in people's head. And all of that stuff is kind of normalized. So people are sort of business as usual. Now I don't know how all the tariff stuff is going to manifest itself in the second half of this year in terms of the macroeconomic conditions, but the initial shock is out of the system. And people still need to go do their AI thing. They still need to go do their transformation thing.

To margin. So that takes me to AI. You announced Prisma AIRS, AI security platform. Elaborate on -- I'll ask you like I ask start-ups. What is the problem you're trying to solve? Meaning what is the solution? What's the position for it in the market and why?

Right. So as all of us have seen this beautiful AI expression, whether it's a ChatGPT expression or a VL3 expression, make a video. We all understand the answers in the consumer context, right? When it gets to a professional context, our desire for accuracy is higher, right? You can go to deep research from Gemini or Google and go ask it to produce a research report on Palo Alto, but it will pale in comparison to Tal Liani's report on Palo Alto, which I have a few suggestions for that we'll talk about later, but that notwithstanding. That is, one way is our desire for accuracy is way higher in the enterprise world than that you're seeing on the consumer side from an AI perspective. So every enterprise company is busy trying to understand how to deploy AI. The first thing I do is to deploy AI is either use a packaged software from the outside, they are legal lawyers using Harvey.ai or you could be using ChatGPT or you build your own. Now if you think about it, I think AI will end up in the same proportion as we use SaaS software versus self-developed software in companies in terms of how much we use our own, how much we take from the shelf. If you believe that, you have to make sure that whatever SaaS software using or it's going to be call AIaaS, AI as a Service, whatever AI-as-a-service software use is equally secure in enterprise. We're not sending our corporate documents into it. We're not getting it data stolen. We're not getting -- we don't have a large drug discovery AI software being trained by every company's proprietary data. You don't want that. So there's a lot of requirements in enterprises to deploy AI in such a way that data doesn't escape because that's kind of the intellectual property. There are a lot of requirements that you can't hijack my AI. There are -- it's -- so I'll tell you this, the most fundamental difference between AI behavior and SaaS behavior. SaaS software delivers a predictable outcome. You ask a question to SaaS app. It does the same thing at 7:00 a.m. in the morning at 9:00 p.m. in the night. Today, tomorrow, it does the same thing every day. You have Salesforce, Workday, ServiceNow, you pick your SaaS application, do the same thing. AI is fundamentally different because you're constantly training. So you expect as your LLM gets trained, your application gets trained, it's going to give you a better answer, which means there is no longer a predictable answer. Now it's very hard to understand the difference between a better answer and not a better answer, but we know that the output constantly changes, you have to test it for security on a constant basis. What do you say, "Hey, what's the formula for X?" And it says, click on this link, the link is a malware link. We have no idea whether that link is accurate or not. Yesterday it wasn't doing it. Today, it's doing it. So in the AI security world, you have to constantly bash your AI deployment to check that it's not creating security risk. So there's something called persistent red teaming. You have to persistently bash against your applications to make sure they're not hacked by somebody else. That's the capability bought to protect our AI. And that's kind of one of the underpinnings of what we're doing with Prisma AIRS, which is AI runtime security. So when you deploy AI, you can deploy it safely by putting a security envelope around it.

Are we there in terms -- I understand the need, I understand why you're doing it and what you're doing. But are we there in terms of willingness of enterprises to look at it, to invest in it, to buy it?

So I was reading this wonderful deck from Mary Meeker who used to be at Morgan Stanley, I got this 300-page deck on AI. And I saw some of the AI stalwarts on the weekend, and they found out that the use of Gemini token is up 50x in the last 1 year. So somebody is using it. It's probably being used in apps like coding assistance for now or it's being used in -- it's going to get used in a bunch of creative use cases. So you're beginning to see the usage. Now if a coding assistant becomes popular, every enterprise deploys coding assistant. I guess, it needs to be deployed securely. You can't just have any coding assistance and at Palo Alto, you have to go through 17 hoops before you can deploy a coding agent to help you because we're scared, if we run security software, if our software gets exposed to the public world, they figure out how it works or where the flaws are, they can use it to reverse engineer us. So there's a lot of security requirements as you start using the LLM. So I don't know in the short term out of there or not, but I know that the current state of AI is the worst it's ever going to be, probably going to get better. And I remember my first brick phone. It used to be in my car. I had to put a charger into it and it cost $2.99 a minute. And you told me that was the state of technology, we weren't there. But today, you can't leave home without those things, right? So if you imagine 5 years out, what is going to be the state of AI, it's going to be a very different world. In that scenario, who is going to be there? So we have to be constantly paranoid in our business that we've got to protect for something you don't understand fully yet.

Could it -- is your Prisma AIRS, is it part of -- is it the feature on your other solutions? Or is it the product stand-alone? Meaning will we be able to measure it in the future? Or is it going to be just integrated into other solutions?

No, it will be measurable by itself. But like you said, look, there's a lot of people evaluating AI. But if you can ask all your companies to come in and say, how much AI are you using internally? The answer is going to be, we're still experimenting or we have 1 project, we have 5 projects. We're using 37 models at Palo Alto, 37, right? Half of them are experimentation. But that's, again, a stark departure from the notion that enterprise is going to use one. We're not just using OpenAI. We're not using Gemini. We're using 37 different models. And use 37 different models to make sure they're all secure. Suddenly, your security attack surface has increased. If I was using one, Google can help me protect Gemini or Sam Altman can help me protect OpenAI. But if I'm using 37, it's my job to make sure I'm protecting these all 37 models being hijacked. So we made this bet 7 years ago that we didn't think most customers are going to be in a single cloud, which has turned out to be true. If you look at any company, they're on 2 or 3 clouds. So I think the same thing. You're going to be using so many different models that you have to find a way to protect against them.

As you speak, I have -- I ask myself, take, for example, the bank I work for. Publicly, they said that they have 300 security solutions. You are talking...

300 are too many.

Right. But you're talking about a platform that has a pricing advantage, has a technology advantage. What needs to happen and maybe it's already happening, what needs to happen for enterprises to abandon this issue of best-of-breed and optimize each domain separately and migrate more to a platform approach because anything we're going to touch on XSIAM, Cortex and AI, there are benefits of having a platform. Forget the price. There are technical benefits of having a single platform. So what needs to happen in the market for the market to migrate from the historical deployment of point solutions to a single vendor that is going to provide you -- I'm taking it to extreme. So going this way.

Look, it's happening. The problem is we're sitting on $1 trillion of security plat, which is unamortized. People bought this over the last 10 years. It's not -- it's $1 trillion of security spend in the last 10 years, which is in production being used right now, which is a large mesh of all these things you talked about. Getting that sorted out, understanding the renewal cycle of those, understanding the replacement cycles of those. Look, the easiest thing to do is to do nothing, right? It's working. why mess with it? I don't understand the tangible benefit of security. Of course, the day you get breach or hack, you understand the tangible benefits of not -- or downside of not having the security. And that kind of becomes the holy grail for all of us. We all rush to that place and say, "Can I help you transform the whole thing?" That works. But normally, if I walk to your CIO and said, "Do you want to replace this entire backbone?" You're going to say, "I will do. I going to go to AI. I'm busy. I do cloud, I'm busy." So the natural inertia of technology organizations is to not fix what's broken until you get to an inflection point, but it's so bad that you look silly by not changing it. Look at -- and you know this, like look at the endpoint industry, right? For a long time, we didn't have to replace Symantec or McAfee. It was great. Eventually, it's a holy s***. You can't be using Symantec and McAfee. You got to get off that, whether it's Broadcom buying Symantec or whether it's McAfee going through 3 private equity cycles. Eventually, everybody moved from there to CrowdStrike or Microsoft or Carbon Black or you have your -- pick your favorite XDR vendor. So the industry inflects and transitions. At that transition point, you have to figure out how to find a platform solution because that provides the incentive to migrate. So take XSIAM for us. SIEM was not inflecting 5 years ago. We started building 5 years ago. Today, SIEM is inflecting. Every company either has a SIEM replacement or migration project started or about to start in the next 12 to 24 months. Now we should use that opportunity to inflect and replace like today, on average, you need to deploy XSIAM, we replace 4 to 7 vendors. It's a good start.

Yes. Last question on AIRS or maybe 1.5 questions. So there is security of AI. There is AI being used for security. What problem are you solving with this solution, both or one of them?

No. AIRS is fundamentally to help organizations deploy AI securely. If you want to deploy your own AI in your own company, you use Prisma AIRS and it puts a wrapper around it, it protects you. So your data doesn't get stolen, your models don't get hijacked, people don't inject prompts. There are all kinds of fun examples. like example like all of you are in San Francisco, some of you might try Waymo outside. That's an AI model, teaching -- telling the car when to turn, when to stop, when to break, where to drop you off. If somebody was able to crack the security of that car, how do you feel about that? Now if you take that analogy and apply that to every agent you want to run on your enterprise and protect every agent from being hijacked or being taken over by bad actors, that's what needs to be protected. That's what Prisma AIRS do. Using AI for security is a different problem. For that, you need a lot of good data. Our biggest challenge in the security industry is we have a lot of data but not a lot of good data. And that is why I said to you the SIEM industry is inflecting because that's where the data was supposed to be. But the old models were charged so much money to ingest data that there was a natural disincentive to collect a lot of data. So if you look at the traditional companies, I won't take names, but who were in the SIEM business, the #1 drive I see was, it costs too much money. And there's a lot of data, and I have to pay a lot to ingest the data. So they created a natural disincentive. So we flipped that in this head. Today, we collect anywhere from 70 to 100 terabytes per customer of data at the right price for the same price they were ingesting half of that earlier that allows us to deliver a better security outcome.

Yes. So you launched XSIAM a few years ago, grew tremendously.

30 months ago.

30 months ago, grew tremendously. What's your differentiation? There are -- when we talk to vendors, everyone tells us they have a SIEM solution, some kind of a SIEM solution. What's your differentiation in the space?

The intelligent people understand, will understand the differentiation. But fundamentally, SIEMs have been large data stores. You collect a lot of enterprise data and then run solutions on top like UEBA and a bunch of workflow management to figure out how to figure out the most important security issues they have today and solve them. The problem is like playing whack-a-mole. Every morning you wake up, there's a new set of issues. You got to play whack-a-mole, it goes away. Our whole principle was ingest all this data, analyze it using machine learning and use humans as a training element. We said that 5 years ago. Now it's becoming more real with AI, the way it's being deployed. Our intent is to eliminate everything in your security infrastructure, any issues using some version of automation or technology or machine learning. I'll stay away from the word AI, but effectively what is traditional AI or precision AI, we like to call it. So that's the point of view we had. That's how we did it. And the proof of the pudding is in its eating. We have taken -- we've deployed at about 150 customers so far out of 200-plus that we've sold. The average or median time to detect and remediate has gone on average from 4 days to under 10 minutes across every one of the deployed customers. And everyone is a case study. Right? So that's what we go with as our credentials say, you take 3 to 4 years, 4 days to find out what has hit your infrastructure. That's too long. Today, the fastest ransomware attack was simulated by us at 27 minutes. So 27 minutes from start to finish, we can get into a company and take their data and leave and hold them to ransom. If your time to detect and fix it is 4 days and the time for me to get in and get it out is 27 minutes, we have a problem. You got to go below 27 minutes. We're at 10. At Palo Alto, we're at 1. We want to take all of our customers to 1. Eventually, it needs to get as close to real time as possible.

And you spoke about the expansion of XSIAM to e-mail. What are the areas XSIAM can grow to? What can you -- just think about kind of your strategy for the next few years, where would it go?

Look, in the long term, as latency continues to improve, as cost of storage continues to decline, as bandwidth continues to become more and more abundant, you have sensors, which are effectively at the edge of your network or at edge of your solutions, those sensors collect data. They have some technology or models at the edge, that would say, "I know this is bad, stop it." Most security attacks happen when something is bad, you don't know it's bad. If it's bad, you knew it, you would stop it. So there's not a prevention at the edge problem. This is a -- I didn't quite understand this was bad, hence it got through. So when it gets through, it becomes a data problem, right? So our view is that eventually, there will be sensors and large streams of data running into a large analytical backbone, where it's going to be analyzed on the fly, and the announcement of fly will tell you, "Oh, it looked good. It was bad. I figure that out in a second. Let me go back and block any activity post that." That's the holy grail, right? That's where you want to get to. You want to be able to real-time stop bad things from happening and analyze them. Otherwise, you keep fortifying the doors with more and more things to do. You like we -- like we take off our shoes at the airports, right? Every time something happens in the security industry, we have 1 more thing to do, no liquids, takeoff your shoes. All these new rules are written because that was bad. We didn't know it was bad then. Now we know it's bad, so we protect for it. But somebody comes as a new attack vector, which is no longer known. You got to take off your socks as well now, I don't know. Had to use nonsecurity analogies for learning security.

We have 2 minutes left. I want to ask you one more question before I'll give the -- I pass on the mic. QRadar. You acquired the IBM Security QRadar, Software as a Service assets. Can you give us just an update on the experience after the acquisition.

It was great because it allowed us to partner with Microsoft and go to many of their customers and take them on this journey of migrating their SIEM and SOC. It added a tremendous sort of boost to our ability in the SIEM space because if IBM decides that they should partner Palo Alto and effectively allow us to manage their product for them. We've seen a lot of -- many of their large customers convert to Palo Alto. We're still in midst of converting a lot of them as well. You could see that the migration has already begun on the customer end. They were already thinking or planning to move Alto platform. So we showed up at the right time. I think it's still going to go down as one of our best deals that we've done from a -- which is nontrad. We didn't buy product. We actually bought, I'd say, go-to-market in a way. And so far, so good.

Is there any question from the audience? We have yes, 1 here and 1 here. Here we go. We're going to go like 1 or 2 minutes into the...

Okay. I'll be quick. So with all of the capabilities the platform is now powering right, can start-up AI security companies really gain much traction when you've penetrated the -- increasingly, the data stack as well as the network side. Is there room for those players in the existing market?

I'm -- look, that's a good question. Like I can't say no, because if you look at the history of cybersecurity, there's always been 2 or 3 companies come from the side and become big, whether it's a CrowdStrike or Zscaler, now Wiz. You'll see always somebody chasing that. But I will say the environment today is very different from 7 years ago, 7 years ago when I started this job, we all used to soft of stay in our swim lane and live and let live was the philosophy. Today, you noticed most, I'd say, somebody said the top 5 cybersecurity players in the market, they're all paranoid. We're all looking to see what else can I acquire, what else can I expand? I think somebody just bought a SIEM solution to enhance their capabilities. Somebody bought CNAPP capabilities to enhance that firewall business. So what's happening is everybody is starting to cross over into territories, which traditionally they left alone for somebody else. And you're seeing they're getting some degree of penetration. So you're seeing 2%, 4%, 5%, 10% market share in many of these categories by incumbents, which was never seen in the past. In the past, a whole new category. If you think of the XDR category, like McAfee and Symantec were kind of doing their thing. You saw Cylance, Carbon Black, cyberreason, CrowdStrike. There's a whole bunch of them that showed up and all of us just watching. Nobody did anything. right? Suddenly now, that's not possible anymore. You say CNAPP, every security will say, I got it. What it does is it creates commoditization. What it does is say everybody says I have the capability, then you have to go discern which one has it. So I think it's more competitive from that perspective. Having said that...

Scale is going to matter?

Yes, scale is going to matter, but incumbent have to be careful. There's always somebody smarter faster than you. So our job is to identify them and hopefully, they partner or sell to us, like if Protect.ai hasn't sold to me, I have to build that capability myself. I'd be 6 months behind them. So the key is, can you get momentum? Can you move fast enough? And can they see the value of being out of your platform as opposed to going to something else. And platforms take time to consolidate. If you look at cloud security, it's Wild West about 5 years ago. There was DivvyCloud, LaceWork, Dome9, all these things are gone. Now you've got Wiz, you've got Palo Alto, you got Liberam, CrowdStrike on the cloud side. So suddenly, it's beginning to consolidate into existing platforms as the industry matures from the solution set. Today, in AI, we don't know what the solution set is going to look like. We're still evolving it. So we didn't think teaming was important. Now we think it is important. We did not think model scanning was important. Now we think it's important. Now there's a whole bunch of Agentic AI security that needs be built, which just doesn't exist. So can a start-up show up tomorrow inside Agentic AI security? Sure.

The question I had was.

Yes. Let me -- since everybody may not have heard the question, he's asking the question which is, just tell me what's growing really fast and what's not growing so fast, right? So this is what he said. right? Okay, got it.

It's easier to put that in your model. I'll give you the financial answer. The financial answer is we have a portfolio. And in some quarters, some things do a lot better than the others, but they sort of normalize over time. The way I see it is if I do this for the next 5 years. So we can keep compounding our business in the mid- to low double-digit percentage points and this business doubles in 5 years, right? If this business doubles in 5 years, you decide what multiples you want to put on it. But with the combination of the portfolio we have, we're doing -- we are seeing all the right trends. We're in the hardware-to-software migration of network security better than anybody else in the industry. 7 years ago, we didn't have a SASE business. Now our SASE market share, we're probably #2 in SASE after Zscaler in the world. Nobody expected us to be in that space. We didn't have an XDR business. And we're 1 of top 4 in XDR, and we'll see what happens to one of the 3 players, it's Microsoft, CrowdStrike, us and SentinelOne. So I think we're slowly making inroads into categories which are important for the future. And SIEM, we didn't play until 30 months ago. Today, we're 1 of the 3 people that people consider moving into from their existing SIEM. So we are catching the right trends. We'll see how the AI trend shows up. The whole idea of catching the right trends to make sure that the portfolio grows at a certain percentage and over time allows us to platformize. And the more we show up as a solution, the customer says, "I already have Palo Alto, why do we need to go buy something else to attach to this?" So our view is, land the platforms, grow the platforms. I'll give you an example. There was a customer when I joined Palo Alto, whose total spend was $3 million a year on a product that was I'm pretty sure they ripped out had we not refactored the product 2 years in. That customer went from $15 million over 5 years to $23 million over 3 years to $70 million over 3 years, that was an upgrade mid-cycle. They're probably going to be $140, right? That's platformization. You go from $3 million a year to going to $30 million a year. That's the 10x over time when you keep working with the same customer and keep helping consolidate. If I can do that for a few thousand companies in the world, we can get a real business going.

Thank you.

Thank you so much.