Google Cloud Next 2025 | Munish Khetrapal, Palo Alto Networks

Clips
More from Google Cloud Next 2025

Munish Khetrapal

VP of Cloud Engagement

Palo Alto Networks

play_circle_outline Palo Alto Networks' partnership with Google Cloud and platformization initiative

play_circle_outline Role of agents in future security operations

play_circle_outline Transforming Security Operators in Distributed Systems: Integrating Security with Operating Systems

play_circle_outline Use of supercomputing for precision security and threat detection

play_circle_outline Strategic partnership with Google and leveraging their advanced capabilities

Info
Transcript

Munish Khetrapal, Palo Alto Networks

Munish Khetrapal

VP of Cloud Engagement Palo Alto Networks

In this insightful interview from Google Cloud Next 2025, Munish Khetrapal, vice president of cloud engagement at Palo Alto Networks, shares valuable perspectives on the future of cloud security. The session, part of the ongoing theCUBE series, features a vibrant discussion on how generative artificial intelligence (AI) transforms software development and cybersecurity practices in a rapidly changing technological landscape.

With hosts John Furrier, Dave Vellante, Savannah Peterson, and Paul Nashawaty, Khetrapal sheds light on Palo Alto Networks' rec... Read more

explore Keep Exploring

What was the announcement made by Nikesh, the chairman and CEO of the cybersecurity provider, last August regarding a platformization initiative and partnership with Google Cloud? add

What is happening in the enterprise IT world over the last 50 years? add

What is the impact of AI and distributed architectures on enterprise IT security and cybersecurity? add

What company was announced as the AI technology partner of the year for Google and the AI technology security partner of the year? add

What is the current relationship between the speaker's company and Google, and how is it working for them? add

bolt Powered by CUBE AI

Munish Khetrapal, Palo Alto Networks

search

>> Welcome back everyone to theCube's live coverage here at Google Next '25. I'm John Furrier, host of theCube with Dave Vellante, Savannah Peterson, and Paul Nashawaty from theCube Research team, breaking down all the analysis here at Google and talking about everything from AI infrastructure changing, evolving, security paradigm changing. Agents are coming, the role of data and software, super critical, and of course the hardware advances. Palo Alto Networks is here on theCube. Manish Khetrapal is VP of cloud engagement at Palo Alto Networks. Manish, great to see you again.

>> Thank you.

>> Welcome back to Google Next, but also we talked with Stephen Orban recently in the Marvel series-

>> Yes, you did....

>> about the success you're having at Google Cloud with Palo Alto Networks. Obviously everyone knows who you guys are, but you guys have some news you shared here at the show. I want to get to that right away. And of course, we're going to talk about the role of software, how that's changing and how software is built with this new generative AI era. But you got some news around how much sales you're doing through Google Cloud.

>> Through the marketplace. So just taking a step back, we are the largest pure play cybersecurity provider and Nikesh, our chairman and CEO, about last August, '23 August, announced the platformization initiative, which basically said that if customers are looking at a lot of cybersecurity problems, where average we are looking at 25% of the customers today, the data is accelerating in five hours. In 2020, it was nine weeks. They have 84 different security products today and these 84 different products need to be platformized. So we went into this platformization initiative and Google is a strategic partner of ours, and we have driven over a billion and a half dollars of sale through Google Marketplace. We enjoy the partnership.

>> 1 billion, 1.5 billion-

>> 1.5 billion....

>> through Google Marketplace, which means that's a go-to market for you guys.

>> That's right.

>> That's sales.

>> That's sales. That's right.

>> Of Palo Alto.

>> That's right.

>> Okay. So what year did you guys do the platforming? I forgot what year it was. When did you guys really take up the platform view? What year did you guys start really nailing that down?

>> Last summer, right?

>> It was last summer. That's really when we went in and announced the platform view. We have been building best of breed technologies for 14 or 15 years. We have 24 leading technologies. Each of them great in their own area.

>> But you just declared the death of best of breed in our little analyst roundtable, meaning platforms are going to replace best of breed. It was very controversial... So I put out a post and then our John Olsuk, who's with us now, it created a nice debate about best of breed.

>> Sure.

>> Of course, it's an ongoing debate within the customer base, but look, it's clearly working. You guys have momentum with platformization. We're going to obviously measure that. It is hard for practitioners to reduce from, what'd you say 82? Because it's just some new toy that comes out all the time that plugs a hole.

>> Absolutely.

>> And then people gravitate to it.

>> Absolutely. And in the world of AI, I think there's no debate anymore. If you take the Forbes, 150 of them in the last 18 months already gone on Palo Alto Network's platforms.

>> Yep.

>> Pretty much every top 2000 customers has at least used one or two of our platforms. So I think the debate of platforms is over. It's a good question of adoption cycle.

>> BJ gave several examples, I want to throw out Nir, we were talking earlier, Nir, when I met with him in one-on-one, he said, "Forget about today's security stack. AI renders and automation renders all of that essentially ineffective." I'm like, "Explain that." He said, "Look it, we could stop 99% of the attacks previously and then 1%, humans could deal with. But now with generative AI, the attackers can scale phishing attacks, for example, at unprecedented at the levels. Humans can't handle it anymore."

>> Absolutely.

>> So you can only fight automation with automation, period, the end. So AI has to be infused in the entire security stack.

>> Absolutely.

>> Okay. So the question then is, why Palo Alto? Are you an AI company?

>> We are a cyber security company where AI is the most important asset for us. So we have been collecting data from 2007, 2008. We have the largest real-time database of threat intelligence that exists in the world, nine petabytes of data a day. Just to give you a scale of what nine-

>> Unit 42, right? Those guys.

>> Unit 42, nine petabytes is like 15 New York Public libraries, which is the largest public library, on a daily basis, just on threat intelligence.

>> In a beautiful building.

>> Just to give you an analysis of how much data-

>> Massive data. Tsunami.

>> Massive data. And then you need AI to process it because now you have to look at zero-day threats. You have to look at all types of different attacks that are coming through different vectors with AI adoption across enterprise.

>> And the other thing Nir said to me, I learned so much in 20 minutes just talking to the guy. Because I was pushing him on the data consolidation and the single data platform, and he says, "It's not so much... It is the consolidation, but it's more important to have the access to the right data at the right time." Kind of what you were talking about before, that the example you gave off camera, you got to get to four different data sources. Having the right data allows you to respond before the exfiltration occurs.

>> Absolutely. So if you look at the point products and where they really fail is you've got a great product that does only VPN and then you have got somebody else who does the firewall. Then you have somebody else who takes care of cloud security, somebody else who takes care of the data security. What happens is if there's an attacker, the attacker's coming in as a phishing email or has tried to try the password a couple of times, but the agent that provides that information, you lose sight of that. That attacker went in, then hacked into the system, escalated its privileges, and then move those privileges to access data and exfiltrate data. If you have a platform, you can connect the dots. Oh, somebody entered that email twice and then immediately after that he went into a system he wasn't supposed to go, and then after that he looked at the data he wasn't supposed to get to. And that interconnection can happen only with data and platforms, and that's why I think we are unique.

>> Manish, talk about the success with Google Marketplace and how that relates to the AI revolution. We heard Google on stage on the keynote say they're the only cloud that has gen AI in all modalities, so that's good for them. Good job, Google. But what does it mean for security? What are the use cases that you see evolving that the platform's enabling? Where are you guys seeing the gen AI for security and where's the security for gen AI?

>> Yeah, so we have taken an approach where we have been developing over 5,000 models that we have developed over the last 15, 16 years. And in security, we have to be right every time, because if you're wrong once, we fail. The attacker has to be right only once to enter a system. So what we looked at is we looked at machine learning, we looked at neural networks, and we used the combination of generative AI to now start building what we call precision AI, that ties together the best of the AI world to enable us to automate our platforms and technologies to drive further.

>> I've been saying on theCUBE, I don't know how many years we've been saying this, but recently a lot, past two years.

>> Yes.

>> And the past year specifically, cybersecurity vertical or in section or what do you want to call it? Sector, is probably the best use case of where the talent that people actually writing code, building platforms, like Palo Alto represent the broader gen AI space.

>> That's right.

>> Because it's real time, multiple databases, you said platforms to connect the dots are needed and just the inbound data tsunami of red teaming, blue teaming, just all the work that gets to get done to defend, protect, remediate, all those things generations of talent have been working on is the same kind of, it rhymes with all the things we're hearing in other, I would say slow moving verticals that now have to deal with a resilience bar that's super high on production. With gen AI, you're starting to see data threats. So it's almost as if it really is a cybersecurity business now, tech. And just from a technical perspective, even Google on stage today, full stack approach. Again, another theme we're hearing. What does this tell us about our future? Because if software does need to be rewritten, how do you frame that? Scope for us what all this means, because you guys have been dealing with all this, which is why you're first mover on some of these key advantages and these breakthroughs. I think you're a tell sign for the rest of the industry at large. How do you frame that? How do you scope it? What is happening?

>> If you take a little bit of a step back before I answer the cybersecurity question, what's happening in the enterprise IT world? Enterprise IT over the last 50 years has only operated in one model at a fundamental level. There's a system, it's your mobile phone, your desktop or your laptop all accessing something in some central place, whether it's client server, whether it's mainframe access, virtualization, cloud computing, it's always about accessing a central database. But with AI, that's changing. With agents, you're talking about what? 15 billion agents that are going to be in the workforce by 2030. That's the amount of data you're going to generate, and that's the amount of thinking you're going to go. So it's going to be a completely distributed architecture, which means every enterprise will need to spend money to evolve their current enterprise architecture. And even a simple prompt and an example of how a prompt connects to four or five subsystems to collect that data, it's going to dismantle the concept of security in the software. And I think security will become part of the operating system. It'll become part of the middleware. It'll interconnect the different applications. Because one of the value that security can bring, and this is the first time I can say this, cybersecurity is always about insurance. It's always about protecting yourself. But this is the first time it's like, why do you have brakes in a race car? To drive faster, not to brake. Why do you need transformation in your workforce and in your AI enterprise that you're deploying? You need security that allows you to take risks at a higher level without exposing the brand and the company and the IP.

>> So you're thinking about agents right now, I'm sure, because age agents will come into the fold because today's point about who's going to solve all these inbound machine attacks, or things, events. Call them events because you generically say an event's an event. How do you look at the agent wave coming? Obviously security built in from day one, I get that, but what technically has to happen? What's the sequence of events?

>> So agents, what agents became a fancy word six months back? We have seven and a half million agents forecasted to be enabled in an enterprise by the end of this calendar year. 15 billion by end of 2030. So that means your workforce today, give and take is about one and a half billion people in the workforce. You're going to have 10 times more of the workforce. That means you're going to have 10 times more of complexity, 10 times more of intelligence.

>> What is the role of a security operator? I just made the term up, popped in my head, but network operator, we've heard that term. Database administrator, these are concepts that have been around for generations. Now the role of NOCs and SOCs are changing because how do you operate a fleet of agents when they become more than just a technical feature for evaluating packet movement, software patterns of users. Now you got holistically think almost like, what did Jensen call it at CES? IT is going to turn into the HR department for agents. That was the haymaker he threw.

>> It absolutely is.

>> And he's not wrong if you take it... But there will be a role for the human.

>> Yes.

>> How do you view that?

>> Same example of what do you do with taxi and Uber drivers when you have Teslas, an autonomous car that's going to be there? So security is going to be autonomous. It's going to be collecting data, it's going to be managing data, it's going to be managing flows. The role of the security operator will be less about how do you run security and threats and attacks. It's how do you orchestrate and govern it, what policies you want to implement. It'll become more of a policy maker. I use this airport example quite frequently. You have Google, you have NVIDIA who build the best planes, who build the best infrastructure, and then they build the best security as well. But as a passenger, a wrong passenger can enter the wrong plane and it can be a disaster in several different ways. That's where Palo Alto comes in. You become TSA security and you become baggage routing facility and boarding facility. But then you have a role for an enterprise controlled car, an airport control center. Which plane takes off when? When do you land? Which passengers schedule? And that's the role of a new security operator.

>> So it's a shared responsibility model.

>> Exactly. Yeah.

>> Absolutely.

>> So that's shared.

>> Absolutely.

>> Airport's a shared responsibility model.

>> Yeah, it's a joint responsibility model more than shared.

>> One of the things we heard at Ignite in New York City, which was an interesting concept and talked to a lot of people about it, is wartime versus peacetime security. Explain what you guys mean by that and how you see that evolving.

>> So the role of a security operator, if you think about if you're in a SOC operating environment, till a couple of years back, and of course the people who have advanced their SOC operations have seen their lives change. You had 15,000 alarms and events happening on a daily basis. You have a team of 10 people, how are you going to handle 15,000 alarms? So it was always tactical, it was always reactive. You're looking at what's the next best... Whoever cries the most. But if you can automate that and bring them down to 7 or 10 attacks on a daily basis, now that role of that operator spends one hour on the tactical stuff and 7 hours thinking about how to automate playbooks, how to improve security posture, how to increase the ability for security to add value to the business, and I think that's becomes a more strategic role.

>> And then of course, posture management is this big thing, right? Everything's become posture management. What's your point of view on posture management? It's not unimportant, but it's not everything.

>> Posture management is one pillar and one element of our thinking. If you look at the way we have looked at designing our platforms, we have posture management, provides you visibility. Tells you who has the compliance problem, who has the data problem, what's the vulnerability. It tells you what, but it does not tell you how to detect it, how to control it, how to respond it. So you need the combination of posture management with detection, in-line detection. With AI,. You cannot afford anything but in-line detection

>> And that's where agency comes in, right?

>> Yeah. Absolutely.

>> I mean, you mentioned precision, mentioned the platform. You got Copilot is another big thing. Talk about those three things, precision, platformization of security and also the role of Copilots.

>> Absolutely. Precision AI is about technology and implementing AI and getting it right every time because that's the only way we can do it. That's about an AI implementation in Palo Alto networks. Platforms are about bringing all the data together and using precision AI to fight against detection, controls, response, all of that stuff and automate it. Copilots are about making it easy to use the platforms. So three distinct capabilities.

>> In precision medicine, for instance, supercomputing has been a tailwind. How has the advancements in supercomputing. What was it? 45 exaflops I heard on stage today with the new TPUs, Ironwood. You're seeing massive surge on all clouds and on-premise.

>> Absolutely.

>> Nvidia, we've all been covering all that. You got the horsepower now to run stuff. Does that make precision security better and how?

>> So just as one of the announcements we had yesterday, we got announced as the AI technology partner of the year for Google and the AI technology security partner of the year.

>> You are an AI company?

>> We are an AI company and we are the partner of the year. So the reason is we've been able to use technology that Google Cloud has enabled and being able to improve our models, improve the precision and attack capabilities that we can use. Zero trust, zero-day attacks, looking at patterns and recognizing those patterns and forecasting those patterns. Autonomously creating playbooks to prevent any attacks to happen. So we are using this technology, the GPU, the infrastructure, the models, Vertex AI models to improve that.

>> That's a boost for you guys. Big time.

>> Yeah, it is.

>> And the data platform, you guys are big query users as well. I mean, it's-

>> 15 New York public libraries of threat data on a daily basis, which is the largest public library in the world.

>> On the threat data, talk about the landscape of threats. Are there in models emerging? Are there ones that you're seeing clearly? Are there lines forming on the threat landscape, you say these are the five models, or these are the three we see? What is the threat landscape look like to you guys?

>> Excellent question. I think I wish I had an LLM just to tell me what the new threat is going to be because there's so many new threats happening on a daily basis. We have organized it in a very simple way to think about it. It's a framework. Framework is as accurate as a framework can be. This question of what are the threat data-related issues that can happen? Data poisoning as an example. And I like to use the Tesla example, bring it back. Suppose in a stop sign you have 25 miles an hour. You're poisoning Tesla, the thing that's 25 miles an hour zone and guess what's going to happen to the stop sign? So that's data poisoning for you. Then the second thing is model and how you can exfiltrate model, how you can reverse engineer a model based on the outputs and actually find out the models that you've trained with your IP and build your brand, how you can actually reverse engineer it and understand your IP. So that's the model-related issue. Third is inferencing. Prompt injection is a very typical attack of inferencing. You can fool the inferencing engine to tell you something it doesn't want to tell you. Fourth is infrastructure attacks, denial of service, denial of agent usage. Because now if you're going to depend your life on running your business, you got to make sure your infrastructure is secure, and anything entering that infrastructure is secure. And fifth is governance, which is equally and more important. Visibility, posture management. What are the actions? How do you train the models? What's your ethical behavior? What's the social behavior? So those are the five areas of risk that we see and we are building capabilities to address those five risks.

>> I'm sure there'll be more. I know we're tight on time, but I want to make sure I get in the mention and also discussion around Google. Obviously partnering with Google, you're the VP of cloud engagement. That's the account you manage and work with. Again, we've had you on the Marvel series with Stephen Orban. Again, you marketplace numbers are phenomenal. What is the relationship with Google and how are you guys advancing on behalf of the enterprises that are platformizing their environments? They want to have a bridge to the future that's solid and reliable. Talk about the Google relationship and how that's working.

>> Google is a strategic partner of ours. We have stated that several ways in the last few years and we talked about the big announcement from a go-to-market point of view. But the first thing is we build many of our software platforms on Google infrastructure. We also work on other cloud infrastructure providers. Doesn't mean that we don't, but we use a lot of the advanced capability Google has, BigQuery, Gemini to build best-of-breed platforms. If you can call that best-of-breed. I'm going to bring both those words together now. The next places we are using Gemini 2.5 to improve our co-pilot. So how people consume our platform is equally important. You can build the best technology. If you don't know how people can consume it, it's not going to be valuable. So we are leveraging Google technology. But the biggest area of investments is helping enterprise secure their AI journey and accelerate their AI journey. We have over 76 integrations into several Google products that allow customers to easily consume our products and capabilities.

>> What's on the roadmap for this year? What's on your to-do list? Obviously you got the numbers up and to the right. 1.5 billion, not too shabby. It's some real money there. Inference is happening in the cloud. The TPUs are going to give you some good boosts there and more capability, so congratulations on that. What's on the agenda with the relationship? What could people expect?

>> So I was joking with Stephen yesterday. We went on a common panel and he talked about his backlog being $96 billion, Google backlog. So I said, "We should get about 5 or 10% of that share. That's our fair share of market." So that's definitely something that's top of our mind. I mean, that's definitely an ambition. The second area is agents. We are going to be looking at how do we continue to help customers adapt more agents in a secure manner. And third area is we continue to enhance our platform, it's a breakneck speed. Every three months you're going to evolve your thinking and there's so many new things happening on a daily basis.

>> What's your market share today? It's less than 10%, right?

>> It is less than 10%.

>> And so Nikesh talks about 40% off the charts. My feeling is if you can get to 15% of the market, that says to me that platformization is actually winning. So that's kind of my first milestone.

>> Yeah, I would say there are two dimensions to it. Because if you think about market share, that depends on the market size. But if I change the market size, I can -

>> Yeah, if the market triples-...

>> I can still double my revenue, right?

>> Your point, if the market triples you can, yeah. Which is great.

>> With AI, the market itself is going to go to levels that we haven't thought about because AI cannot be adopted without security and we don't know what that... We have estimates of what it's going to be, but I don't think we know what we don't.

>> And we talk to a lot of customers of yours and other vendors, security resilience bar is so high on gen AI right now.

>> Absolutely.

>> You can't run AI without security.

>> Yeah, absolutely.

>> Manish, you got a great spring in your step. Congratulations on the performance.

>> Can't run security without AI.

>> security for AI.

>> It's a data problem we said many times on theCUBE. It's going to be right, Dave, looking back at the videotape as we say, looking at our CUBE commentary. Manish, thanks for coming on again.

>> Thank you.

>> All smiles here on theCUBE. Business is good in security. I'm John Furrier with Dave Vellante, your host of theCUBE with Savannah Peterson and Paul Nashawaty. We'll be right back after this short break.

>> Thank you.