Welcome back, everyone, to theCUBE here in our New York Stock Exchange studio. I'm John Furrier, host of theCUBE. This is our East Coast studio, overlooking the show floor. Part of our new brand, the NYSE Wired program, it's a CUBE Original, plus the open source community of leaders are part of that, curated by Brian Baumann at the NYSE. We got a CUBE alumni going back to the origination of theCUBE, 17 years. Sanjay Poonen, CEO of Cohesity. Sanjay, great to see you in our new set. What do you think? Not too shabby?
It's amazing. I got to tell you, I knew you when you were small-time in this small little place in SAP, and I may have been your first or second, or third interview. I used to come here to NYC to talk to CNBC. Really, I love this place.
Great. Congratulations on all your success.
And how did you describe me to Lynn Martin, the CEO? What did you call me?
I said, "Vellante and John, I was told they are two very interesting characters." I forget exactly what it is.
Jim Cramer. You mentioned a Jim Cramer reference.
Yeah. You are the Cramer of tech, okay? Cramer has a more business perspective on everything.
His stock picks are great, legendary. Got a great community. You've done it well. And you've covered SAP, VMware, AWS, Dell. I've been to many of the events. You and Dave Vellante do a great job.
Well, appreciate it. You're a special friend of theCUBE. More importantly, we've been following your career from SAP, of which Bill McDermott's now the CEO, was a co-CEO of SAP, and now he's the CEO of ServiceNow.
Very timely to get into that discussion. Progression through VMware where you transformed VMware, set the agenda on their growth strategy.
And now at Cohesity as the chief executive officer, and now with the combination with Veritas, you told me in theCUBE, better together, one plus one is five. I think it might have been five or three or four...
11? Okay. One plus one equals one, one, one, one. There you go. So, a lot's going on. The performance is great. Congratulations. You're in New York for an event. You have news. I want to get into the news real quick. Then, I want to get into what you see your strategy is. This is our Cybersecurity Leaders. You had some news with ServiceNow and a variety of other select targeted ecosystem winners. Take us through the news you had today with ServiceNow.
Yeah, John, it's been a great year. It's been about a year since we finished the acquisition of Veritas. It's gone really well. We're making our numbers. We're the number one market share. So, Cohesity's gone from number seven in this space to number one. And really, we're building a rule of 40 type of company. Very strong success in the Fortune 500. 70% of the Fortune 500 are our customers. Extremely strong in banking, public sector, healthcare, telco and technology firms. So, we had a big event here in New York and a little bit like the Rolling Stones tour, we're going from New York, Paris, Singapore, three events, bringing our customers here with hundreds in the room and then thousands and tens of thousands watching on livestream. We think that's the new model of how some of these things are going to play out. And we announced some strategic news around the notion of how AI and agents could be used for good and how do we play defense when they're used for bad. The agent-resilience story was a key part of what resonated. And we made a strategic announcement at ServiceNow, where what they are building is an Agent Control Tower for a place where you can build, register, manage, secure and observe agents. And that Agent Control Tower, if one of those agents go awry, remember that fat finger delete of a file you had in the past? But this time it's not a human, it's a non-human doing that. They would detect that and they will send us a signal. And we're just basically a platform that's built for zero-trust, immutable. We can then restore the data to the previous point in time because we have all of that data available to you. We announced that with ServiceNow. Bill McDermott and I were quoted. I put out a blog post on LinkedIn and he echoed that to his community. We also had an announcement with Datadog. Those were the first major pieces.
I want to get into this ServiceNow because they've been on a tear lately with agents. Obviously, they have their business platform, been a great business model. Agents are particular. Everyone loves them. They're going wild. They're running wild. The number one conversation is tracing, observability, tracking, and then trust. Is it explainable? So, in the enterprise, these are the two hot buttons that people are talking about. Some say guardrails, but really, it comes down to, can I observe them and manage them, metrics to that? And then, two, are they trusted? Can I explain where it came from to manage around hallucinations? This is a huge factor. How does that relate to the Cohesity-ServiceNow story? And what's your role in that agent, protect, secure, scale?
Yeah. So, we are a data-resilience platform. So, we have 200 exabytes of the world's data. And by the way, that number is significantly bigger by a factor of 5 or 10 than all our competitors put together. So, all of the largest companies store their historic data. Secondary data, backup data is just a time series index in an immutable, secure fashion of everything you have, especially if your retention requirements are forever. So, we're not the platform to build, manage, secure, register, and observe those agents. In fact, some of our competitors built an agent cloud or acquired an agent cloud, we think that's an utter waste of time and money. The companies that are going to do that are going to be companies like ServiceNow, like Datadog, like AWS that has AgentCore, like Microsoft that has Agent 365, like Google who have Agent Foundry, CrowdStrike bought a company called Pangea, Palo Alto, they're all focused on some aspect of that agent platform. So, we went to each of them, starting with ServiceNow, because if you think about agents and managing and securing and observing those, it's like a CMDB problem. If you are a CMDB, which is what ServiceNow does, think of agents as just being one more asset that they're tracking. And Bill made it very clear to me, "We intend to be the Agent Control Tower of the enterprise." So, we said, "Listen, that's great. We're going to integrate with you. So, where you see an anomaly because of an agent doing something..." And the beautiful thing about Cohesity, because we were a platform, we could integrate into them with our APIs in a matter of days.
I want to just interrupt you for a second because you mentioned CMDB. Explain what that is, some people might not know what the acronym is.
It is really the database for the configurations.
There you go. Configuration management database?
Something like that. But basically, it's for all the files that configure the past, the static rules, the rules. You could think of it like the SAP for IT. It's an ERP system for IT. You store all your configs in your CMDB. This is what Remedy was in the past, remember those old days? ServiceNow has modernized us for the cloud. But when you track your assets and you track your IT workflows, you do that with ServiceNow. ServiceNow's become the SAP of IT. So, if you do that same process with financials with SAP, you're doing that with IT with ServiceNow. So, agents are just another IT workload that you're going to track within their platform. The question is, will they build a rich enough platform to register, manage, secure, and observe agents? And they're well on that way with the Agent Control Tower. Now, they won't be the only company, but I think they are a leading company to be that, and that's why we picked them first.
Yeah. So, it's kind of like an agentic infrastructure. I was talking about this with another guest earlier today around how you got agents and AI-native software layers and people using all kinds of code assistants and copilots and all that kind of stuff. But the infrastructure piece feels like cloud-native, but it's mostly AI-like infrastructure where, when you inject intelligence, then you have a generative, as Jensen always used to say, generative, not static rules. So, this notion of a config file, which we've seen security breach, "Oh, I didn't update the patch," or, "The rules were broken." The agents have to respond to the intelligence. So, how does that translate into the cyber resilience? Because you have to, one, protect and restore. And data security is the hottest area and that's what you do. So, as you have more intelligence, what does that do for your customers at Cohesity?
Yeah. So, there's two things. One is when that agent goes awry and it detects data that's been deleted, we can restore and recover any point-in-time data. But there's another aspect of it. That agent infrastructure is kind of like a virtual machine or a container. It has state to it. It has a vector database or other things. We announced, for example, the ability to back up and restore Pinecone, a vector database. We can now restore that entire agent infrastructure, all of that, because it's just another workload that we need to be able to restore fairly well and fast and easily. So, as we think about this, we announced both the ability for data resilience, but also the way in which we can do agent resilience for that agent itself.
So, agent resilience is basically cyber resilience, it's data, but they're now workers?
Yes. Non-human identities, that's the way to think about them.
So, when you talk to customers, what are they saying in terms of the work that they got to do to prepare? We've seen examples. I was just interviewing at Barcelona at MWC, AT&T senior engineer, and he said, "We had 19 miles of open source because the frontier miles couldn't speak telco. They didn't have all the data." And he said when he did the analysis, when they have their agents successful, the grind out of the compliance, the hard work that they do allowed the agents to scale. If they didn't do the work on identity or compliance and governance, they went off the rails.
What's your view on that and how are you thinking about that with your world because you're protecting all the data?
Yeah. So, there's two things that I talked about here. One is AI and agents done for good, I want to get to that in a second. And how do you protect from the bad? We've discussed all of what you protect in the bad and partnerships like what we're doing with ServiceNow, Datadog and others to come. But on the aspect of how you use this for good, we have, for example, transformed our entire engineering effort inside Cohesity with agents to write code faster. One of the things that in our space is very common is you build connectors to many of these source systems. We have thousands of connectors to virtual machines, databases. We're now using vibe coding techniques to build connectors, and it's being done six times faster. So, that's a productivity improvement. We have filed for a patent for how you can use this technology to build connectors. Another thing that we did, and this is where our friend Jensen and NVIDIA comes in, we've been able to say, "Once that data that we are backing and protecting lands on our platform," remember it's a software-defined secure file system, "rather than just thinking of it as a storage problem, let's now transform that data to a lake." So, we built the world's first vector database, a semantic layer to interrogate the data that we back up. So, for example, imagine our customer has hundreds of millions of PDFs and you want to write a query, "Show me my summary contract terms I've done with them," or maybe it's healthcare records. Any unstructured data, we back up. All the historical data is there. Search it just like you search with ChatGPT. So one of the other things we announced was that product that we built with NVIDIA. Remember NVIDIA invested in that company, Jensen talked about us, the only company he invested in two years ago at GTC. GTC is coming up in a week. We'll have some more announcements with them. What we did was now we said, "We're going to open up that layer of Gaia to not just be our own app, we will allow search tools like Glean and Copilot and Amazon Q, or ChatGPT, to talk to our semantic layer. But more importantly, we'll allow our catalog, we announced also a Gaia catalog, to talk to Databricks or to Palantir." So, it's opening up that platform now of this data that we have to an ecosystem of data consumers, and that's going to open up a whole new opportunity for this secondary backup data to participate in gen AI apps.
I think the ServiceNow announcement really talks about how their AI platform with your product works better together. It's a good announcement. But you mentioned something about the relationship with Palantir and all these different environments. The AI world, it's not your yesterday's data backup and recovery because you have different modes. You could have this platform... Because you're essentially data security.
Yes. You're all clean data, right?
It's data security, that's the way I look at it. Maybe it's the wrong category or...
It's not a Magic Quadrant, whatever. But if I'm a customer, I might have a little bit of Palantir. I'm going to have Glean. I'm going to have all these systems and the user journey is multimodal, multi-platform...
We are a source of data to those tools, right? Yes, and you're right. I mean, 20 years ago when we first met, we were talking about BusinessObjects, and that was the BI system of SAP. Now, it's Palantir, it's Databricks. We were talking about BW and HANA. Now, it's Databricks and Snowflake. So, yes, the technologies have evolved, but our job is to ensure that if we're protecting the world's data, hundreds of exabytes of it, we feed that data into a Databricks and that's a novel concept. Nobody in our space has built that catalog to be able to feed into these data lakes. And that's why Databricks was actually one of the other announcements we made yesterday. So, I think if you think about the ServiceNow announcement for that agent resilience story and the Databricks announcement for this catalog story, those were the two highlights yesterday.
Yeah. And I think that also the need that we see for horizontal scalability across different use cases and touches. Talk about how that translates to NVIDIA? I know you mentioned NVIDIA, that you're going to have some activity there. We both know Jensen in his speech, he says computer science on stage more than any CEO I know. I mean, you have to work on that, Sanjay. He says it like six times in a keynote. He says computer science, distributed computing. So, clearly, AI factories are pumping out tokens and driving intelligence. Physical AI is a big theme and robotics. That's the two worlds coming together. So, as you look at your enterprise, what is NVIDIA enabling you? More horsepower, supercomputing capabilities? Are you getting leverage out of the performance of the GPUs and the systems? And is there an ecosystem play with things that they've been expanding because they've been really rolling out their ecosystem?
Yeah. Let's back up for just a second and just talk about what an incredible moment we are in. I think Jensen is the Steve Jobs of the AI era, the godfather of AI. People went to WWDC and the Apple launch events to watch Steve Jobs. And I think in the cloud world, they went initially in the SaaS world to Benioff and then Andy Jassy, his three-hour lectures were amazing. If you're in the AI world, you must be at GTC because this person gives you a complete, from the hardware all the way to software, a three-hour lecture. I mean, he's a professor at heart, that Jensen. So, it's amazing.
And he's completely transparent and authentic.
And authentic. He doesn't have the canned answers.
There are lines of people at the San Jose SAP Center lining up to get to the show. I mean, I'm fortunate to have a VIP seat next week. You'll be there, I'm sure.
Yeah. So, the way we think about it is when we built this solution, I went to him. I'd known him from my VMware days. I went to Satya and a few other friends and said, "We have a large amount of data. Can we build a ChatGPT-like capability on top of this data?" And he and others opened my eyes to RAG, retrieval augmented generation. I had no idea. John, if you'd asked me what RAG was, I would have told you piece of cloth you wiped your windshield with two years ago.
Retrieval augmentation generation.
But now I'm like, "This is so exciting." So, we applied for a patent to be the first company that uses RAG for backup use cases. And Jensen saw this and said, "This is amazing. If you're backing up the world's data and you're building Gaia, I'd like you to build Gaia," this RAG-based application, "on top of my platform," the NVIDIA enterprise AI platform. Now, for me, what's it then for me to do RAG? Well, the enterprise AI platform NVIDIA. Think of NVIDIA as a GPU, CUDA on top of that, like an OS layer, and then this enterprise AI platform. If I build on that platform, I get my app to run on top of AWS, Azure, Google, Oracle, and on-prem for free because he's building his platform on all of those five environments. So, it was an easy answer. They put money into the company, exclusively one company in this space, us. And we're working very closely with them.
That's a huge endorsement for NVIDIA to invest in Cohesity.
