AWS re:Invent 2025 | Clint Sharp, Cribl

Clips
More from AWS re:Invent 2025

Clint Sharp

Co-Founder & CEO

Cribl

play_circle_outline Bridging the Gap: Addressing the Disconnect Between CEO AI Enthusiasm and Compliance Hesitations in Enterprise Adoption

play_circle_outline Cribl: Transforming Logging Infrastructure to Tackle Data Growth Challenges and Enhance Scalability with Budget-Conscious Solutions

play_circle_outline The future of data management lies in efficient structuring and accessibility.

play_circle_outline Differentiation of Cribl's Lakehouse from traditional data solutions in telemetry.

Info
Transcript

Clint Sharp, Cribl

Clint Sharp

Co-Founder & CEO Cribl

In this broadcast from AWS re:Invent, Clint Sharp, co-founder and chief executive officer of Cribl, joins theCUBE’s John Furrier to discuss the infrastructure realities facing the enterprise transition to agentic AI. Sharp highlights a critical disconnect between executive ambition and procurement reality, noting that while CEOs are pushing for AI adoption, legal and compliance teams remain hesitant due to risk. The conversation centers on the massive data implications of autonomous agents, which generate exponentially more queries and logging "exhaust" than ... Read more

explore Keep Exploring

What are the observations regarding the implementation and procurement of AI solutions in large companies? add

What challenges do organizations face with data growth and budget constraints in their logging infrastructure? add

What is the difference between Cribl Lakehouse and traditional data technologies, and what makes it modern? add

What challenges do IT and security teams face regarding data management and resource allocation? add

bolt Powered by CUBE AI

Clint Sharp, Cribl

search

John Furrier

>> Welcome back everyone to theCUBE's live coverage of AWS re:Invent. I'm John Furrier, your host of theCUBE. We're here for three days of wall-to-wall coverage. Our 13th year covering re:Invent. We've seen it from the early days when it was just a small little cloud, some basic building blocks and primitives, abstracting away servers and infrastructure. Now, we're seeing the agent error coming where the ability to abstract out the data and prepare the new infrastructure to automate work and abstract out work and all that heavy lifting. Clint Sharp here, he's the Co-Founder and CEO of Cribl. He's doing a lot of work in this area. He's been a leader in the infrastructure side with data. Clint, great to see you again. It's almost a tradition every year at re:Invent. We sit down and you give us the scoop on how the data pipelines are going, who's moving what around, what's the infrastructure look like. Great to see you.

Clint Sharp

>> It's great to be back, John.

John Furrier

>> So on my little preamble there, I want to get your thoughts on this, because one of the things we've been covering in New York, at the NYSE studio in Palo Alto and the events is it seems like the enterprise is like stuck in the mud here on agentic and AI. Security resilience bars are high. Everything's in POC mode, one year contracts. There's no ARR, there's no final R on it. It's an annual revenue. It's not a recurring. So you see kind of like people trying to figure out what to do. It's almost as if they're not ready yet. And Amazon actually pretty strong announcements with Nova half-baked with your data. Okay, that's cool. Frontier agents, that sets the direction. If this continues, the enterprises now will have a path to get some wins and some ROI, but they've got to be set up for it. What's your view on this? I know you have an opinion on this.

Clint Sharp

>> Yeah. So I think one of the fascinating things that we've observed, and we're in 49 of the Fortune 100, we're in 30% of the Fortune 500. So we sell primarily to the upper end of the market, is the disconnect between the top levels. So their CEOs are, "We're all in on AI." And then as you start to talk to the people trying to procure AI solutions, like their lawyers and their compliance departments, "No AI. Get AI out of these agreements. We can't buy anything with AI in it right now." They're eventually going to get there. But what they're going to find once they get there, and I've been talking to a lot of CISOs, a lot of CIOs, and I'm like, "Hey, look, agents are going to be the future. Eventually, we'll be able to eliminate a lot of your tier one analysts and we'll be able to 10x investigate, turn them into 10x investigators, make them way, way, way more productive. But if I need to quadruple or quintuple your logging infrastructure in order for you to get that benefit, does that ROI work out?" And they laugh because they're already spending tens of millions on this. If I have to spend 50 million on it, I'm not spending 50 million.

John Furrier

>> Explain that point because this is a really key. This is the promise and the dream on one end and then the reality on the numbers. Explain that again about the cost of the logging. Just explain that.

Clint Sharp

>> I mean, Cribl, since its inception, we're solving this problem of data growth. Your data is growing at a 30% CAGR, your budget is not. And what got you to 2025 is not going to get you to 2035. And so the reason why we even came into existence and the problem that we solve is that customers are like, "If I keep trending this way, I'm already spending $10 million a year on my logging infrastructure. In five years, I'm going to be spending $25 million, but my budget isn't going up like that. So I need to do something different." This problem is slightly different in that agents are so productive. So as they're doing their investigations, they're going off and for every one hypothesis that a human would run, they're running 10 or 50 hypotheses.

John Furrier

>> So they're throwing up more exhaust, more data?

Clint Sharp

>> They're generating more queries. And so if you talk to people about their infrastructure, it's not like they're running their infrastructure at 10%, they're running their infrastructure at 80%, 90% because it's already so expensive that I can't just have idle infrastructures.

John Furrier

>> They're clocking it up as fast as they can?

Clint Sharp

>> And so if I then go generate 5 times, 10 times, 20 times more queries, I'm going to need that much more infrastructure in order to accommodate that workload. And so that's where the ROI is going to mismatch is that the promise is like, "Hey, it's going to do the work of humans so much faster than what humans are going to do." But the humans are generating queries today. They're logging into search bars and typing in terms and trying to find bad actors and find problems and all that sort of stuff. Agents are going to do the same thing. They're just going to do it a much more rapid clip. So we've been talking a lot about agentic telemetry as the future and trying to get your infrastructure ready to get to a modern lakehouse architecture. The data needs to be structured rather than unstructured. It needs to have a well understood semantic model so that the agents will be able to understand the data. It's not just a soup of text that actually is structured into fields and with a well-known schema. And if you start working now, by the time your lawyers and compliance people allow you to go procure an agent, then you'll be ready to go get the benefits of our future agentic future.

John Furrier

>> So AI telemetry, if I hear you properly, it's the setup for AI, meaning you got to understand what's going on and that's going to still need to be logging data too, right? I mean, explain that further. I might need to understand that because is it the same as observability, or is it different?

Clint Sharp

>> Yeah. So I mean, it is a update to our view of observability and security. And what we're saying is, okay, the agentic future is coming, but agents are a little bit different than humans. So as an example, in my own data warehouse, when I go to my BI tool and I ask it questions like, "Hey, how much revenue do we make by segment?" Inside of my data warehouse, I have six fields called ARR. And so which one does it ... It gives me a different answer every time. That's not what I want. I want it to be consistent and deterministic and reliable. We're going to have to do this same type of work for security data, for observability data. Instead of kind of pouring a soup full of logs that is just completely unstructured, we didn't do any prep work to it. We just loaded it in there so that we could search it and get it back. And a human can read that and a human can reason about it. And a human can say like, "I read the log. It makes sense to me. It's not well parsed. It's not structured that well." But AI and agents are not that intelligent. And we're not going to want to customize them so heavily for every enterprise. We're going to want to be able to have these agents workouts out of the box. And so we're going to have to do a lot of work to structure that data on the way in so that the agents will have a rich understanding of the data model and be able to give back the correct answers to these questions.

John Furrier

>> So the Cribl Lakehouse is what you guys have, right?

Clint Sharp

>> Correct.

John Furrier

>> So what is it different? What's modern about that? Explain the difference between the legacy or modern. I call a modern legacy version of that. What's the upgrade with Cribl Lakehouse versus the others?

Clint Sharp

>> Yeah. So we're bringing general purpose data technology, the type of technology that Snowflake, Databricks are building for business data, but we're bringing that to the telemetry space. And there's a lot of work in abstracting away data engineering. So one of the problems that IT and security buyers have is that I have about as many people running my data warehouse as I have in my security department. And that makes sense because they're doing a lot of data prep work and ETL jobs and getting all the data loaded and structured and building reports and all that stuff. But my security team is about the same size and their job is to secure everything. They have to handle all the identity, they have to understand all the vulnerabilities and the remediations, and they also have to ask questions of their log data. And so if we need to have a data prep team that's as equal in size today as the people doing all of security, that ROI just doesn't-

John Furrier

>> That's not even possible.

Clint Sharp

>> It just doesn't work out. And so really what we're doing with our lakehouse is we're saying, "Hey, we're going to give you the benefits of what you would get out of a cloud data warehouse, a lakehouse technology, separated storage compute, very fast query execution, but we're giving you that for telemetry data. So we're abstracting away the concepts of tables. We're abstracting away all of the data prep work that you need to do. You can think of it like you thought of the solutions from 10 years ago. You can just pour data at it. You don't have to do any work, but we're going to give you the properties and the performance of ..."

John Furrier

>> So you're using AI agents to one, make the data lake better. Is that right?

Clint Sharp

>> So we're doing a lot of work on the data prep side, so building parsers so that you don't have to go build all those.

John Furrier

>> When you say dump your data in, just give it up, don't have to worry about it.

Clint Sharp

>> Right. Just pour it at us and make it our job.

John Furrier

>> So prepping the data on the way in?

Clint Sharp

>> Yeah, absolutely. Absolutely. Hundreds of parsers, out of the box structuring, normalizing to industry standard schemas like OCSF or OpenTelemetry, but we get the data come in and automatically put it into a format that's going to work really well to give you very fast query, but also very cost-effective performance because the volumes of data and telemetry are orders of magnitude greater than business data. So if you look at a Snowflake petabyte data warehouse, that's a big customer for them. My larger customers are moving a petabyte of data per day and they're retaining data for a year. And so you really need to customize and build things specific for the telemetry market to make it work cost effectively.

John Furrier

>> And you think that's going to feed the main neural systems of the agents, if my word, but like basically be the key for the agents to perform?

Clint Sharp

>> Yeah. And so then the agents are going to layer on top. And so we are building agentic technology that will help you kind of marry this human generated data. So if you look at how people troubleshoot. Okay, something goes wrong. There's a Slack channel created, there's a Jira ticket created or ServiceNow ticket created. They go start browsing a bunch of Wikis, trying to figure out if there's runbooks and things like that. What agents are going to do is marry all that with your telemetry data. And so we're building that, and then there's also other companies building that. I don't know if you know Resolve AI has a huge booth out there. They're going into this space. ServiceNow is going into this space, but ultimately all of these agents are going to need data. And we are the company that's providing the plumbing and the infrastructure that's going to make it both performant and cost-effective to have access to all of this data.

John Furrier

>> Right. So for people watching out there that are in this mode of, they're super excited, they see what Amazon is announcing. It's clear that it's AI or die market. I mean, clearly AI is here for a long, long time. What do they need to do to get ready? I mean, give your advice on how to set the table for agentic to go on right? Because that's the number one conversation here this week is like, what's the foundational thing? Now, Amazon has things like Connect that was built from the ground up because they built their own stuff, but not everyone's going to have that capability, but now they need to start thinking about what that looks like.

Clint Sharp

>> Yeah. I think we have to consider that agents are really good at doing things that were a path a human has already treaded, right? Something that a human has done before. And so it's going to make investigations way, way, way more productive. It's going to make humans way, way, way more productive. But in order to do that, you're going to have to have better quality data. And telemetry data today, I mean, it's kind of an open secret, but it's mostly garbage. Most of it doesn't get read back. So we're going to have to get really good about deciding which of this data we want to onboard, getting it well-structured and deciding how we can make it ready for agents to be able to interpret it.

John Furrier

>> Yeah. One of the things, as you mentioned, the garbage data telemetry that I'm seeing in kind of the AI factory kind of vibe with NVIDIA with simulation. You're starting to see simulation environments get built just to handle the garbage. It's like the wastewater treatment plant. It's all this stuff. And agents are going to throw off more code. They're going to do a lot more things to track. So your point about data, I mean, we're seeing it with vector embeds. I mean, just words now have context. That's more storage. So there's more data coming from this. So what's your view on simulation, these environments? Am I off the reservation or-

Clint Sharp

>> Well, I think one of the things that you said that hits at that point is your data was growing in a 30% CAGR in a pre-AI era. So I'm already breaking the bank today. And what's going to happen ... And so you can think of telemetry data as this exhaust from machine to machine interactions, but it usually starts with a human, right? I open up an app on my phone, I rebook an airline ticket, right? Behind the scenes, there's all this log data that's generated. Every time I click from screen to screen, there's data being generated. Now, imagine that's an agent doing that for me, but it's doing it at warp speed. And so how much more data are we going to end up with? So back to agent-

John Furrier

>> By the way, on the telemetry side is one thing with just when agents be functional. These are transactions. It's almost micropayments, OLTP like conditions. That actually needs to then go to a system of record. That's more data. It's like, again, just telemetry aside, agents will throw off more data.

Clint Sharp

>> The data growth problem is not going away and the cost challenge is like the budgets are not going to increase at the rates of data volume increase. So what ultimately are we going to do? We're going to have to get smarter.

John Furrier

>> Yeah. That needs a radical solution. Data gravity is still the problem.

Clint Sharp

>> Well, that kind of brings me into my data pawn sort of thing, which I think we thought we're going to bring all the data together, right? And data lakes became a thing. We're going to put all the data into one big data lake, but what my customers are telling me is that like, "Hey, I've got a data lake from vendor A, B and C. Why do I have to pay to move it and centralize it?" And so, one of the things that we're doing with our search technology is federation and we do it really, really well. And we think that the future is I want one lens on all of my data. I want one lens, but I don't want to care whether it's in a vendor's cloud, if it's in my cloud, if it's in my data center. And we think that, that's going to be a key capability going forward, especially as we start to look at agents need one lens on that data. Do I want it to configure an agent to talk to all my different data lakes, or do I want one tool that can figure out how to talk to all of them right?

John Furrier

>> Well, that makes sense because why move the data? It's already landed somewhere.

Clint Sharp

>> I'm paying for it.

John Furrier

>> Yeah. It's sitting there.

Clint Sharp

>> And now, I have to pay for another copy just so that I can query it from one central place. But that seems excessive to me. If we can federate, then we can save ourselves a ton of infrastructure.

John Furrier

>> Clint, always great to chat with you again. You always have the finger on the pulse. I think you're right. The data prep, in quotes, not the old definition. Preparing for AI is definitely real on the agentic. I guess in general for your business, talk about your business, how things going on the business front, revenue, momentum. Can you share some stats on what's going on?

Clint Sharp

>> Yeah, for sure. I mean, the business is growing at 50% year-over-year. We announced last year, or I think it was earlier this year that we had surpassed 200 million in ARR. You can do the math in terms of how quickly that's growing. We're continuing to grow at a very, very rapid clip. We're beyond 1,000 employees now, well over 1,200 customers. You mentioned the Fortune stats, the business is growing at an incredible rate.

John Furrier

>> A lot of big enterprises.

Clint Sharp

>> We love selling to big companies. It's just because I grew up-

John Furrier

>> They have a problem too. That's the problem space.

Clint Sharp

>> In Silicon Valley especially, people like to get started selling to startups, they make decisions fast. They're smart people, they're future looking, but the real money's in the enterprise. We have the patience and we also have the understanding of their problems. In the enterprise, nothing ever goes away. I like to joke that the whole thing stood up by a server sitting under somebody's desk, you accidentally kicked that thing, everything comes crashing to a halt because they've made a lot of choices over the years and we really-

John Furrier

>> They're brittle, but don't kick it. Just don't touch it. Just leave it alone.

Clint Sharp

>> There's probably already a sticky on it that says, "Don't touch this."

John Furrier

>> Password, because the person died.

Clint Sharp

>> Yes, password. Password is password. But we understand their challenges and as a customer's first company, we just love supporting them.

John Furrier

>> Yeah. And the enterprises have the need, Matt Garman quoted on his keynote, "80% to 90% of the value is going to come from the enterprise with agentic." So good tailwind for you guys. Congratulations. Great to see you. Thanks for coming on the theCUBE.

Clint Sharp

>> Great to see you.

John Furrier

>> Breaking it all down here. Again, the data problem doesn't go away. Even with pre-AI, AI will throw off more and more data. If you don't have the strategy for it, you won't be ready to have the agentic workflows, of course, doing our part to bring you the data here. Thanks for watching.