In The Arena by TechArena - Data Insights: Krishna Subramanian on Data & AI Costs

Episode Date: June 26, 2026

In this episode of Data Insights, Allyson Klein and Jeniece Wnorowski speak with Krishna Subramanian about the growing impact of unstructured data on enterprise infrastructure and AI strategy.

Welcome to Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome in the arena. My name's Allison Cline. Today is a Data Insights episode, which means I'm joined by Janice Naraowski. Hey, Janice, how are you doing? Hi, Alice. Thank you. How are you? I'm wonderful. I am really excited for the conversation we're going to have today. Why don't you just start with a introduction of the topic and who you brought with you. Yes, I'm excited today. We're going to talk with Krishna, who is the co-founder and CEO of Cobrise. Krishna, welcome to you the program. Thank you, James and Nelson. Happy to be here. Krishna, I am really excited to interview you

Club Prize first time on the show. So why don't you just start with an introduction of Club Prize and a brief overview of the focus of the company? Sure. So Comprice is a very surprise. It's a is a Silicon Valley company. We're headquartered in Silicon Valley, and we've been around for about 10 years, and we focus on managing unstructured data. And unstructured data is just all the data that isn't in a spreadsheet or a database. So this video that we're recording today, this creates unstructured data. Every time you drive a self-driving car or an electronic vehicle, that creates unstructured data. Every time you go to your doctor's office in an X-3 or an sonogram, that creates unstructured data. So 90% of the world's data today is unstructured. And we created comprise because

customers from our prior to companies told us that enterprises are now starting to drown in unstructured data. It's getting bigger and bigger. And nobody knows how much unstructured data they have or how to really feed it to AI safely, et cetera. And that's why we created comprise. Wow. Got it. That's very cool. Right in far alley, Krishna, being a data storage company, right?

So we're really interested in learning more. But you recently introduced what you're calling a flash storage assessment aimed at large enterprise environments. Can you tell us what you're seeing in the market and what has made this necessary right now? Yeah. Yeah. You're probably hearing some of the news, even as consumers, that hate new phones or Apple Watches, etc., may get delayed in production. And that's because most computers use memory to store information.

And right now, memory chips are facing an acute shortage because AI is actually consuming a lot of the high performance memory. And because AI is using up so much memory, and it takes a long time for suppliers to add more to their supply chain, that means that even, Enterprise storage has less memory. And because of that, enterprise storage costs are going up. So just in the first few months of this year, most storage companies raise their prices by anywhere from 30 to about 75%.

And if you're a buyer in an enterprise, you have more data piling up, but you didn't budget for such massive price increases. So somehow you've got to squeeze in all this data growth in light of these rising prices. So what we're doing is we are product because we look at data across all their storage, we're able to give them an assessment and tell them, hey, you have data going at 30% or 40%. But look, so much of this data is actually cold.

Nobody's actually using it. But it's consuming expensive storage right now. And if we are able to just offload that cold data to the cloud, let's say, and make it look like it's still local, we're freeing up all that space which you can now put new data onto. So you're reclaiming existing capacity without having to buy at these exorbitant prices.

So our assessment, our flash stretch assessment, helps enterprises plan this out and figure out how much they can reclaim of their existing capacity. Now, you are introducing a really interesting value proposition. When you talk to enterprises, how are they thinking about the changes around the interplay between capacity cost and infrastructure planning? And how does your solution help them with that journey? Yeah, a lot of enterprises, to be honest, are just like somewhat surprised at the sudden problem.

Overall, most companies do know that a lot of their data might be cold. But let's just think about yourself, probably on your phone, you take a lot of pictures. if you're like me, I have two pets and two kids, so I have a lot of photos. But yeah, you're probably taking a lot of photos and videos on your phone. But how many of those photos and videos do you look at every day? Right. You're probably keeping a lot of them around, but you're not looking at everything every day. Now, multiply that by 20,000 people in an enterprise.

And you can imagine how much data and enterprise is generating every day, but how much of that is probably cold. Nobody's actually actively using it, right? So most enterprise IT organizations know that they have a lot of cold data, but they don't exactly know which data is cold or why data is growing at the rate that's growing. So they lack the analytics and that visibility into all their data because different users are generating this data, different applications are generating data. So what comprise does is we come in and without interfering with anything they're doing,

we're able to assess their environment. Our software actually analyzes all their cloud in on-premise storage and is able to show them exactly how data is growing, what data is cold, who is using the data, how much it's costing them, and all of that. And then we are able to buy policy, move the right data to the right place. Maybe a longer answer, but yeah, that's how we're helping that. It was great. Yeah, no, it's helpful.

And many organizations today have unstructured data. spread across multiple environments, whether that's on-prem or in the cloud or even across different vendors. What challenges does that fragmentation create when teams try to come in and try to manage and optimize for storage? Yeah, you're right. That most 96% of organizations are heterogeneous. They don't just use one vendor because most of them use the cloud and they have on-premise storage. And the challenge is that many might think, hey, look, can I not just get my storage vendor to just optimize their footprint? But then when they do that, what happens is the storage file system can move data around within its own architecture, but it's proprietary. So then if you actually want to use your data in the cloud or if you want to switch away from that storage vendor, you have to pay a penalty.

It is called a rehydration penalty. You have to actually, ironically, buy more of their expensive storage to get off their storage. So it creates lock-in. So, you know, what we provide is a vendor-neutral way to look at data and to tear data and to move data across environments without any of those penalties. When you look at enterprise storage today, how much of the problem is actually about data growth versus visibility? in you've described this, what data is actually being accessed and how valuable it is? Yeah, they go hand-in-hand, Alison, because if data weren't growing so fast, maybe even if you're

inefficient with storing it and managing it, it would be okay. But given that data is growing on average anywhere from 20% to 100% plus year over year, especially unstructured data, if you take that and the fact that this data, there's very poor visibility. to it, the two compound one another. And there's a third factor nowadays, and that has to do with AI and data security. Because a lot of this data, it's no longer just about even managed the cost of it. Unstretched data, one of the problems with it is there's no common structure to it. It's not that unsructured data has no structure. It has no unifying structure. And what I mean by that is this video is a different format than the document that you might have on your

computer, which is a different format than your X-ray image. So there's no one way to search across all of these. And so what we do is we actually build a metadata database, a global metadata catalog where we impute the information about the data into a metadata catalog. So we bring structure to the unstructured data, which is important when you want to use this data for AI, because you need to be able to find and feed the right data for AI. So that's the third use case that's emerging now, which is, hey, yeah, we want to control costs, we want to get visibility, but we also want to feed the right data to AI. So as I understand it, Krishna, Comprice approaches this by identifying and moving data off

high-cost storage tiers, but from a practical standpoint, how do you do that in a way that doesn't disrupt users or workflows? Yeah, it's a great question. We all want to have a clean house, but we don't want to stick everything in the attic and not be able to get to it, right? So similarly with data, we want all the data available at our fingertips, but if there's a way to do that and still be efficient, then we'd be okay with it. And so that's where this notion of transparency comes in. So what we're able to do is we have a patented technology called Transparent Loop Technology. And what that does is the core data, when you want to take it off your expensive storage and

put it, let's say, in the cloud, we actually, when we move it, we leave behind what we call a dynamic link. And basically, it looks like that, let's say we moved an X-ray image to the cloud. It looks like the X-ray image is still local. It's like a shortcut. And you can still see it. Your applications can still open it. But when they go to open it, then we would stream then data from the cloud instead of it sitting locally.

So it's transparent to users and applications so they don't see any change. Maybe they may see a slightly higher latency when opening it other than that there's no change to them. And so there's no friction in the process. Now, you've used an analogy of moving seasonal items to free up space while keeping them accessible. How important is that balance between cost optimization and seamless access when organizations evaluate solutions like this? Yeah, it's a great question. you said seasonal, our winter coats are not always in use, but when it's winter, you want to use them. And data similarly has sometimes sickly the life cycles. So let's say you've finished a project

and you're done with this project so we can move that data off to the cloud, but you may be building another project that relies on some of the data from this project. And at that point, you want this data again. And that happens often for unstructured data. And so we not only make it transparent and easy to get to the data, but we also have mechanisms like bulk recall, where we can bring the whole project back if it's active again. It's important to have flexible recall mechanisms for customers to address these different use cases. And beyond just freeing up capacity, what kind of operational or financial impact or organizations

typically able to achieve when they're more strategic about approaching and managing unstructured data? Yeah, it's a great question. In unsurited data today, it conceals about 30% often of IT storage budget. So it's a pretty sizable cost factor. And the cost of unstructured data is not only in the storage. It's also in the backups and the ransomware recovery copies. So there's many copies made of that data.

And when we take the cold data and we tear it, we not only shrink the storage, by 80%. We also shrink the backup and the ransomware DR copies by 80%. So the savings are about 80% of storage and backup costs, which can be pretty significant because unsuitary as a sizeable portion of their state. And if you add to that, we're actually cataloging the data as we move it, we're building this global metadata database.

And if you're able to feed the right data to AI, we're seeing that AI accuracy, increases by about, in some cases, like 135% when we do that, because we're able to feed AI a smaller subset of the right data as opposed to the wrong data. Now, as storage costs fluctuate and infrastructure demand continues to rise, how do you see enterprise data management evolving over the next few years, especially as data intensive applications like AI continue to scale? I think this whole memory shortage is front and center in people's minds right now.

But I'm confident it will get resolved in a year or so, right? We either have more memory supply or we'll get more efficient techniques. It may be not be as urgent. But power is becoming a short supply. A lot of data centers don't have power anymore to power even the workload they want. So I think the broader trend is that resource scarcity will continue. And so you need to be more efficient about how you store and use data. But more importantly, storage is going to become important for compute as well.

And let me tell you what I mean by that. I don't know if you've seen recently, but some of the organizations that really start adopting AI are suddenly seeing that their AI token costs are just skyrocketing. And so how do you reduce the AI token costs? meaning how do you reduce how many tokens you need to use for AI? The best way to reduce token costs is to reduce how much data you're giving to AI. Give it less and more relevant data.

And that also needs good data management. So data management is useful not just to reduce storage costs and memory costs, but it's also useful to reduce AI token costs, which is going to become really important. I have to say this is like one of the best conversations that I think we've had, Alison on this show. So, Krista, I've learned so much from you, and I would love for you to tell our listeners and infrastructure leaders listening today, what's the first step they should take to start freeing up the pressure of that rising storage costs and capacity constraints? How can they manage all this? Yeah, Janice, you can't manage what you don't know. And most people, most IT leaders, I think the problem

is they don't know really what the issues are with their unstructured data. So it's thoughts with that assessment. So I would say that's why we added this flash stretch assessment service. And it is complementary if the customer is a qualified customer, if they have at least 500 terabytes of data, they can reach out. And if they want us to help them, we're happy to provide this assessment to them so that they can understand because visibility is the first step, I think, to solving this problem. Thank you so much for being on the show today. That wraps another episode of data insights. I think that when we look at everything that we've talked about, is there a place that you think we would send the audience for more information? And how is

the best to engage your team? Yeah, thanks, Alison. Yeah, our website, www.comprice.comprice.com. Has a lot of information. And there is, if somebody wants to talk to us or reach out to us, they're contact us buttons right there. But they can also see demos. They can reach research the situation better if they'd like on our website. Thanks so much for being here. And Janice, that wraps another episode. Thanks so much for the collaboration. Thank you, Alison.

Same. And thank you, Krista. Thank you, Janice. Thank you, Alison. Thanks for joining Tech Arena. Subscribe and engage at our website, Techorina.ai. All content is copyright by Techorina.