Scaling the AI Era: Next-Gen Data Storage Solutions with Pure Storage CEO Charlie Giancarlo

Sriram Viswanathan: 

Today we're diving into one of the most fundamental aspects of our digital world. Storage. Joining us today is a true industry leader, Charlie Giancarlo, a dear friend of mine from a long time, who's the chairman and CEO of Pure Storage. Charlie is a true Silicon Valley legend who has helped shape the very fabric of the modern internet during his career.

Charlie has been at the helm of some of the biggest technology shifts. Over the last 30 years from his long tenure at Cisco, where he pioneered foundational networking technologies to his role in Silver Lake as an investor, uh, and to his various boards, including Accenture, Netflix, Arista, and others. And Charlie is truly a person that can represent the current pulse of Silicon Valley.

And so, Charlie, thank you. For joining us in this podcast. It's an honor to have you on this podcast. Well, thank you.

Charlie Giancarlo: It's a real pleasure to be here.

Sriram Viswanathan: Yeah.

Charlie Giancarlo: Yeah.

Sriram Viswanathan: So Charlie, you know, you and I have known each other. I was just mentioning prior to the podcast, uh, to one of our colleagues that, uh, you and I have known each other since, uh, 1999.

Yes. Uh, you were at, uh, Cisco at that time, and Cisco and Intel. Mm-hmm. And Microsoft had. Gotten to close to 500 billion. Yes. Slightly above. Yeah. In that, uh, period. And you really were instrumental in shaping the internet, uh, for the next 15 plus years. Mm-hmm. And you came into Cisco through an acquisition I did.

That Cisco made, that's right. Called Pana. Can you talk about that first and how that helped shape your, your career into Cisco and how it evolved from there?

Charlie Giancarlo: Yeah. So, uh, Calpana for, well, prior to Calpana, I'd started a company in Silicon Valley that worked on, uh, what at the time, uh, was, uh, high speed, uh, switching and networking.

Uh, and we actually had built, uh, a, uh, a couple of products, uh, one of which became the first. Uh, and, and again, this is all in context of the early nineties, the first high speed backbone for what was then still a very private, uh, government sponsored internet. Capability. And when I sold that company, I went to Calpana, uh, Calpana had a, uh, an early technology, uh, that, uh, eventually became ethernet switching, uh, believe it or not.

So now I'm really aging myself.

Sriram Viswanathan: And me too. Exactly. I remember.

Charlie Giancarlo: to it. Yeah. So, uh, Calpana actually developed the first, uh, literally the first ethernet switch. Um, and several years later, uh, we were, um, acquired by Cisco, which was great fun. And of course, with the Cisco engine that existed at the time, you know, just an incredible sales engine, very strong technology in the area of, um, of networking and routing.

Uh, you know, we were really able to expand that, that switching and routing environment very, very rapidly. And of course, the, the internet started to become known. To the greater public and to organizations after it became a public entity in the mid nineties. And it was just an incredible period of growth.

Uh, you know, Alan Kay, who was one of the, um, inventors of the early personal computer, had a phrase, which was that, uh, perspective ads, 40 IQ points. Mm-hmm. And, you know, the perspective that we had was that the internet would eventually take over everything. Uh, and so that really led, and that's how you and I, uh, came, uh, to know one another.

It led to a lot of consolidation, uh, in what was then a fragmented communications world.

Charlie Giancarlo: Yeah. Uh, as every, uh, type of communication started to be consolidated. Uh, onto internet protocol. Yeah. Right. That's the 40 IQ points. Yeah. Uh, you didn't have to be, uh, a genius. You just had to be, um, have that perspective that everything would collapse, um, into the internet.

Yeah. And that drove, you know, a lot of our activity. Now I have to be honest. Uh, we were able to predict a lot of things such as, you know, all communications going on to internet. Yeah. We didn't predict Facebook for sure. Yeah. You know, and what all the social consequences Yeah. Uh, would be of having this ubiquitous Yeah.

Uh, environment. Yeah. But yeah.

Sriram Viswanathan: Yeah. Just, just double clicking on that and we will talk about Pure and some of the other things, but at that time, in the early days. You one could argue that Cisco got made because of two acquisitions that happened. Yes. Mm-hmm. One was yours, Ana. Yes. And the other was crescendo, correct?

That's right. So you know, and Jeri and your colleague That's right. Came from Crescendo. That's exactly correct. And switching and routing effectively made Cisco what it is. So, so you seem to have picked the right. Sort of inflection. Yeah. In the evolution of telecommunications and sort of networking at the right point in time.

And I wanna draw that parallel to Pure. Yeah. But before we get into pure that journey, in your mind, you know, uh, is that still evolving? I know that Cisco has got its own challenges. Yeah. And things have changed. But where do you see Cisco's role in the networking space As they move forward? Yeah. Uh, in the data center and the enterprise There, lots of others.

That seemed to have sort of ridden that wave much better than Cisco has. That's right. So now that you are beyond the statute of limitations, you can talk about

Charlie Giancarlo: Yeah. Cisco. Well, of course, you know, uh, just when your alma mater falls on on hard times, uh, you know, it, it, uh, it affects you, it affects one personally.

Right. Uh, one never likes to see that. Right. Yeah. Uh, I do think that, uh. Uh, that, well, let's start with, uh, the, the industry itself. Networking. It's like roads and highways. Yeah. You know, it's something that we perhaps don't give much thought to these days, uh, but is so important in terms of the way that we live our everyday lives that we can't imagine life without it.

And so I think that that business, while. While, like roads and highways, we don't think much anymore of when a new one comes in or, or how, uh, instrumental they are. But the, the ongoing development of that area is incredibly important. One of the things that that has become so important, uh, is that, uh, a lot of the.

Uh, competition. A lot of the new technology has gone into the way that modern data centers work. In particular, the hyperscalers could not operate Yeah. Without these incredibly high powered, uh, ethernet switches. Mm-hmm. In particular. Uh, so when at Calta, you know, when we produce our first ethernet switch, it was.

10 megabits. Mm-hmm. Uh, per interface. Yep. So that is 10 million bits per second. Yep. Per interface. The current state of the art is 800, almost, uh, a terabit per second, which is a million, uh, it's, sorry, it's 100,000 times faster, faster. Than what we produced. Yeah. In the, uh, yeah, in the, in the mid or early to mid nineties.

Yeah. Yeah. So, you know, the, it's just incredible.

Sriram Viswanathan: Yeah. So, so let me just sort of step back since you brought up the data center. Mm-hmm. It, it sounds like the first wave of. Uh, evolution of the networking really took advantage of the internet. Yeah. Took advantage of computers getting connected, all of that.

But now it sounds like the second wave is really the evolution of the data center, correct? Absolutely. Right. And the move towards the cloud. Can you talk about the, the three pieces, the compute infrastructure. The networking infrastructure, you know, perhaps the storage and, and memory infrastructure and, and how do these things fit in?

Yeah. And which one has sort of the, uh, you know, sort of the pull position right. In driving this innovation forward? Yeah,

Charlie Giancarlo: so you're exactly right. I've always thought about the data centers having three core technologies, right? Or infrastructures, uh, compute infrastructure, which has been up until. Very recently, largely Intel based, your alma mater.

Uh, the second is the switching en environment, right? And by the way, the, the, uh, capital spend also kind of goes in this direction. Uh, so then you have the, the switching environment by which all of these, uh, Intel computers, la largely Intel computers are held together and communicate with one. Now they're end communicate of course, with uh, uh, with the, uh, applications and the consumers that are, are getting access to 'em.

And then the third one is the data environment, which has to be stored. When people say they store data in the cloud, eventually it has to be stored on something physical. I. There's not literally a cloud in the sky, right? Yeah. It goes to a data center, or several, in some cases, several data centers, and gets stored on a device that is, uh, largely today, uh, hard disk.

In fact, 90% of the hyperscaler, uh, storage environments are still largely hard. Mm-hmm. Now. If you go into an enterprise environment, by which we mean let's say a large bank's, uh, data center, or, uh, most companies, most large companies that you do business with have their own private data centers, even if they are also in, in the cloud, in those, in those environments.

About half the data now is on, um, what's called flash. Memory, which is a semi just to be, since we're bringing it up for the first time, it's a semiconductor, right. It's a chip.

Right.

Charlie Giancarlo: Whereas a hard disc is a literally a spinning magnetic disc. Right. On which, uh, data is written.
Sriram Viswanathan: Yeah.

Charlie Giancarlo: Uh, and so the enterprises has shifted.

They're halfway through their shift to go to semiconductor. Yeah. In, in the hyperscale environment, it's still about 90%. Hard. Hard. Yeah.

Sriram Viswanathan: But I think it's interesting you brought up this analogy between compute infrastructure. Networking and, uh, storage in the computing, uh, infrastructure. One could say the Moores Law sort of is the pulse of the whole correct.

Innovation cycle. Yes. Uh, Metcalf's law perhaps is where networking is. Yes. And, and you, you help drive that Yes. Uh, to its fullest potential. And there is a equivalent storage law. There's an equivalent

Charlie Giancarlo: storage law. So, yeah. Uh, let's go through it. So Moore's law for, for, um, uh, to review for chips Yep. Is roughly a doubling of, of, um, density every two years.

Yeah. Right. You, you know, this, uh, Metcalf's Law was actually a tripling of performance every two years. Yeah. So it was even a faster. Transition over the

Sriram Viswanathan: last, well, when you say that in terms of tripling in performance, you really mean the raw bandwidth, raw bandwidth speed, or interconnect speed

Charlie Giancarlo: between speed between systems, right?

Yeah. So, you know, when, when actually I was in high school, uh, there'd be, uh, I think it was, uh. 75, what was called 75 Bo modems. Yeah. That was 75 bits per second high.

Sriram Viswanathan: I hate to admit.

Charlie Giancarlo: I, I know it. Yeah, you would hear it, right? You could hear it. Yes, I could hear it.

Sriram Viswanathan: Yeah.

Charlie Giancarlo: And then if you go up all the way through access that people get, have people in home, normal people get a gigabit now.

Yeah. A second. Which is huge. And if you were to do that calculation, you'd find that it's about three, uh, tripling every two years. Um, uh, so that was networking. Storage has its own, uh. Hard disks have been in existence for 70 years. Mm-hmm. And they've had a consistent, if you look over those 70 years and you extracted, let's say any particular year, you know, it might be up, uh, might be down in terms of price per bit, but if you take it over the 70 years, it's a straight line.

Yeah. Uh, and, uh, you know, constantly going down downwards. Uh, flash memory, which is also semiconductor memory mm-hmm. Uh, is semiconductor follows Moore's Law.

Sriram Viswanathan: Mm-hmm.

Charlie Giancarlo: It's a steeper. Uh, uh, decline mm-hmm. In terms of price performance. Mm-hmm. And so eventually these lines will overlap, but yes, you're, you're correct.

All of these have a, uh, they, we call them laws, but really what it is, is, uh, the, the return on investment Yeah. Of r and d that drives a very consistent improvement in performance over time. Yeah.

Sriram Viswanathan: It's funny. You should, you should. Uh, we're both dating ourselves with, with technologies that are decades old.

Yes. Uh, I, I recollect, uh, Andy Grove's comment that the laws of physics are the same for everybody.

Sriram Viswanathan:  The laws of economics are different. Yeah. So in a way, most of these laws are really laws of economics. Correct. How can you drive costs down and drive higher performance? So in that context, um, before we talk specifically about pure itself, do you think that the storage.

Evolution is sort of tracking some of the pace at which the, you know, semiconductor related or networking related, uh, transitions have happened over the years in terms of performance, cost, you know, um, manufacturability, you know, any number of parameters. Yeah. Would you say that they are all tracking similarly?

Charlie Giancarlo: I would say it's generally yes. With a big caveat though. Yeah. And the caveat is that. Uh, storage is a particularly conservative market. Yeah. And so users are less apt to switch to a new technology because of perceived risk. Yeah. And the perceived risk is there because if you lose, I guess. Users are, are, uh, somewhat accustomed to a computer, you know, malfunctioning.

Yeah. Or, or not working properly. Uh, they like less than that, uh, when the network stops operating. Yeah. But they're kind of used to network not being perfect. Yeah. If you lose da, if one loses data, people get fired.

Sriram Viswanathan: Yeah.

Charlie Giancarlo: Uh, because you've lost all these transactions. Yeah. You can't prove it. Data loss is considered to be very, very bad, and so it tends to drive a very much more conservative attitude in terms of trying things that are new.

Yeah. Even if proven better. Yeah. Right. So it's a slower moving market. Yeah. Uh, but outside of that, yes. Largely they track one another the same way.

Sriram Viswanathan: But, but you know, in the consumer side, I know that most of what Pure does is in the data center and the enterprise, and you are, you're not really focusing on consumer.

Is that, is that a fair, that is correct. That's. But I think from a consumer standpoint, the user behavior, much to what you just said, um, also had happened in the last 10 plus years. I'm sure both of us remember when, you know, we had to go and delete inbox That's right. Uh, in our Outlook or in our Google Mail, and at some point we stopped doing that.

Yes. Because there was. You know, unbelievable. Unlimited, unlimited storage. That's right. Uh, what drove that? Why was there a inflection where people stopped worrying about storage as a limited, constrained commodity? Yeah. For usage? ‍

Charlie Giancarlo: Well, because when you ran up against the limits of whatever storage you might, might've had before.

‍Sriram Viswanathan: Yeah.

Charlie Giancarlo: It caused a huge number of problems, right? Uh, you know, whether it's the enterprise or whether it's consumer. Uh, let's go back and I'll use this analogy maybe a couple of times, uh, today. Yeah. You may remember that even a decade ago, but certainly 15 years ago, if you needed more storage for your laptop or your desktop, you bought an external hard drive.

Yeah. Right. And you a great, initially fantastic. You had ex extra storage. Right. But what happens when you started running up against the limitation of that? Well, a big hassle. You have to go buy a new one, a bigger one, and then you'd have to move everything you had on the. Smaller one over to the bigger one.

Right. And you couldn't easily share it or anything like that. Mm-hmm. But that was your only choice. Right. So I think two things happened that fundamentally answers your question of why things changed. First of all, we all went to the cloud for storage. Right. And once you're in the cloud, you don't have to worry about your own device.

Mm. You also don't have to worry about having a limit. You have a limit as far as what you've paid for.

Sriram Viswanathan: Mm-hmm.

Charlie Giancarlo: But in a sense, the cloud can have a lot of capacity and so allowing you to go even further is just a matter of paying, uh, just a matter of economics. Mm-hmm. Uh, as you, um, as you discussed. Mm-hmm.

Uh, before. So first of all, it became a lot easier for a user to just add more and more and more. Mm-hmm. They just had to shell a little bit out of there. Uh, out of their pocket. The second reason is because of that ec, those economics of storage, which has just consistently, um, you know, improved over time to the point where, you know, it's a few dollars a month to pay for your extra storage.

Yeah. And as far as people are concerned, that's a lot easier than the hassle of everything else.

Sriram Viswanathan: Yeah, yeah, yeah. So in this context, can you. Um, can you sort of delve into, it's not just more storage, it's not just, you know, cost Right. It's also architectural efficiency. Correct. Which you keep talking about in some of your, uh, your, your speeches and other, uh, uh, writings talk about what is architecturally more efficient about moving to flash as a semiconductor Right. Thing versus hard drives or, or SSDs or any of that. Right. So talk about why is that difference important.

‍Charlie Giancarlo: First of all, uh, like almost like every other transition that we've made in the past when you went from a mechanical system to a, a, a semiconductor base, a chip based system.

Yeah. Right. You got a lot of benefits. You got efficiency, you got cost, uh, you had a smaller space. Mm-hmm. Right. And the, the, the, uh, the semiconductor could do more than the mechanical system, which was relatively fixed. Mm-hmm. Right. In the case of storage. Flash is simply just much more performant and much, uh, much denser, meaning more compact and uses far less energy.

Now, the challenge for Flash in the past was that it was a lot more expensive than Disc. Yeah. On a per. A bit level per byte level, however you want to.

Sriram Viswanathan: Notwithstanding the pace at which Moore's Law was driving the cost down in the semiconductor fa

Charlie Giancarlo: That's right. Faster than, than, um, um, than magnetic was going down.

Yeah. But still higher than magnetic. Yeah. Right. So, and as I mentioned that. The lines are approaching one another. Yeah. Uh, it's still a little bit apart, but, and that's where we are right now. I'll come back to your question. Yeah. That's where we are right now. We're getting close to the crossover point. I see.

And so we're gonna see a massive transition eventually to flash. To flash now. Um, the, uh, the reason for it is, okay, so you get better performance, better density, l lower, uh, lower, um, uh, lower power consumption, and then much more reliable as well, which is very important. Hard disks are mechanical devices.

Yeah. They fail. Yeah. Uh, they fail fairly often. You know, their, their annual failure rate is in the two to 4%. For a year range. Uh, they almost all fail after five or six years. Yeah. So they have to be replaced, you know, fairly often. Whereas in, at least in the products that we sell, uh, our failure rate is down to about, you know, almost a 10th of 1%.

Mm. Uh, we, and, and you know, we guarantee them for 10 years. Mm. So they, they last an awful long, long time. Yeah. And, uh, when you put the full system together, we're literally one 10th. Space, power and cooling. Yeah. Of a mechanical disc device system. Yeah.

Sriram Viswanathan: So just that I, so for our audience, so, so hard drive to SSD, which is really flash based, but it really mimics, it mimics the disc, the hard drive.

That's right. And then there is pure flash. Correct. Which is, which is far more efficient power. Efficient cost comparable to the decline that you're saying in, uh, in hard drives. That's right. But. Isn't it easier for data centers or enterprises that are looking to moving away from hard drives to go through this intermediary STR step using SSDs?

Because that's a faster adoption curve, isn't it?

Charlie Giancarlo: That's correct. Well, that's what they've done. Yeah. And let's talk about this for a minute. You know the SSDs a brilliant invention. Right. Yeah. Uh, flash, uh, works very differently than a hard disk, and we don't have to go into detail. Yeah. But it works very differently than a hard disc.

Faster read, write, get faster. A faster read than write, but you can't overwrite it. Yeah. There's, uh, some maintenance involved. Yeah. So, um, what the flash manufacturers did 15 years ago when they wanted to get into a laptop, when they wanted to get into computers. They thought to themselves, oh my gosh, if we have to force the operating systems to fundamentally change the way they use storage, it's gonna take forever.

You know, it takes 10 years to make changes to operating systems. They said, well, we've got a really bright idea. We'll take a flash and we'll make it look like a hard disk. Everything. It's form factors like a hard disc. The, the, uh, the, uh, the connectors are like a hard disk.

Sriram Viswanathan: Yeah.

Charlie Giancarlo: And there's software, uh, uh, inside the SSD to make it look like a hard disk.

Yeah. Um. The benefits being, you don't have to change any of the operating system software to use it. Yeah, right. The disadvantage is the disadvantage you might imagine of making a chip. Look like a mechanical device, sort of like making a microcomputer look like a typewriter. Yeah. Does great as a typewriter, maybe that's a good analogy, but maybe doesn't do anything much else, right?

Yeah. And so when we got started, uh, because we were the first to use flash, we knew we had to compete against much cheaper, hard disk. And so he said, we're gonna design our own software to make a flash perform. Uh, at its best possible, uh, way, in the best possible way that's both performance and cost. And so by, um, by operating it directly with software, we're able to get about somewhere between a 30 and 40% price performance advantage.

So yes, you're, while you're correct that. Um, it is far easier, uh, to, to be able to use, uh, flash as an SSD, uh, in a, in a large scale environment because you don't have to change any of your software. Yeah. Uh, we're now at the point where the economics, the, the price performance economics with us are so good.

And if you're operating at very, very large scale, like these hyperscalers that. 30 to 40% makes a big, makes a big differe difference. Yeah. Right. Yeah. And so we won, as you may, uh, know, uh, yo meta recently. Yeah. I saw that. Announced that, uh, that we were the ones they've selected as their next generation, their design for their next generation data center.

They can always use the fallback of either disc or SSDs or SSDs, but. You know, their going forward design is, uh, gonna be focused on our TE technology.

Sriram Viswanathan: Yeah. So I wanna, I wanna unpack a bunch of things that you said here, but I wanna get to the software piece that you mentioned. Yeah. And I wanna draw the parallel to what you did while you were at Cisco.

Yeah. Because arguably the, the, uh, Cisco dominance, if I can use that word in that period Yeah. Was largely because of the stickiness that they were able to create with iOS, right.  

Charlie Giancarlo: Which is the Cisco software stack. Well, and actually the definition of the network effect. The very definition of the network effect, the very definition of the  network effect, but really the, the Cisco glue right.

Sriram Viswanathan: For across all the switches and routers and all of that. Right. Got enhanced because of iOS. Correct. So is can you draw a parallel to how you're thinking about software in the storage area? Yes. Akin to that?

Charlie Giancarlo: You bet. Two different ways. Alright, so. Uh, method number one, uh, because of all the limitations that existed in hard disks, if you were to, uh, go into the enterprise storage market today, what you would find is more than a dozen different software operating systems.

Um, operating different types of storage and there are different types of storage Yeah. Which we don't necessarily need to go into now. Sure. But you can imagine high performance, uh, what's, what's called generally cheap and deep. Yeah. There are different file formats or, or block formats. Uh, there are large scale systems.

Hot storage, cold storage, hot storage, cold storage, large systems, small systems. And given the way that a hard, the limitations of hard diss. Uh, vendors would have to create all different operating systems, uh, to manage in those different environments. Uh, we, because, you know, because we could leverage the, the inherent performance advantage of flash and because we chose to only do flash.

‍Sriram Viswanathan: Yeah.

Charlie Giancarlo: Uh, and, and nothing else. Knowing that eventually the economics would work across the, uh. Uh, the landscape. Um, we kept ourselves focused on one software environment, one operating system. Mm-hmm. So every one of our products that really covers that entire space, we mentioned large, small, high performance, cheap, um.

Uh, all the different fo uh, formats of data. We have one operating system that does all that. And that's called purity. That's called purity. Yeah. We call it purity. So, uh, that first of all gives us a tremendous advantage 'cause we can focus all of our attention on the one operating environment. The second way I.

And, and I'll have to go back to the analogy that I used before of the, um, external disc drive. Hmm. If you were to go into enterprise storage today, right. So everything except the cloud and you go into the big banks, uh, the credit card processors, uh, you know, uh, large manufacturing companies, uh, any company actually that manages any amount of, uh, data processing in their own environment.

What you would find is that the data storage systems that they have, each individual system is tied to an application stack.

Sriram Viswanathan: Yeah.

Charlie Giancarlo: Okay. And they might have, you know, uh, several or dozens of application stacks in the same data center, and of course multiple data centers. And each of those, um, uh, each of those arrays operate as a standalone entity.

Sriram Viswanathan: Yeah.

Charlie Giancarlo: Not networked with anything else. Yeah. So these are data silos.

Sriram Viswanathan: Yeah.

Charlie Giancarlo: Okay. Uh, independent. Um, and you know, one of them may be filling up, this goes back to that external disk drives. One of them is filling up and one of 'em is empty. But you can't share Yeah. The capacity or the performance across them.

Yeah. So you mentioned Cisco. I mean, that is my network, is my background. I said, well, how, how dumb is this? I mean, why aren't these things network? Why doesn't operate like a cloud of data? Yeah. Rather than individual arrays, right? Yeah. So instead of that external hard drive, why not take the enterprise?

Why not allow them to make their own cloud? Why do they have to go to the cloud to get cloud? Cloud? Cloud is a way of operating, not yeah. Uh, not a destination. Yeah. Yeah. And so, uh, with, uh, uh, uh, work that we've, uh, been at for three or four years now, all of our systems are upgradable to just the, the next, uh, latest code where they automagically transform from individual arrays into a cloud of storage.

Sriram Viswanathan: So it's almost like what you're really saying is the. Data that somehow got really tightly integrated with the application in which the data is used for. Correct. You are sort of decoupling it and creating data as almost like a data lake available in the cloud for all application and the intelligence That's right.

Shared between all of that, and you are making the point that. Flash and your sort of software solution allows cloud providers and hyperscalers to be able to do that. That's correct.

Charlie Giancarlo: Now, well, hyperscalers have been able to do it on their own. Yeah, right. Because they had that, they built that software on their own ground up.

But the enterprises, enterprises could not,

Sriram Viswanathan: but, but let's, so I want to go back to your Cisco days. I'm sorry, I'm trying to No, please do this comparison with Cisco because these are very relevant and we're gonna talk about AI in a second. But you know, Cisco was very big in cybersecurity and there are a lot of security related products.

Yes. I from an enterprise standpoint, isn't it? Um, isn't it more prone to cyber attacks if the data is all completely shared? Because then once that is compromised, every application is compromised, right? Potentially.

Charlie Giancarlo: Well, it, it's an interesting analogy as, as you know, uh, also if a network is penetrated.

Yeah. Right. Uh, the, one of the problems in enterprise networks was that once an attacker is able to get on the network, then they're able to to everywhere. They're able to get around to everywhere. Yeah. Right. And so the same, you need to have the same protections regardless of whether the data is also networked.

'cause the data is also accessible. In fact, most attacks against data. Uh, today are attacks against the application stack Yeah. That the data is attached to. Right. And that's where the leaky bucket is. Ah. So, um, I would say that it doesn't fundamentally change the, uh, attack profile. I see. Yeah. Yeah.

Sriram Viswanathan: So let's, let's talk about.

Ai. Yeah, because I think you are sort of leading to, you know, cloud infrastructure and, and this whole wave towards every enterprise really looking at their workloads as AI workloads. That's right.

Sriram Viswanathan: Um, can you, I mean, is there a hierarchy of, of, uh, significance or importance between compute, networking, storage?

Memory and power, if you were to take these five things. Yes.

Sriram Viswanathan: Uh, I, I, I don't wanna sort of, you know, you know, minimize your role, not at all in storage, but is that a tail versus the dog in this thing? I mean, which is the bigger constraint? Oh,

Charlie Giancarlo: well, so first of all, there's an economic hierarchy, right?

Especially now, right? Yeah. In terms of where economics are being realized, right? Yeah. Now, we'll, we'll. And we can walk through that, right? Yeah. Uh, and, and then each of these things have their own, um, you know, challenges and opportunities. So let's start with the, you know, on the, um, on, on the chip side of things.

Yeah. The semiconductor side. Clearly, you know, Nvidia and in gen, much more generally GPUs. Yeah. Uh, are, you know, driving, driving the, all of the economics right now. Well, it's

Sriram Viswanathan: almost like saying that the margin profiles is heavily loaded. Towards the providers of GPUs.

Charlie Giancarlo: That's exactly right. Yeah. Now, by the way, that's also very heavily focused on the hyperscalers.

Yeah. You know, uh, I would say between hyper hyperscalers, uh, what's generally called tech titans, you know, uh, companies like X and Tesla. Mm-hmm. Uh. And, uh, sovereign clouds mm-hmm. Are 90% of that, of that market. Mm-hmm. Right. So it's still very strongly felt there. The second, uh, economic winner is networking because all of these systems have to be tied together.

And by far, the, the, uh, most popular, and I think increasingly so. A connectivity vehicle is ethernet and ethernet switches. Right. So they get a lot there. Mm-hmm. Much smaller on that scale, uh, is storage. And that's has surprised a lot of people. Uh, we, we were clear on that. And you would put that ahead of power, energy.

Well, no. So I'll come, let's come back to power. Yeah. 'cause I think, no, I think power. Well, let's come back to that. Yeah. Uh, uh, but, uh, much smaller than, than either networking or, um, or, or, uh, GPUs is storage. Mm-hmm. So very surprising to people, but all of GPT-3 was trained on a few terabytes of data, which is almost nothing.

Yeah. Uh, you know, we estimate that last year, last calendar year, 2024. Uh. The total amount of storage sold into AI was probably about one and a half billion, somewhere less, less than one half billion. Now that, that's nothing to, uh, you know, I think yeah, either one of us would like a billion dollars.

That's right. You know, uh, even just in, in terms of additional revenue, but in a overall 50 to $70 billion market, yeah. It's still relatively, uh, relatively small. Um, power is, is, is going to be a huge constraint. But, you know, the estimates that are made as a straight line on where AI is today, I think underestimates the degree to which, uh, efficiencies will be found in ai.

Uh, one of which by the way, is, uh, I had mentioned that, uh, that our storage is about one 10th of the space, power and cooling mm-hmm. Of traditional storage. Of that is hard dis storage in the hyperscaler. If 90% of their storage is hard disk and about 25%, and this is true of their total power in a data center is taken up with storage, we effectively give them 20% of their power back.

Think of that as a power source. That's great. It's a huge power source. Yeah. Uh, against. All the data centers that are currently in existence. So, you know, there's, that's a huge power source. So I think what we're going to see is as, uh, GPU power efficiency improves, uh, and as, uh, as we start switching over to, uh, flash, we're, we're going to see that there's less power requirements than, uh, which is very good because building power and, and commissioning power plants and power lines and all of that.

Yeah. Is a, a multi-year process and, and very, very challenging. Yeah.

Sriram Viswanathan: And you were gonna, you were going to touch on, you know, memory because memory is the other piece. And memory. So

Charlie Giancarlo: memory is the other piece. So memory, you, the, the challenge with memory and so memory is a growth business. Yeah. And that's gonna be a very strong business.

Yeah. And part of the reason for that is memory has largely reached, uh, sort of the end of Moore's Law. Maybe not the end, but it's really leveled out in terms of. The kind of improvements memory he led Moore's Law. I don't know if you remember that. I remember

Sriram Viswanathan: DRAM was the thing that started the whole revolution.

Exactly. Right,

Charlie Giancarlo: right. Yeah. And so that was really what led the doubling every, every two years. Yeah. And that's flattened out a lot. Yeah. For a variety of reasons. So, um, increases in demand for memory is not gonna be met with decreases in, uh, you know, in price as we go forward. And that means a, a, a strong growth business there.

Sriram Viswanathan: So I want to touch on another common friend of ours. Yes. You know, uh, Sanjay Rora. Ah, yes. Who, uh, who's been on our podcast in one of our earlier. That's wonderful Episodes. Uh, you know, he's the CEO at Micron. Yes. And prior to that, he was the CEO at SanDisk. Correct. And, you know, and

Charlie Giancarlo: I, I, I don't know if you knew this,

Sriram Viswanathan: yeah.

Charlie Giancarlo: That

Sriram Viswanathan: we were lab rats together in, in, uh, graduate school. I, I am aware of graduate school that that's why I brought this up. Yeah. Because, so he, you know, built SanDisk Yeah. As, uh, as a, uh, uh, as sort of a, you know, storage story with flash and, uh, and then he made the, uh, choice of, uh, selling himself. Mm-hmm.

To a traditional hard drive provider. That's right. Yeah. Which, you know, a lot of us in the industry were scratching our heads as to how does that make sense? Right. But it sounds like, um, Western Digital is splitting, they just split. Right. They just split, yeah. Between flash and the traditional hard drive business.

Right. And the reason I bring this up is this question related to the data center mm-hmm.

Sriram Viswanathan: Where you have compute mm-hmm. The GPUs or, you know, acceleration, uh, CPU class engines mm-hmm. That power it. Mm-hmm. And then you have a, a, uh, HBM, high bandwidth memory Yes. Activity, which is what Micron is all about.

Right. And then you have storage. Yes. So, not to put you on the spot, but is there a scenario where the storage providers think of consolidating with memory because they also offers significant benefits to the large hyperscalers? Yes. To see that as being a consolidated offering. For a variety of reasons.

Yeah. Including benefits and power and all of that. Yeah. So, so your thoughts on that?

Charlie Giancarlo: Well, it, it's an interesting conversation and as you may know, just, uh, last week, uh, we introduced, uh, what we believe is going to be the world's fastest storage system for AI environments. Yeah. Right. Uh, you know, we'll, we will beat the bench, the current set of benchmarks by a factor of five or 10.

Yeah. So, very exciting in, when you say fastest

Sriram Viswanathan: is largely latency access? No, it's

Charlie Giancarlo: actually, uh, speed of, of being able to deliver data. Okay. Okay. So, um, GPUs are, are, um, notorious for notorious for, for sucking, you know, needing to suck in, but also write, uh, a lot of data very, very rapidly, right? Yeah. So. The faster you can get the data to it and the faster you can write the data from it.

Yeah. Uh, that's what, uh, delivers most of the, uh, value add in those, in those environments.

Right, right.

Charlie Giancarlo: Um, uh, and of course, uh, on the computers themselves, they have this high band bandwidth memory, memory. Which is extraordinarily expensive. Yeah. So there's this constant tug of war. Yeah. Between how fast can you deliver the data from a less expensive storage system, and how much a high bandwidth memory do you need?

Yeah. So for that reason, well, it turns out as well that the semiconductor manufacturers, you know, mi you mentioned Micron SanDisk, which is, you know, a portion of, uh, Oxia.

Sriram Viswanathan: Yep.

Charlie Giancarlo: Um. They, uh, they produce both.

Sriram Viswanathan: Yeah.

Charlie Giancarlo: Right. And, uh, for a company like us, uh, we are always trying to enable the customer to get the best economics.

Yeah. By producing our product to be able to operate faster so that they need less high bandwidth, uh, high bandwidth memory.

Sriram Viswanathan: Yeah. So in a way you are, you are saying that. If in C two, if memory was there. Yes. And if you can deliver, uh, you know, much faster access, much faster read and write and memory, oh, sorry.

In storage, your potential, your stealing from the memory opportunity. That's right. Right.

Charlie Giancarlo: And by the way, this is classic. This has always been the case. Yeah. Uh, and uh, there's always been this battle between the speed of storage and the need for more and more memory. Right. And for that reason, you know, in a sense we're in the business, but it'd be, I think it'd be unusual to find a company that will compete in both, especially since high BA with memory is sold to the computer manufacturer.

Right. Which is storage is sold to the customer. Right. And, and so it's a different customer. But that blends in the data center though. It does. Well, you're correct about that, right? 'cause the

Sriram Viswanathan: data center wants to have access to. You know, if I have, uh, the, uh, Nvidia GPUs, I want access to a lot of HPM. Right.

And I want access to a lot of, uh, flash, um, uh, storage as well. That's right. Uh, but let's talk about AI in this context. Yeah. And, and data. Mm-hmm. Uh, because you brought that up as, uh, the opportunity to sort of decouple application specific data to more of a horizontal data lake. That's right. Uh, that's a huge opportunity.

But yeah, I wouldn't

Charlie Giancarlo: call it a data lake 'cause that has a particular opinion. I understand. I I, that was a statement. I call it the data cloud.

Sriram Viswanathan: Data cloud, yeah. Is, uh, is data lake is more in the database context and, and that's different, but, but I wanna ask you this question about, uh, training. Yes. AI training and inference.

Yeah. And there's this move towards slow thinking. Mm-hmm. Slow reasoning models and inference versus, you know, fast thinking, GT four kind of instantly gimme a result. Yeah. How does that. Affect you when you are talking about much better performance and being able to offer you the solution very quick.

Doesn't that fly in the contrast of what you're saying?

Charlie Giancarlo: So now we get into the intricacies, uh, or the complexity, if you will, of the AI market. Yeah. Right. You might remember that I said that you. My view of, of the total market that existed for storage in AI last year was, was less than one and a half billion dollars.

Right? Yeah. And it's because of the reasons that you're identifying, which is the developing the models. Ai, machine learning is really mostly being done by these very, you know, by the, the hyperscalers, the sovereign clouds. Yeah. You know, and the tech titans. Right. Everybo, nearly all other companies, with a few exceptions, are gonna be focused on inferencing or re um, retrieval, augmented, uh, generation, which requires far fewer GPUs, um, and does not require this extraordinarily fast storage.

Right? So in my view, the real opportunity, uh, you know, for us. Is in something quite different, which is if you were to go to the most enterprises, uh, who have been studying how they are going to use AI for their, uh, for their own benefit. Yeah. What they would tell you is that their data storage is, or their data itself is so fragmented around the enterprise.

Mm-hmm. They don't understand how they're going to get access. Uh, to it and pull it together to pull insights out of it.

Charlie Giancarlo: And this is why our, uh, concept of, of aligning their various different types of storage into something that looks more like a data cloud. Yeah. Gives them that access to the data Right.

And will make it easier for them. Yeah. To be able to access it.

Sriram Viswanathan: Yeah.

Charlie Giancarlo: And so I believe that's the big opportunity in the enterprise is more along the lines of organizing the data, right? Making it accessible, uh, eliminating the silos, uh, and create and, and doing one of the thing as well, which is creating more uniformity.

This data cloud that we're creating allows customers to create rule sets around how data is, is handled.

Sriram Viswanathan: Yeah. And

Charlie Giancarlo: to do so. In an automated fashion rather than manually

Charlie Giancarlo: You know, on pen with pen and paper. Yeah. And so it, it enhances cybersecurity because the data is managed the way that they set by policy and software.

Yeah. Um, and at the same time makes the data accessible, uh, to, you know, whether it's analytics or, yeah. We don't even have to talk about AI per se. Right. Any kind of analytics or data analysis. Yeah. Even ai. Yeah. It makes it more accessible.

Sriram Viswanathan: So I wanna just create a mental model for, for my own self. Yeah.

And perhaps for the audience. So in a way. Um, the software architecture that you, you have Yes. You know, maybe it's, it's purity. Purity, yes. Uh, is akin to what Jensen is trying to do with Cuda, although it has different purposes. Correct. Uh, it enables a whole class of developers to write to, uh, you know, his GPUs and stuff, and akin to what Cisco was doing with iOS in a way.

Correct. So, is that the mental model? Yes. In fact, I might

Charlie Giancarlo: EE even employ another one. Yeah. Uh, so you, you know, one of the things that you, you mentioned Cisco, one of the things Cisco did was effectively virtualize net, uh, communications. Yes. Right? It no longer every, everything used to have its own communications network.

Yeah. And then it became one thing, right? Yeah. So it was virtualized. Well, another, uh, company and technology that virtualized something was VMware, of course, right. I was gonna get to that. Okay. And it virtualized compute. Yeah. You may remember vCenter. Yeah. So in addition to VMware saying, well, you could run more than one application.

Yeah. Virtual, you know, more than one environment. Virtual machine on a machine. You became a virtual machine. What vCenter did is it said, okay, now you have all of these virtual machines. We'll make it easy for you to run an application on any one of them. Yeah. And to manage where the applications are, are operating.

Right. That's what we're doing with purity, is that it's like vCenter in the sense that now data is managed in an automated fashion. Yeah. Rather than manually. Yeah. Uh, and you could classify, you know, different types of data to be managed in a different way. And again, have that happen in software.

Yeah.

Charlie Giancarlo: Uh, and so. It, it, it's another strong analogy.

Sriram Viswanathan: So it strikes me when you say that, um, I was gonna get to VMware and virtualization, but in the storage context you can talk about containers. Yes. You know, Kubernetes or Anthos, what h HP was doing. And it strikes me that Google came up with Kubernetes. Yes. HPE came up, was one of your competitors came up with Anthos?

Yes. Uh, in sort of virtualizing con container. Right. Technology. And you bought Port Works. We bought Port Works. So, so talk about that. How, how important is that for your core strategy of building virtualization from a storage standpoint? Yes. And containers.

Charlie Giancarlo: Well, actually I'll go into just a little bit of detail around containers.

Yeah. Because, uh, containers is another virtualization technology, but it's even one layer lower, uh, if you will, than, yeah. Uh, than, uh, virtual machine, right. And containers in Kubernetes was designed so that applications could scale very rapidly, right? And then descale, if you will, or right, get reduced also very rapidly, right?

And when that happens, uh, you have to set up connections to storage. Again, very rapidly. Right. And, and so it operates very differently than the way the world operated before that. Right? Generally, storage systems were tied very, very closely to their application. Right. And it didn't change very often.

Right? When it did, it almost always changed manually. Right now in milliseconds, you want it to go up and down and that's what Port Works does. What Port Works allows is for applications to scale for our

Sriram Viswanathan: audience. Port Works is a company, a small private company that you acquired, that we acquired some years ago. Yeah.

Charlie Giancarlo: Um. We saw that the world would, would over time, I think very few people would, would argue that 10 years from now, applications will all be containerized. Yeah. Today they're mostly all virtualized,

Sriram Viswanathan: but if they're containerized, is it, is it orthogonal to what you said about data cloud or is it in the data cloud?

There are many containers per application.

Charlie Giancarlo: Well, containers are the, um. Are focused on the compute side of things. Yeah. Okay. So it's containers, virtualize, compute, yeah. And applications. Right. Um. It actually is orthogonal to the data cloud. Data cloud. Right? Yeah. And whether or whether or not you even have a data cloud, right?

You can still just have one array tied to Yes. A, a container stack. Right. But the fact that you want something to be virtualized on containers, in my view, is you also want your data virtualized. Yeah. So that it doesn't matter where your data is. Yeah. You don't wanna be constrained. Yeah. Uh, by where it happens to be.

Sriram Viswanathan: Yeah. This is all, you know, quite, quite fascinating. We can keep talking about a lot of the technology elements of it. I mean, you seem to have really identified the, the possibly the next great innovation and, and, uh, explosive sort of use, uh, uh, in the data center, in the storage area. But you have a lot of Com competitors, uh, you know, Dell and HPE.

Yeah. So talk about who do you, who do you run into all the time?

Charlie Giancarlo: Well, uh, Dell is by far the 800 pound gorilla. Right? Uh, they, but they came from the compute site. They came from the Well, but they bought EM cm. Yeah. Which is how they got into, uh, storage, right? Yeah. And, and EMC was the, uh, the, the 800 pound gorilla.

Now I. Because they started and still really their legacy is from the hard dis Yeah. You know, in their case they have nine different operating systems for the different, all the different types of storage that are, that are out there, which I think is quite constraining. Um, I. And, and another thing is, is just their business model is much more of a commodity business model.

Uh, and in fact, when I joined Pure, the dominant, uh, uh, um, consensus was that storage was a commodity

Sriram Viswanathan: Yeah.

Charlie Giancarlo: Area and was going to either go to the cloud Yeah. Or become completely commoditized. And, and my view. Was, it's one of the three, we talked about this right at the beginning. It's one of the three major areas of spend in the data center.

In fact, usually it's the second largest area of spend. Um, and for data centers to continue to advance. It has to advance. Yeah. If you treat it like a commodity, uh, you know, there's an opportunity for someone who treats it like a, uh, like high tech. Yeah. To give you a sense, we will spend, even this year we're, you know, as, you know, over $3 billion now.

Publicly traded company, very profitable, but we'll spend 20% of our revenue on r and d mm-hmm. Uh, this year. That's great. Um, many of our competitors spend less than five. Yeah. Uh, because they treat it as a commodity commodity. So it's a very different point

Sriram Viswanathan: of view. But to your point that Dell had EMC, HP is your competitor also.

Also? Yes. And these are your two largest competitors. Uh,

Charlie Giancarlo: and then NetApp was also very

Sriram Viswanathan: large. Yeah, they're very large. But in the case of at least an HP and, and, uh, and Dell, yeah. They have a core business which is much larger in the compute side. So, so do you worry about the fact that for you it is just storage?

And I understand that you don't wanna make that, at least you have your customers perceive it as a commodity. But for the other guys, they have other margin pool that they can sort of. Aggregate storage with Yeah. And go to market faster. Yeah. Is that a concern of yours?

Charlie Giancarlo: Uh, well, it's always a concern in, in the sense that, uh, they, one of their selling techniques is to sell what they call full stack.

Full stack, yeah, exactly. We'll sell compute, storage and, and networking. Although neither one is terribly strong.

Sriram Viswanathan: Yeah. In the networking. Well, which is one of the reasons why HP tried to acquire Juniper. Juniper. That's correct. Right?

Charlie Giancarlo: It's still attempting, I think.

Sriram Viswanathan: Yeah. It's still attempting. Yeah.

Charlie Giancarlo: Yeah. Um. So that, so that is a sales technique, but I, I think that first of all, you have to realize that compute in a sense is a terrible business.

Yeah. Uh, it's the semiconductor companies that make money, not if you're not,

Sriram Viswanathan: if you're

Charlie Giancarlo: Nvidia, not if, well, if you're the semiconductor manufacturer, Intel, and now Nvidia, you make a ton of money. Yeah. But the assembler or the manufacturer of the server. It is in terrible shape business. Yeah. It's a terrible, uh, business.

Yeah. So, uh, they're interesting. I think for both HP and Dell, their top line is driven by compute. Yeah. But their bottom line is driven by storage. Storage. Interesting. Uh, and so it puts them, I think, also in a challenging, uh, position. And I think the fact that. That on the compute side, they spend almost nothing on r and d.

It hurts them from the attraction of good engineers. Yeah. On the storage side.

Sriram Viswanathan: So let's just in the final section, let's just talk about pure a little bit. Uh, you know, your, you, the company went public in 2015. It's about 10 years now. Uh,

Charlie Giancarlo: that's correct,

Sriram Viswanathan: yeah. And you started in 2017. 2017. So you've pretty much seen this explosive growth of, uh, of pure.

Yes. So, so. So do you see this continued growth forward organically? Do you see yourself doing some inorganic moves? And if so, in what kind of areas would you think about it?

Charlie Giancarlo: So, uh, yes, it's been nearly all organic. 20% of r and d is a lot of, uh, you know, capital is a lot of investment. Um, we do, we, we are very active looking at, uh, opportunities, uh, for inorganic.

Sriram Viswanathan: Yeah. But.

Charlie Giancarlo: Even there we're looking, I would say largely for technologies that help us in this march towards, uh, you know, the data, the enterprise data cloud vision.

Yeah.

Charlie Giancarlo: Right. Which is really more around managing data Yeah. Than it is adding storage. If part of our value is that we satisfy all of the customer's storage needs with one operating environment, you can't buy another storage company.

Yeah. Right. If you do, now you have Yeah. Two completely different environments. Uh, and that, you know, uh, reduces, uh, if you will, the value in the attractiveness Yeah. Of what you called iOS. That's right. Of the ecosystem. Right. Yeah. So we have to stay very focused on, yeah. On, um. Being able to complete the entire storage set of capabilities with one operating system.

Yeah. Now in the area of making it look more like a cloud and data management. Yeah. That's an area that's ripe for Yeah. Uh, adding more capability and value.

Sriram Viswanathan: Yeah. I would have to assume that you're probably trying to go up the stack and not just be thought of as a storage provider, but you're actually providing.

Some degree of full stack, you know, maybe security, maybe manageability, and all of those.

Charlie Giancarlo: All of those things. If you think about it, once we, uh, complete this idea of a, of a data cloud of storage, yeah. Then it's all about managing the data on top of it. Right. Right. And that's what's really exciting. Yeah.

You can't manage the data on top of it if you have a different system for every different type of storage. Right. Once you make it, once you virtualize it. Yeah. To use that phrase. Yeah. Now you're managing the data on top of it. Yeah. Or now. We provide the tools for our customers to manage it. Manage.

Sriram Viswanathan: So on that data point, the last last point on the data, uh, uh, point you made, uh, in the AI landscape, in, in the, in inference.

Yeah. There's all this talk about, you know, um, uh, peak data. Yes. And, you know, moving into sort of synthetic data and, uh, all of that in creating new data for training and inference. Right. Do you have a point of view on where that leads, uh, and how that can be helpful to you?

Charlie Giancarlo: I do. I, I'm, my, my view is that more than anything else, customers want to use, they want to build models.

Yeah. On created data, or, well, let me rephrase it. They models will, um. Create data that is the fundamental rule set for the model. After that, what they want to use is as real time data as possible, as well as historical data to come up with insights. Yeah. So it's really more a matter of organizing your own data environment.

Yeah. To be able to gain insights.

Sriram Viswanathan: Yeah. Yeah. It's interesting you say that because, uh, you know, as you know, Celesta invests lots of companies in Deep Tech and some of our companies are actively engaged with your team. Yes. And you know, I, platform nine is correct, uh, very involved with, uh, the whole VMware replacement, uh, opportunities as is.

Cca, which is working very closely with your team. Yeah. So, you know, we are all rooting for you to go all full stack because I'm sure they'll find lots of opportunities to look at in our portfolio. Oh, thank you. Yeah. Uh, but let me just, you know, finish in the last, uh, section. Um. Charlie, you, you, you have been a very understated, uh, uh, executive, uh, but just had just an amazing track record of, uh, achieving and riding some very crucial waves.

Mm-hmm. Uh, is where does this wave lead you to, I mean, do you, do you see yourself sort of continuing in the storage landscape and, and evolving pure into a much larger business? And what's, what's your, you know, what does the next five years look like? I do.

Charlie Giancarlo: The, the next five years, you know, honestly, uh, well, as, as you pointed out, uh, I'm not a spring chicken anymore.

Sriram Viswanathan: Well, notwithstanding, uh, the, the gray beard and the professorial look right. I sort of think of your energy. You are an avid skier, so I am, I am. I know. I look you. Yeah, I think a lot of spring chickens will compete with you. Yeah,

Charlie Giancarlo: well I can compete with a lot of 'em. That's, that, that's actually true. Uh, but no, my, I'm, I'm fully committed to driving these, uh, I'm very mission oriented as an individual, right?

And, and, uh, the missions I, I really wanna complete right now is, uh, driving this capability for enterprises to create their own enterprise data cloud rather than these silos, right? Yeah. That's, that's a multi-year mission. The second one is, I mentioned the win at Meta. Yeah. You know, I do believe that five years from now.

All of the major hyperscalers could. Now I, I don't know that we'll, uh, be able to achieve all, but, you know, a significant fraction of the major hyperscalers will use our technology, right. To replace their existing storage environment. Right. Maybe not exclusively, but to, uh, to a large extent. Honestly, you know, I think as we get older we really enjoy developing people as well.

Sriram Viswanathan: Yeah.

Charlie Giancarlo: Uh, and so developing the, the, the team here at Pure to, to really lead, uh, or to be the leaders of a leading company. Yeah.

Sriram Viswanathan: That's fantastic. Yeah. You've also been very active. In, uh, investments. I mean, you, you ran a big chunk of the Cisco investment. You were at Silver Lake. You drove, I guess the Skype acquisition.

Skype. Yeah. It was one of, one of myself and one of my

Charlie Giancarlo: partners. Yeah.

Sriram Viswanathan: Yeah. So do is, uh, does Pure have lots of investment activity around your,

Charlie Giancarlo: we're just getting started in that area. Yeah. Uh, do we put it in, uh, uh, an investment in Core Weave? Yeah. Uh, we have a, a small investment in a company that's pursuing alternatives to.

Uh, tape as it turns out, uh, which, uh, you know, continues to be, uh, you know, an important, uh, area for data storage. Yeah, yeah. Uh, so, and then we're looking at, at other investments as well. So no, we will continue doing that. I, I would say that, uh, you know, my full-time job here has probably caused me to, to, uh, uh, curtail some of my, my personal investment activity.

Yeah. Just don't have as much time to Yeah, that's right. Pay attention

Sriram Viswanathan: to it. Well, Charlie, this is just truly fascinating to talk to you, and I've always admired your, your career and your journey and all the things that you've done. And really thank you for the opportunity to sit down and talk to you about what makes the storage landscape, uh, important for the data center and the role that, uh, pure plays in that.

So thank you for joining this podcast. You're very,

Charlie Giancarlo: you're very kind, uh, Sherah. It's always been a pleasure working with you. I remember our first meetings, uh, yeah, over at Intel.

Sriram Viswanathan: This was in, uh, you know, and you and I, uh, battled a lot of, uh, issues around wifi.

Charlie Giancarlo: I I was just thinking that myself. Yeah, we, we really did the early days of wifi, which, uh, yeah, today

Sriram Viswanathan: we take it for granted.

But you and I know some battle scars from that period. We absolutely do. Yeah. Yeah. Well, thank you so much, Charlie. I really appreciate you joining us. Thank you. Take care. Thanks for tuning in to the Tech Surge Podcast from Celesta Capital. If you enjoyed this episode, feel free to share it, subscribe or leave us a review on your favorite podcast platform.

We'll be back every two weeks with more insights and discussions on all things deep Tech. Thank you very much. Bye for now.