AI Data solutions are not One Size Fits All with Solidigm

I'm Scott Shaley. I'm currently the Director of Leadership Narratives and evangelist for, uh, soy. I am Not a Day Warner.

Oh, I'm one of the, uh, lucky few that have been able to join the organization and help expand, uh, the capabilities and concepts of what SOY is doing. I have a long history in storage. Many of you may know me from either one of my startups or a local competitor, if you will.

Uh, but what I'm here today, uh, this for, for this part of it, uh, is to talk to you about the one size doesn't fit all concept, right? So we all have things we know that companies make and they try to make sure that they fit the right boxes. And we're, we're talking about square pegs and round holes, things like that.

And if you hadn't caught on just yet, we didn't quite coordinate a tire. So, you know, I, I hope I didn't make it too stuffy in the room. I know it's after lunch and all that good stuff, but we're gonna have some fun with this.

So, um, when we went to GTC, love the keynotes, love where the market's going. I love the refrigerator event. We're gonna talk about that towards the end of this with the liquid cooling concept of it.

But everything that was mentioned, especially even in Jenssen's keynote, was about GPUs and memory. And we all know that we're here at an infrastructure event, and you all now know that we're an SSD company. So of course we have to play that into, uh, that you, you need something more.

The data's gotta sit somewhere and it's not always gonna be in memory. It's gotta reside in a permanent location at some point. Um, and then the capacity and type of data storage you have also matters as far as it goes for developing these tools.

So I was having a little bit of fun with, uh, some ai, um, co-pilot grok. Take your favorite one. And I put in a, a nice little, uh, prompt here from my favorite movie of all time, which is really sad 'cause it's a cartoon with, you know, Robin Williams.

But phenomenal cosmic capacity, itty bitty SSD living space. And after 19 iterations on that prompt and three legal reviews to get a properly legally approved, uh, picture, this is what we came up with as far as what an AI generated image could be for trying to fit as much data as humanly possible into an SSD. So, I, I, I had to keep playing around and going back to my roots of being, uh, way too old, I typed in the concept of a picture of lemmings falling off a cliff of technology with a hero figure.

And there were multiple, multiple prompts here as well related to how to get the right picture. I, last I checked, that's really not a lemming that looks more like a geral to me, you know? Yes.

Um, the, the one I had that I actually liked looked more like a mouse than a leming. But then legal said, you use the wrong prompt tool. You can't use that.

We can't promise that we can put it on a public stage. But the idea is all of this is around concepts of having the right partnership engagement with either your AI tools, your AI infrastructure, and being able to know how to deploy them appropriately in a way that makes it engaging and successful in, in use of that particular tool. So, again, trying to have some fun with uhis and prompting since we're at a field day event and Tagging it, Huh?

Thank you for tagging it. Of course. Uh, that was also legally approved.

Um, and if my lawyer watches this, I'm in trouble. So please, if I need a job, I'll be reaching out. It's Really a pika, huh?

It's a pika. It's a pika. PAA, it's a Mouse.

Um, so obligatory in all presentations when a marketing guy gets up in front of the room is the up into the right graph. 74. I'd love to know how they get the orders of magnitude on that.

If you're math petition at all, it's almost impossible to get two orders of magnitude on a daily basis. But terabytes per day, 174 zetabytes in 24. And, uh, oh, there it goes.

Um, there's your up and to the right. This is the expected data growth in, uh, from 20, I think it's 2010 on there, all the way out to where we're gonna have 181 zetabytes this year. 8 trillion parameters.

And I highlighted a data set that exceeds a petabyte. Now, as we all know, a petabyte doesn't fit in HBM very easily 'cause there's not a lot of HBM that fits A petabyte further emphasizing that the unsung hero of AI is gonna be a storage product. Now, this particular data source is talking about data we generate as human beings, as sensors, things we're all using.

If you take a spin on that and doesn't want to move unless I point at it, um, there's another aspect of this where we're going to exhaust our amount of data that we're, we're creating and we're gonna start using the synthetic data. And this is getting into rag and all those kinds of things like that. And so, um, this particular graph shows the graph from the previous page exponentially impacted by the synthetic data that we're implementing real data and cycling through it and cycling through it and cycling through it and having a lot of fun with it.

And so the synthetic data, of course, is all being done behind the scenes. It's being done in a combination of memory and storage. When you get into the data pipeline later, you'll see petabytes going to terabytes, going back to petabytes, but it's all net new data.

And I see a shaking of the head to the up and to the right. I'm curious. So, uh, I would love to take credit for it, but of course we do have to label everything.

So our friends at Gartner, uh, came up with this, uh, particular view of the world for that AA model. Now, one of the things that we don't tend to think about a lot as well, especially when you get in front of DTC and Jenssen's doing these big pitches and he's showing these amazing racks from all the vendors is getting into the edge and what we're doing around the market in the edge space. So this is a precursor to a recommendation of a future event from one of the, uh, delegates in the back.

Uh, the Edge AI market is broken out by hardware, software, infrastructure services, and it talks to, um, data magnitude, low latency and privacy and security as some of the major concerns for what's happening at the edge. But at the end of the day, it all comes down to what we talked about at the beginning, which is every watt matters. Every square inch matters.

So we have a fundamental problem. Uh, you know, if you're into modern technology and stuff like that, XAI has recently fired up a whole bunch of illegal methane, uh, generators to power a facility they couldn't get actual power to. That adds to your power concept.

Uh, when it comes to space efficiency, we've got a whole bunch of information where you can share. We, we have, we'll touch on it a little bit later, but around the concept of TCO and how can we shrink things appropriately versus continue to expand them. Um, and so one thing that we spend a lot of time inside soddy is the definition of the word edge.

So this is the first audience participation slide. So this is meant to be a little bit more technical, a little more deep dive, a little more fun, and not so much fluff. So ACE and team spent a bunch of time putting this together for us to consider.

So this is our way of trying to use actual data metrics and aspects of generation of AI data and or edge data to define the edge. So we talked to an analyst from one of the many different analyst groups two weeks ago and asked him, what is your definition of the edge? And he says anything not in a data center.

Yeah. Mm-hmm. And I talked to another one and he says, anything that's a single rack somewhere.

And so I can get all these different answers from analysts and from people. What is the edge? So the, the question to the room is, does something like this actually add value?

What do you think of it? Are we down the right path of how we try to redefine or define the edge market as we see it today? So you can see examples we've pulled from, uh, potential, uh, use cases for our definitions.

So we picked on XX AI and Colossus, you could say, you know, the new star or whatever, uh, is in there. Schneider Electric does a great job of defining a box that the servers go into. So it says, here's what the footprint looks like, here's what the power delivery looks like, here's what the air and cooling is.

You go stuff it appropriately. So that's a great one for this kind of standalone. Uh, we have the micro or modular data center.

Uh, we actually have another partner we recently spent some time with where we went and looked at one of their cargo container, uh, immersion cooled data centers, which is sitting out in the parking lot. And there's a great video on solid I about that, uh, where we worked with a couple of partners, uh, Doug, which does not stand for anything you might think it sounds for. And our friends at, um, Hypertech.

And so that's some very interesting stuff that goes around the modular aspect. Then we kind of looked at, well, what does this mean by cabinet? So my definition of a cabinet when we were having these internal discussions is think of Walmart.

So just three or four buildings that way as Walmart Labs. Uh, 1 26 or something like that, I think is what they call it, where they have an infrastructure group that's designed around how to fit the right amount of it in each store. IE the cabinet in every retail store they classify as an edge point and they use interesting data.

And one of the purposes of that is like, if you think about the weather today, little breezy. Overcast. So let's sell something at the front of the store that fits the weather in the mood.

If it's a bright sunny day, guess what? And it's, you know, summertime, of course the watermelon's going up front. Well, strawberries are not really in season, but it's a perfect day for strawberries when you have a sunny day in the middle of winter because people want that fresh fruit and it's so bright and things like that.

Walmart actually has a group that defines how to put a cabinet level data center or edge data center or edge platform in each and every one of their, their, uh, retail stores. Hey Scott. Mm-hmm.

Jack Paul from Paradigm Technica. Just outta curiosity, I noticed that, um, your size is in square feet and earlier you had, uh, square inches squared for the the drives. I'm just curious why not volume and cube feet or Cubic?

Um, yeah, this one is, it's kind of going off the idea of floor space. Floor space, okay. Not the volumetric aspect of it.

Like, 'cause when you start talking about rack heights and all that kind of stuff, it gets kind of messy and all that kind of thing. So we, we centered on what foot print of concrete take up that space. And much like we are with storage, where it was megabytes, it went to gigabytes, it turned to terabytes inches doesn't work for this kind of scale.

So it becomes feet. But when we talk about square inches and our metrics, we're talking about the drives, right? And, and I'll explain that.

Yeah. Okay. That's fine.

Hopefully that makes sense. Agree. Vol volumetric has a lot to do with it as well.

I mean, yesterday for example was Earth Day, as we all know, another member of our team, Dave Sierra, actually wrote a paper that talked about the aspects of sustainability and that TCO modeling of shrinking a large scale footprint. And it actually dives into the cost of concrete to build that data center as an actual impact to a data center design and the impact that that has on the carbon neutrality and all that other kind of stuff. So it's a very interesting read.

com website. But we figured Earth Day was a good day to put out something that shows the actual impact of the concrete used to build a data center floor. Um, so this tried to cover everything we can think about, you know, so for example, we get into the cooling techniques of liquid air immersion all the way out to, you know, you have passive, so recently I was in an event in Dallas around power substations.

Now what is a storage guy going to a storage conference when the two words do not mean the same thing. You walk up to a booth at that event and you say the word storage, they say, oh, how big's your battery? Yeah.

Think about that. But those, even those points in time are dictated on what they can have at a substation level. So the, the power we're getting to power this building is coming from a substation that is now getting virtualized and that virtualized data, that virtualized substation has some form of an edge footprint and it's actually dictated by several very stringent rules on what can and cannot be in that box.

And we've actually had a partnership with one of those companies to deploy our SSDs in a passive non-air cooled environment sitting on an edge in a substation because that's how far out we can go with some of these products. And part of the reason that we're successful with that is the amount of data generated by substation is crazy. And the amount of power allowed to that server, even though it is a power substation, is infinitesimal compared to even your standard, uh, Lenovo think edge platform.

They can't even put a high on Zon in there. They can only go as far as, uh, the Sapphire platform. So this, this concept is something we've been working on to try to define the word edge and is something that we're very curious to see if there's any feedback off the first base.

Yeah. Okay. So this Is a great backup slide, but I, the whole time you're talking, I'm trying to figure out what the difference metrics are.

What's the, what's the X axis, what's the right axi, uh, the what axis? And so I think, I mean you, uh, what I would start with is a graphic to begin with. Okay.

And have it laid out as far as, 'cause also the, the nomenclature up top isn't something that I immediately get core data center makes sense. But then edge data center might, so maybe if you have them laid out, it might make sense. Okay.

Or might be easier to, to get to. And then just pick a underneath it, a couple of these that you think are, are most important. Um, and then I don't see anything like sensors or smaller devices out on the edge.

Yeah. So for the stuff that we're doing, we're data center class products. So getting out beyond where a server's placed is not so far as where we're aimed with our, our current portfolio and technology that gets in more into IOT and client based type solutions where you would see micro servers and things like that running on an M two or whatnot from a storage perspective.

So we didn't step that far out. Right? We Do know think for greater context to figure out, here's the whole thing, this is where we play, This is our, this is our, I look at this and I thought the whole idea was to make, explain the edge to people because there's different definitions of it.

And your definition would then be one that's, it can't be complete 'cause you're just looking at what you're doing, which is sort of the, the next step I would think. Yeah. Seeing the forest for the trees concept, right.

Yeah, exactly. And again, I appreciate that this was really actually meant to be a little bit of a bring your minds up and make you make you talk back to me. 'cause I hate being the guy that presents a dozen slides and never hears from people.

So it's great feedback. I do appreciate it. Sort of asked and answered.

But you, you mentioned that, uh, your edge definition stops at where a server would go to, right? That's, that's kind of that far edge, right? Yeah.

Talking the micro stuff, so, so asked and answered. Got it. Okay.

Yeah, and it, it's challenging 'cause they're, like I just said, there's servers further out than we have ever seen them before and they're gonna continue to be there. Yeah, for sure. So, well I really appreciate the feedback on it.

It's, it's great to get that, that kind of view of things. So, um, what I wanted to do next is kind of walk you through, uh, our family of products. 'cause you know, every good person has to put a, a a family of products up in front of you.

If I'm gonna talk to you and have a whole bunch of people looking at this later. So this is our portfolio. This is where the one size doesn't fit all comes into play.

Um, so we have a plethora of solutions for a reason because not everybody needs everything. Um, I'm gonna give you a little, uh, history lesson here in a minute as well. But what it also comes to shape, uh, shape, size and that whole inches square matters to you.

So I, I'm a hardware guy, I gotta have toys and I'll be back with some more toys later. But these represent one of the newest things in innovation that I was a very proud member to be a part of, which is the invention of an SSD only form factor for data center storage. So previous to something like the EDSF family, you had a two and a half inch, a three and a half inch.

They were all based on, I have a hard drive and I gotta put something next to it. So it's got a look and act the same. And a bunch of us got together, including members of Intel, myself, from another organization and a gentleman led it out of Lenovo.

He says, I'm done with that. I want SSDs only. And born was the E-D-S-F-F or E one S form factor.

And then we brought it into standards bodies and they said, well we version's not good enough. We need at least another one. And there's actually three or four of them now.

And you can see they're slightly varied simply by heat sink and things like that because everybody starts putting in their 2 cents and they say, well those are pretty, but they're just not quite big enough. So why don't we go one step further and make it a little bigger, but still a little skinnier. And again, the next iteration of the E-D-S-F-F is now what we call the E three s.

And this is the high capacity drive because you can see board space wise, we're, we're a board company, right? We've put components on a board. I'll show you a board later just for fun on one of the slides, but how do I get more space for that?

And I don't have to worry about spinning mechanisms, all that kind of good stuff. 5 millimeter, which is generally used in platforms that can support, uh, high levels of fan cooling because there's limited gap between it. Then our friends at Meta and a couple other organizations said, no, we want heat sinks.

We're gonna run it with a few less drives and a few more fins on it to cool it off. So this is the E one s 15 millimeter and there's actually one that's 10 millimeters taller. So you can slow the fans down even more because again, we're starting to get into this, fans become a problem in a data center.

Uh, so I'll pass those guys around. And then this is the E three s and this guy, again, same kind of concept. Let's find a way to put it in a server.

I don't wanna redesign the back plane or the, the metal box 'cause this requires complete re-architecture of the frame. But what if I take a two and a half inch drive footprint and just do the back plane and the little bit of metal that holds the cartridge of the drive. Here comes the E three s and there's now compliments of, again, standards bodies and everybody's voices coming into play.

'cause uh, I'll let you know, I, I'm poking on Sander's body 'cause I actually sit on the board of snia. So I get to have fun and poke at it. Uh, we came up with four versions of this.

There's the E three S one T, so it's length one, thickness one. There's a E three S one L where we extended a little bit longer because the thought idea was to be able to put an FPGA in there and still have room. And then we said, well, well we need to make a little thicker too.

So we have a two T and a two TL. So we've created, now we've created our own monster by going to SSD specific form factors. And this, this wide variety of products is supported in the E one s, the E three form factor and the traditional UDOT two that's represented by your Lego kits on the table.

So what is the T Terabyte? T stands for thickness in the case of a form factor, right? You gotta mix up acronyms and form factors everywhere.

You know, I think, uh, we're up to about 675 acronyms in the dictionary at soy for what does stands for what? And there's at least 20 or 30 overlaps. So thanks.

How many of these were you expecting to get back? Um, you, you, you'll notice there's a little bit of black plastic hanging out of it. Aw, yeah.

I I I have been to one of these rodeos before and lost things. So that one works better. Yes.

So how much storage do these hold? So the E one s today can go up to eight terabytes. Uh, we're working on 16 terabyte versions of that.

It's all based on the media type media package, things like that. The E three s form factor can get up to 32 terabytes today and we have plans to go to 64 and beyond on that. 5, double the height extra board double the capacity.

So that's why we're able to get up to those kinds of footprints. Um, you'll also note on the E one s is the fun purple hook. Those will actually come in a little bit later as far as the end of the day when we're talking about cooling.

Um, the E one f uh, form factor was ACT E one s form factor was designed that the latch, uh, could be designed by the manufacturer of the server frame so that it could match their color schemes and patterns. Because it's one of the very first form factors that is 100% frameless. There's no carrier for it.

It just plugs directly into the system. And that's another one of those things where solid I being just over three years old still has the roots in. We helped invent and design and, and um, nurture these new form factors into existence.

So now that we've shown you a whole bunch of products, and I have not mentioned any of them by name or what they do because that's really not what you're here to learn about, but I'm gonna put it up there anyway, um, we've started mapping it to workloads. And so, um, the AI one shows a nice big block because that's Ace's next section. He is gonna talk about the pipeline and how products fit into it.

But you can see we've been looking at this from a workload perspective from the very beginning. We've never designed a product because, oh guess what, PCI Gen five was just designed. I need a gen five drive.

That's not how you should do things. There are people that do it that way. Uh, but we've certainly been around the block long enough, whether it's you take it, uh, from the day one people all the way through to people that have joined the company.

'cause we see the vision. You have to design a product that's, that fits the true need, right? It's, it's one of those things.

We're not a field of dreams. I can't just build it and they will come. I need to talk to customers, I need to partner with customers.

I need to go in and say, you know what? I've got this really cool thing, but I know it's not gonna work for you. And be willing to have that hard conversation.

And then you start having these kinds of capabilities come out of it. So while this is an AI Field Day event, our products aren't designed just for AI and they're not designed just for one specific use case. We have the capability to look at it and you can see we've got the, the little balls on the, or the circles, uh, on the end to talk about where you can get your capacity from versus your read performance from versus your right performance from, and again, the different products as they fit through that different category of things.

So this is something that's been very near and dear to me because I hate talking about bits and bytes and is it S-L-C-T-L-C-Q-L-C? That's not really what matters to consumers of these products. It matters to me how I build it.

I have spec specific customers that need to know those details. It comes into a whole part of the supply chain aspect of it. But at the end of the day, what's your workload?

How are you gonna solve your problem? Because we've been around the bus long enough if you're as, as young or as old as you wanna call me, that this pyramid has existed for quite some time. And I'm gonna shift it kind of from how SSDs were designed into the concept of data temperature.

Yes. So one thing Scott, I'll bring out, um, and I I had to get through this with you guys in the, as long time as I've worked with you, you guys talk about talking to um, customers. Yes.

And for the longest time I felt like that customers is talking to people like Dell, vast, those kind of guys. But as I've learned from working with the team here quite a bit, customers actually means real end user customers. Yes.

And so that I, I just wanna bring that out because that's different from, that needs to be emphasized in how you talk to folks because that's different from a lot of companies that we've talked to that deal with components Yeah. That go into the work that they're, they consider customers or really just the vendors not the end user. Yeah.

So I I appreciate that call out Kimberly 'cause it is very true. That is one of the things that that made me so excited about joining this organization. And I get the term evangelist on my name 'cause I get to go talk to people that you don't normally talk to and the true end customers that you're talking about.

Like I mentioned Dallas, I went to a storage event that was about batteries because it was actual people that are deploying substations. I met the guys from pg e that are powering this building. I met the guys from Southern California Edison, they're powering my house.

Those are the types of customers that we're actually finally getting to the point of talking through them and understanding with them what they need in these particular solutions. 'cause a lot of what you hear in the day or right now about AI is all about these great big models, right? And everybody has this need and Jensen gets on stage and touts about things like Core Weave who happens to actually be a partner of ours, but who is using Core Weave and how are they using it.

And 95% of the enterprise down under that, you know, very top level haven't even got AI deployed yet. We've got all kinds of modeling on that and surveys that we've produced around talking to these customers at events that are not just the OEMs or the, the VARs and things like that. So I appreciate that.

So How much of that is, is the value you use when you're selling? So you have customers, you have end users, yes. You're Going out and looking at end users even though it's gonna get there, say through Adell or somebody else.

So then do you come to Adele and say, Hey, let me show you who's, who's, who are your customers and use that as part of the fact that to say that we understand this process Process? Um, so yeah, we actually do do that quite a bit. Um, we actually host what we call a customer summit.

And the customer summit involves anyone who's not a tier one OEM for example. Mm-hmm. It's mostly, you know, core partners that are that next level down that are consumers of those products.

They don't bend their own metal, but they buy the metal from someone. Mm-hmm. And so we have to partner with them and find out what they need and then we go back and say, okay, who is your supplier?

And then we make sure that we can get the product they need through that supplier and things like that. Whether it be a Dell and hp a super micro from down the street, whatever that organization may be. Right.

The the IOT platform I was talking about in the substations is from a, a little known company called Lanner Electronics. Right. But they happen to be based in Fremont and they work in that market.

We needed to find a way to get a customer, SCE or pg e or whomever a product. And they're one of those supply paths. Cool.

Yeah. So, um, I put this together because I like to talk about how long I've been doing this a little bit. Um, but when we first started the SSD Rodeo for Enterprise SSDs, it was the cash drive 2007, 2008 STEC sold to EMCA 73 gigabyte fiber channel SSD in a three and a half inch footprint.

And it left the doors at STC at $37,000. And then EMC got ahold of it and their supply chain said, oh, it's just another drive. We'll do our usual 300% markup.

Okay. So now I'm selling a single drive into an EMC platform that costs five times the cost of the whole platform became a problem. So we had to start thinking through it.

The industry evolved. We did sell quite a few of those by the way STC did live on get acquired and is still has its legacies in our marketplace. But we started to say, okay, there's another opportunity.

So that was the SLC drive. It was well overbuilt, it was over overdesigned. We said, okay, we're gonna start looking at this concept of a warm SSD.

It doesn't need to be just a cash drive. And I'm gonna start putting a few more of them in the system. And so I'm defining those as warm here on this chart.

So I'm going from a single drive to more drives. They're a little less on that high speed, high performance, high cost. They're trying to bring that price point back down where we can see the realization of the market.

And then we move into what I would classify as the volume drive. And that's kind of where we started or are at kind of today with a lot of the standard SSD. So you get a bunch of different people selling you a standard drive, it's gonna give you a good performance.

You can put 24 of them in a system, you can put six of them in a system. Make your choice, do what you want. And then we said, oh what we still really like that really hot stuff.

So we started doing memory drives of various different types, including one from our, our alma mater if you will. And so there was a resurgence of, okay, I gotta have a really fast one again. So we get the hot drive back, we have an SLC drive in that space.

So we started with an SLC drive, we migrated to MLC, we moved into a little bit of QLC and then we jumped right back to SLC for the simplicity of the media progression, if you will. And then comes along the AI era and we get to what we call the massive drive. And as you can see, I've kind of been bracketing on the page where it fits in the, the thermal temperature, if you will, of these drives.

And this is one of those lovely debates that's ongoing constantly about where is the hard drive going to die? We are not killing the hard drive. We will never kill the hard drive.

We will, we will play that game. But that doesn't mean that the footprint can't keep moving. And so my question and my offer to you is what's next?

So in the space of ai as we're talking about this, you've got a representation of 122 terabytes in one block, in one deck of cards if you will, uh, available to you today. I can put 24 of those in the system. I can get petabytes of of storage.

So now we're at petabytes scale. What, what's the next logical step? Do we actually continue to look at the high performance end or do we start looking at ways to create products that fit better into even further out into that footprint?

And how do we use those drives for example? So if I'm gonna give you a slightly less performant but larger drive, say I go to 2 56 'cause we've got that on a roadmap and we've made that kind of publicly uh, known that we have that roadmap out there, we can get to a point where we're looking at having five 12 and beyond in these single boxes. What does that mean to the ecosystem?

Do we see it moving further down into it? And what does the articulation of that really mean? Right?

Because right now in the AI space, we're all talking about my GPU is getting so much faster and I can put so much more HBM with it and we know hbms expensive. So then we go to dram, which is still expensive. And you're gonna find out here in the next section as we get into the demo, things like that, that there are ways to leverage the storage and the AI infrastructure that's not really being looked at as well today.

And so that's kind of why I put this. What's next? Do we start trying to look at, okay, I've got a disparity no matter how you look at it, I've got a hard drive and then I've got even the slowest SSD on the market.

There's still a chasm there. And how do we address that as we continue to see the need for the data footprint to stay warmer longer and then still be able to archive it appropriately? And so it's just a, a pondering question to you is, is there something out there that's next that doesn't look at, I have to be the fastest, right?

You've got everybody announcing I've first to gen five, hello, you know, we've got competitors first to gen six. But if you think about it, 24 gen six drives in any server that you have, do the math in your head, that's that's, you're never gonna get the full performance out of those, those gen six drives. You just simply can't, there's not enough lanes in the system to give you full line rate for all those drives.

So why am I putting so much effort on making it so fast? So that's kind of a, a nuance of where we're thinking about this kind of what's next and How much of that can be done in software if compression or fabrics that you lay on top of it. And To what, what extent do you work with people who are developing this so that you can get the most out of your drives?

See that's exactly the point, right? Because there was this long longstanding debate about when will the SSSD kill the hard drive And our friends at IDC and Gartner put this three x number out there, but that was solely on price. It was never about performance of the drive, it was never about the capacity.

It was literally does the dollar per gigabyte come three x? And so that's exactly what we're trying to think about is where does it fit in that we've got these relationships, we talk to these people, we know what the end customers really asking for, how do we get them the product they're looking for in the right way and design. So some of it, yeah, you can do it in firmware.

You, we already have firmware hooks for things like throttling. When it gets too hot, we slow it down because it needs to cool off. We can play with those, those metrics.

But then does that help with the other side of it, the cost picture of it or some other aspect of it. So it's finding the right tuning of those knobs to fit the what's next. And on this screen, I apologize, it's supposed to be a nice solid dime light purple like my shirt, but it really does blend right in on the screen.

I just looked. So, um, it, it kind of almost says it's taking over the whole thing and I wasn't gonna be that bold on in the room. So, but I think to your point there, that's what I'm, that's what we're thinking through.

That's the evolution that we're looking at is stop trying to just be I am the best, I'm the first. And I recently penned an article that legal had some fun with, so I didn't get to say exactly what I wanted it to. But the idea is if you look at most sporting events, there's always one winner at the end.

But do they have to win every game? Do they have to win every race? How do I get to that end game where I'm still delivering what customers want?

So they call me first, you know, I don't have to be on the podium as number one, but I want to be the first person they call. And part of that is designing what they're looking for and and working with them to figure out which knobs. 'cause if I put a product into the market that impacts a current customer, like a direct customer, not the end customer, I wanna make sure I'm not p*****g 'em off.

Right? I worked for a company, we produced this really cool platform. Everybody in the industry thought it was cool except for the two number one consumers of that particular drive 'cause we became their direct competitor doesn't work well.

You gotta find that happy balance. Alright, so a question. So one, one place, you lost me back a little bit.

So do you make the hardware that those drives go into Also? Yeah, we, we make the drive form factor and everything that it plugs into a server, whether it be a Dell and hp, a Lenovo, whatever, we don't make the server. Okay.

We, we, we learn from everybody else's mistakes. We don't need to make the server, we need to make the storage fit the server. Cool.

Um, and then as far as what next, I guess that's probably what keeps you up at night. Yeah. Because it depends on, it depends on so many variables that are outta your control.

Pretty much. Yeah. Well, and the but the variables are out of our control.

If we're not having the right conversations, I can get the list of variables from my customer and try to articulate a product around it. If I spend the time to work with them to make them what they want and not what makes the headline. That's where our prowess of soy comes into play.

We spend the time to not, you can go look at some of the recent announcements in the SSD space and we're not at the top of every list. We're at the top of a few of them and we had the first product here, the first product there. But from a volume perspective, all those guys making the number one noise, they're behind us in market share because we have customers that believe and trust us and we built the product they want, not the one that makes the headline.

So, so let's knock the quote marks off heat for a minute. Let's talk about power consumption and the heat required to Cool. Let's be honest, any storage device.

Yes. Is that something that would fall into that magical what's next? Or is there just as we make 'em bigger, they're gonna run hotter.

Is there any way to get around that? Yes, there is actually, and I, I actually hate, uh, through various iterations of the slide, I had the word cool on here for one of these and it was gonna be too confusing for, for, for us to figure out, okay, cool. From a data temperature versus cool from a physical temperature.

Sure. So yes, there are hooks that you can do. When we first started putting these products out, they were basically on par actually a little less power consuming than a hard drive.

When the very first drives came out, even though they were super fast, they were lower power. But because we're starting to get so big, right? There's so many active components, they all generate power.

The controller, a gen five versus a gen four controller. There's just simply more heat, there's more electrons moving 30 terabytes to 60 terabytes. If you don't do power management on the drive level, which is something is a, which is an innovation in the 1 22, everything's lit up that's starting to drive 25 plus watts.

Hmm. And that's now three times out of a hard drive. So yeah, I replaced three hard drives from a capacity and footprint, but I'm consuming the exact same power.

So it's another knob in that TCO story that we have to play with this to actually cool it correctly. And then as we get to the last half when we talked about the liquid cooling concept, that's how we do even more of that. Help the hardware designers design the hardware to make it easier to keep things cool.

Okay. Thanks. So yeah, a question on this one as well.

Isn't this a little bit why QLC is there? Because I know it goes further than just the price per gigabyte or terabyte, but it's just about, um, if you double the capacity and it's gotta be fast, it, there's more, uh, electronics, there's, it's normal that it's more hot, but when using QLC, the performance goes down, the capacity goes up. But I assume that there as well the particle front still be lower as well.

Um, it can be, I I can't speak for other products in the marketplace, but one of the nuances of QLC is as you program QLC, it takes more steps. Yes. Those Steps take longer, which means the device is active longer, which means if it's not designed right and you don't do the right power management across the solution, that one part will be so warm that it defeats the power reduction you get from QLC.

Mm-hmm. So there are QLC drives that run very, very hot whether they, whether they should or not, but they do. 'cause it's not the, the depth and breadth of the design is not there on some of those products.

Ours, yes, we, we can actually manage that and we've got all the power metrics and we could show you a, a litany of data sheets on what we do to make sure the power budget on the E one s is 15 watts. The power budget on an E three is 40 watts, but we keep it at 25. So 'cause 25 is what the U dot two which is represented by the 1 22 runs at, so For the cooling part of it's stuff that you work on, part of it's the bigger Systems that Dell works on to make it cooler.

And then there's probably consortiums, you, you said you were part of some Yeah. That are driving standards that helps to make it cooler. Yeah.

So SNA works on the form factor side of it and the interfaces. So the Sada SaaS and N VME stuff to some extent. 'cause there's also NVM Express that does specifically NVME and then you have uh, OCP, which is kind of driving the rack level stuff.

When we looked at the different racks, they've got a big push on the open compute. Yeah, the open compute OCP, uh, have a very big impact on that. But there are others out there as well that are working through.

Um, the wonderful thing about those consortiums and the reason why we sit on boards of them and we participate them in so much is you can get 155 companies together, which both good and bad can give you perspectives that we don't all just see ourselves. We're not competitors when we walk in that room and we know that we have to be careful what we say, but we can be so much more valuable to the market by participating in these than you would've been previously. So how much of your own competitive advantage compared to your competitors would be the amount that you can cool.

Compared to what they can cool theirs. Um, I'll explain a little bit more of that in our cooling section. Okay.

'cause we've kind of got it cool partitioned out and we've got a, a couple sections. Cool. So yeah, I I promise you'll enjoy that.

Okay. Um, so as part of this, this is gonna wrap up kinda the first portion of this. So we told you what we're gonna tell you.

We told you now I'm gonna tell you what I told you so that we can wrap up this section. So growth of data is just not going anywhere. It's going to continue to be a problem for us no matter what.

Um, it needs storage, but it needs the right storage. Hopefully we've given you a little piece of how we interact in the, in the market around finding the right storage. Uh, and hopefully I can deliver the right storage to everybody, right?

'cause no one SSD is gonna solve the problem. You're gonna hear from SSD vendors all day every day. And as you do everything that you do, I'm not about the drive as much as I am about the solution.

And that's why we're gonna bring a, a demo in here to show you how we wrap solutions around the drive we sell because that's what's important. But we also know we're not smart enough to be the complete solution. So we need partners and customers.

Uh, we, we've put together an amazing, uh, uh, portfolio of products to help solve those problems. And then, um, we will get into, I showed briefly what they look like aligned to the AA pipeline. But you're gonna get now a deep dive coming up here, uh, from Ace on exactly what products and how we define how products align to the different phases of the AA pipeline.

So I will bid you a do for now. I will be back a little bit later and we'll bring Ace back up on stage.