What Cost per GB Doesn’t Tell You About Long-Term Costs

Is your storage strategy built for the future or setting your business up for failure?

In this episode of the Fortuna Data Podcast, Ray Quattromini sits down with Matteo from Kingston Technology to break down what IT leaders, CIOs, and infrastructure decision-makers need to understand about modern storage.

With data growing faster than most organisations can control, and AI workloads driving unprecedented demand, the wrong storage decisions today can lead to higher costs, downtime, and long-term risk.

This conversation covers:
How to choose the right SSD, NVMe, or memory for your workload
Why total cost of ownership (TCO) matters more than cost per GB
The real differences between SLC, MLC, TLC, and QLC NAND
How AI is reshaping infrastructure and driving on-premise demand
The hidden risks of cheap storage decisions
Why storage is now a board-level conversation (not just IT)

You’ll also gain insight into enterprise storage reliability, endurance, and performance trade-offs, plus what’s coming next in storage architecture, including PCIe Gen5 and composable infrastructure.

Who This Is For
CIOs, CTOs, and IT Leaders
Infrastructure & Data Centre Managers
Businesses scaling AI or high-performance workloads
Anyone evaluating storage, backup, or long-term data strategy

00:00 The storage problem businesses are ignoring
00:22 Introduction: Fortuna Data x Kingston Technology
00:40 Kingston’s history and evolution in storage & memory
01:47 Matteo’s role: bridging customers and engineering
02:44 Storage shortages, rising costs, and supply strategy
05:07 Military-grade encrypted storage explained
06:17 NAND types explained: SLC vs MLC vs TLC vs QLC
08:46 Matching storage to workloads (and why it’s often missed)
09:48 Endurance vs capacity vs cost: making the right trade-off
10:42 What makes Kingston enterprise storage different
11:29 Testing, reliability, and lifetime warranties
13:43 Enterprise SSDs, NVMe, and performance benchmarks
15:25 AI workloads: hype vs real demand
16:14 Why businesses are bringing data back on-premise
18:08 Why cost per GB is the wrong metric
19:23 The biggest storage mistakes businesses will regret
21:02 NVMe vs SATA: where each fits
22:18 Why SSDs are NOT ideal for long-term archiving
24:01 The future of storage: architecture over hardware
25:10 CXL, memory pooling, and composable infrastructure
26:09 Silicon photonics and the future data centre
31:34 Why storage is now a boardroom priority
32:45 AI, cyber resilience, and data growth pressures
33:46 Energy efficiency and smarter storage decisions

If you're rethinking your storage strategy or planning for AI-driven workloads:

👉 Visit Fortuna Data for expert guidance on building resilient, scalable infrastructure
👉 Or request a storage architecture review to ensure your decisions today won’t cost you tomorrow

#DataStorage #EnterpriseStorage #StorageStrategy #ITInfrastructure #DataManagement #ITCosts #CostOptimisation #TotalCostOfOwnership #EnterpriseIT #CIO #CTO #DigitalTransformation #TechStrategy

Transcript

SPEAKER_00 0:00

Data is growing faster than businesses can control it, and the wrong storage decisions today will cost you tomorrow. From AI workloads and rising costs to supply chain pressures, organizations are being forced to rethink how they build and scale their infrastructure. So how do you choose storage that actually lasts, performs, and protects your business long term? Let's dive into the conversation.

SPEAKER_02 0:22

Hi, I'm Ray Cortramini from Fortuna Data. Welcome to Smarter Strategic Thinking. Today we have Mateo from Kingston Technology. Mateo.

SPEAKER_01 0:30

Hello. Nice to see you. Thanks for having me.

SPEAKER_02 0:33

First of all, do you want to tell our viewers a little bit about Kingston and who you are, where you're primarily based and founders, etc.?

SPEAKER_01 0:40

Kingston Technology in General was uh founded in 1987. We're still privately owned by two gentlemen, David Sun and John 2. We've started back in the days with just one product that was a single inline memory module that was a revolutionary memory solution due to uh shortage. And since then, we've developed into different product categories much more. So we started with just that one, but the plan was to expand across the board in the technology. In fact, for the past three decades, almost four decades, we've become a trusted brand storage and memory, not just storage and memory. Fun fact, people mostly know us for our USB drives, which are really fancy, they're very popular. But not just USBs. In fact, Kingston is much more than that because we cover the full spectrum from the home overclocker to the Internet of Things to the data centers. We have SSDs, all sorts of SSDs, memory flashcards, USBs, standard as well as encrypted, and of course, our memory solutions. We're committed to reliability, to very high quality products, to expertise that is open to everyone.

SPEAKER_02 1:42

How long have you worked for Kingston? What did you do before, and what your role today?

SPEAKER_01 1:47

So I've done all sorts of things. My background is very diverse. I've been in Kingston for almost for a bit more than three years now. So I've started as a technical support specialist where I was taking care of post-sales, dealing with uh issues, you know, trying to solve problems, troubleshooting, that sort of space. Whereas right now I've I've developed in these three years and I've become a field application engineer, which might mean something or might mean nothing. Yeah, it depends how you see it. But basically, what I do is I am the bridge between our customers and our engineering department. Customers can reach out to us, to me and my team, literally directly. And what we want to do, we want to find the best solution and solve deployment and compatibility problems before they become production issues. We validate, we optimize, we support the deployment of memory and storage solutions.

SPEAKER_02 2:44

From your perspective, you know, we all know that there's a a severe storage shortage. How how do you think that businesses themselves should handle this shortage that's two, three, four years, depending on what who you read, you know? And the prices are obviously going up and up and up as well, aren't they? So, what what is your take on this? And how does Kingston as an organization try to circumvent some of those shortages?

SPEAKER_01 3:11

We managed to play with as a result of our long-term relationships. We have worked with key manufacturers for uh for decades, and these these relationships were not did not just happen, they were they've always been strategic. We have a disciplined multi-source strategy, so we don't just source the components from one manufacturer, we use several of those, so we always have a choice where we want to buy our components and then assemble them together because we're still privately owned. We prioritize our long-term relationships, so we don't just look at the short-term win. We want our standing relationships and partners to know that yes, exactly. It's more of a continuous approach with the supply. We're also proudly channeled first, so we always prioritize the channel. I know that Tom from QNUP was here. Yes, he was. It's one of our latest and most successful relationships with our partner, for example, with QNAP. We've worked together with their engineering department to cross-validate products. Right. So in their QNUP, in their QNAP systems, our DC 600 series and our DC 3000ME series enterprise grade drives are now certified and validated to work in their QNAP systems. So that's an example of the outcome that our close relationship with our partner can have.

SPEAKER_02 4:26

Yeah, that's how that's how we met you guys. We've got a QNAP over there that's uh for customer evaluations. And um, we met Kelly and her team at the QNAP event in London. So we heavily rely on QNAP. You know, we work with QNAP, we're a QNAP distributor, so that's really good. This is off tangent, but last week I went to um Space Expo at Excel. And um, do you do any sort of military grade space type memory?

SPEAKER_01 4:57

We support those with mostly with our encrypted solutions. So that will be the Kingston Iron Key product line. These are hardware encrypted USB drives. We have customers that use them in military environments because they cannot be they can they physically cannot be tampered, and they have some certifications that protect them from physical tampering to physically to physically access the chips. So you cannot physically like open a drive and access the chips and remove them and put them in some other mission impossible then. Nothing on mission impossible, no. Unfortunately, no. With those with those iron key drives, your data is is is not cannot be accessed unless you have the password. You cannot physically get to the chips because these military grade drives are filled with uh are filled with epoxy, which soldiers every all the components together, and if you were to open that drive, you couldn't really physically access to the single chips because everything is soldered together. So the moment you try and break and and and open the drive, it'll break. And it the drive will detect as well some physical tampering, so it'll kill itself before you can actually get to the chip.

SPEAKER_02 6:07

From our point of view, there's a lot of different types of chips. You've got SLC, MLC, TLC, QLC, and then you've got DWPDs, which are drive rights per day. You know, if you could explain to the audience what that means, because I know some of our partners and vendors have bought storage in the past and said, Well, after six months it stopped working, and then you say, Well, what's your application? And then they go, Oh, well, we're doing this or we're doing that, and then they start scratching their head because they they're not aware of these issues. So, do you want to just explain what that is?

SPEAKER_01 6:38

Sure. So you have different types of NAND and different grades of NAND within the same categories as well. So you have SLC, which can accommodate only one bit per cell. That means one piece of data, either a zero or a one. We're talking binary code. Yeah. So either a zero or a one in one single cell. Then you have multi-level cells, types of NAND which can accommodate two, triple-level cells can accommodate three, and quadrile cell can accommodate four. So basically, it's all linked to the endurance of these of these cells. You need to think of it like, you know, imagine a massive building with a lot of cells that are identical. And then you want to put some people in there. These are the bits of your these are your bits of data. You put one person in a cell, this person will not argue with anyone, perfect, nice and chill, they're fantastic and can stay there for long. You start to put two people in that cell, they may have some arguments at some point, you know, there may be some problems. You put three people, even worse, four people, even worse. And the things that you use in this cell will wear out quicker because you have more people using those facilities, say, for example. So since this is related to the endurance, basically a single-level cell will have will last longer and have more endurance than a quadrile cell generally. But then you, as I said before, you have different types of NAND. You have enterprise grade NAND and you have consumer grade NAND. So this does not always mean that a quadrile cell will last shorter than a single level cell. There's always a balance how you use the NAND. You can also program, for example, a triple level cell to work as a single level cell. So that would be a PC P C SLC, pseudo-single level cell. So there's all these sorts of decisions that need to that need to be made beforehand when choosing what NAND you put in a storage device.

SPEAKER_02 8:28

What companies need to work out is what their workload is. If it's video and post-production, they're doing a lot of editing, then I'm a I'm assuming you need like an SLC or a TLC, not you know, and then work out how many drives you're gonna be doing and drive writing and everything else. So and that and that's that's often overlooked.

SPEAKER_01 8:51

It's often overlooked, that's true, and that's where we come. That's where my role, for example, is therefore. So I'm there to help customers make the right decision. Now it's not just about SLC, TLC, MLC, QLC, it's not just that, because the based on the different type of NAND, you can have a TLC that has great endurance, this happens. Our enterprise drives, for example, they're not all SLC because you wouldn't need that. That's also more much more expensive. So it's always about finding the right balance. For enterprise great drives, the Kingston ones, we always advertise the D the drive right per day value. It tells you how long the drive will last over its um and how much you can drive you can write into the drive over its life cycle. So this can really help you make the right decision. If you need to replace a number of drives, you want to pick the one with the DW, uh the drive right per day value that suits your workload.

SPEAKER_02 9:48

For example, if you've got a two and a half inch S SSD with SLC chips in it, that is going to be a smaller capacity than maybe a similar drive with QLC, but it's you've you've got to weigh up whether you want drive rights per day or you want capacity, that's what it boils down to in this longevity or performance or capacity.

SPEAKER_01 10:10

It's always about that the price, the price, longevity, capacity, because of course SL SLCs that you will use more physical space on the PCB because you need more chips, because in in those famous cells that in the building that we mentioned before, if you if you can if you can put only one person in each cell, then you have less bits. If you can put four people in the same cell, then you have more capacity, but lower cost.

SPEAKER_02 10:31

What makes your storage memory different from other vendors? What do you do? Do you do any extra check-in, or or you you've got certain vendors that you always go to? What makes Kingston different?

SPEAKER_01 10:42

What makes Kingston different is generally the reliability of our products. And this this comes from BOM bill of material control, for example. We've mentioned that we can buy from from different manufacturers in terms of hand and controllers and memory. We apply a strict bomb control for enterprise grade products. We only select certain components and we try to maintain the bomb as stable as possible. So we don't want to make many changes there because we know that stability is critical for enterprises. Every time you make a component change, organizations might need to revalidate and retest that build. So we want to reduce the downtime for them to retest the drives. From this also comes firmware stability, because so the firmware is as stable as possible because we try to change the components as the components as less as possible. All our products are 100% factory tested. So every single product that leaves our factories is was tested beforehand. Now, depending on the application of this product, we do different tests. For enterprise grade, for example, we do stress tests for memory, burn-in tests that simulate the workload of several months on a single memory module to expose some bad cells, you know. May there be any. And if one single cell is defective, then the product does not leave the factory. So this gives us the trust to offer a lifetime warranty on our memory on our memory products, which is also something we're really proud of.

SPEAKER_02 12:07

Yes, quite unusual.

SPEAKER_01 12:08

It's a lifetime. And on the SSD, Enterprise Great SSDs, we offer a five-year warranty with NAND. This will wear out at some point. And this is normal. It's not just Kingston's. Again, we're channel first. For the configuration that you need, you can use tools, for example, the Kingston Product Finder, what we call the configurator, which is on our website. You simply look up your system and you will find a list of products that are tested, verified to be compatible with your system, be it an OEM, be it a general motherboard, uh server, desktop, anything. You have you'll have the products that are verified to be compatible there. We're happy to share BOM details. If you need consistency over the components, we offer solutions like full lock for memory or DRAM lock. Right. For full lock, we we lock all the components of the memory module so that you're sure that whatever you buy now is gonna be the same the next time you buy it, because the part number tells you exactly the components that are built on that memory module. And this is important for enterprise customers when they want consistency. Because you know, if you change the components, that might be something. Fluctuation. We all we're also happy to share transparent roadmaps. So whenever there's big purchases over years or months, we're happy to be transparent and tell you, okay, this is going to be available in some months, or you may want to pick this other one because it's going to be more available. You know, it's always very it's always changing that.

SPEAKER_02 13:31

And what what is your largest product at the moment in terms of capacity and what interfaces can you support from an enterprise product? For SSDs, yeah.

SPEAKER_01 13:43

So the largest one at the moment is 15 terabyte with DC 3000Me. For enterprise grade SSDs, we cover uh SATA and NVMe, PCIe Gen 5. So SATA, we have DC 600 and DC 600ME, which is the 2.5 inch SATA solution up to 7.6 terabytes. We have the standard version, the 600M, and the encrypted, self-encrypted version, which is DC600ME. So that supports self-encryption with AES2256 for those who need it with TCG Opal or similar encryption encryption encryption solutions. And then on the NVMe side, the DC 3000 me that's used for AI workloads because they are purpose-filled for these types of workloads where you need that kind of speeds and you need to move massive data sets into the memory and you need to do it fast and efficiently because memory will always be faster than storage, of course. And you want the storage to be as close as possible in terms of speed. So if you use MVME Gen 5, then you can really move those data sets really, really fast.

SPEAKER_02 14:51

And you said that you support PCIe Gen 5.

SPEAKER_01 14:54

What are the speeds that that's a theoretical maximum of 14,000 megabytes per second? That is really shifted. And we can achieve those speeds. Uh but the benchmark charts are quite clear. We can achieve those speeds with this drive with the DC 3000ME, and also Sumer Grade SSD, we can get to those speeds with our um Fury Renegade G5 SSD. That's also a PCIe Gen 5.

SPEAKER_02 15:17

Are you as an organization seeing a big uptake in AI or is it vaporware? And what are your thoughts on the whole AI market? Well, this is this is a big question.

SPEAKER_01 15:28

There's lots of cash being we're seeing a lot of demand in terms of being ready for it. Because you know, everything there's a there's a switch towards AI workloads. Organizations want to use they want to have their own AI workloads and they want to run them locally, you know, for cyber resilience and for data security. So you want your AI engines to run locally and you need the hardware to do that.

SPEAKER_02 15:55

In your opinion, are organizations bringing back some of their workloads because of you know, if we're talking about AI, this it's quite bandwidth intensive, you know, and whilst you can you can have a little piece of wire like that going to the internet, it's not exactly like this, is it, when you can get on premise? So so are they bringing them back specifically for those workload cases?

SPEAKER_01 16:18

We are seeing organizations bring back data closer to where it's generated, especially in the European Union, a bit for uh data sovereignty and data security. But yes, generally organizations now I say I see that they are condensing more workloads into more powerful servers, but they want to have them running locally instead of relying on um on external other on external data centers that do the work for them.

SPEAKER_02 16:43

People like to have their own their own machines, so they're sort of building a fortress around their environment.

SPEAKER_01 16:50

Yeah, it's it's related to the to to risk assessment and to data security, yes.

SPEAKER_02 16:55

Regarding memory, I don't I I I don't know how many memory manufacturers are there, but obviously you work through the channel. So if we've got viewers looking at this, what how would you you know we're we're it's not all about price at the end of the day. No, not at all. Sometimes in in a business they think, well, memory's going up in price as as storage is going up in general. Being with Kingston as a as a provider, what what you're telling me is you've got that historical relationship, you've got certificates of conformance in terms of you can track back in terms of you built that component using that system, and that that to me is quite a valuable thing from an organization, especially when they're building these systems, trying to build these systems if they could get their parts together, and they need that ruggedness, reliability, and performance. Is that is that what Kingston's bringing here?

SPEAKER_01 17:49

Definitely, yes. So, as you said at the beginning, it's not just it's not at all about the price, not at all. In fact, we're there to help the decisions makers ask the right questions prior to an upgrade or a replacement. Because usually conversations start with uh, so what's the price per or what's the cost per gigabyte? Yeah, it's not just about the cost per gigabyte. That is a fair question, don't get me wrong. But if I were to advise a CFO, I would suggest to look at the c total cost of ownership of an SSD, because that covers all the costs related to the single drive over its life cycle. So it's not just the acquisition of the SSD, it's its maintenance and eventually maybe its replacement as well over the life cycle. So you look at the endurance of the drive using this those these values like the drive rides per day, the terabytes written, the meantime between failures, so you know what to what to expect. You pick the right drive, which may be more expensive, of course, but there is a reason behind it. In general, in technology, you get what you pay for. So if you pay a bit more for something, there is a reason behind it generally. Yes. So if you if you need to upgrade your your your organization's storage drives and you only focus on the cost, you may get a short-term win there, but eventually you may be forced in the future to forced and rushed upgrades because those drives are not built for that type of workload.

SPEAKER_02 19:23

You saying that, you know, this brings me back to my question 12, and this is quite pertinent in terms of I'll just read this out. What decisions today will business regret in five years? And this is what you've said if you if you buy something that's you know maybe 20 or 30 percent cheaper than the competition, and then and it fails in two years, and guess what, you can't even buy a replacement, then you know, have you actually saved any money? And I for me, I think it's all about and I've got Kingston drives in my systems, and I'm you know, and they they do run quite reliably, and I don't think I've ever had a failure to be honest, but that's that's not bigging you up, but it's a fact.

SPEAKER_01 19:59

It is a fact, it is a fact, they're reliable, and yes, probably probably because you've also made the right decision when you picked your drive. Decisions today that businesses will regret at some point are those driven by short-term thinking and short-term cost thinking in general. Yes, you pick the right drive with the right endurance, you look at your workload, you see, you you talk to an expert, for example, which can who can help you pick the right product. It'll be possibly more expensive, yes, but then what you want to avoid is downtime, because downtime is costly and it's reputationally risky as well. You want your uh your your organization or whatever you're doing to continue working, you don't want any interruptions there, you don't want to be forced to replace your drives because they weren't out much before then what are you thinking?

SPEAKER_02 20:46

You you have NVMe, you have obviously memory at the top, which is the fastest, then you have NVMe, then you have SSDs, and currently you have HDDs, and then you have potentially tape or the cloud at the lower level in terms of performance, the performance performance layers that you get. So, where from your point of view, where do you see Kingston sitting? Is it NVMe SSD? Is that where your primary or are you moving more towards NVMe?

SPEAKER_01 21:16

There's a shift towards NVMe with organizations, but that doesn't mean that we're not supporting SATA anymore because we're still covering those, and SATA basically is basically selling itself still because for what we're seeing is a shift towards NVMe-based drives for hot data. So for data that's being constantly used, and especially for workloads like AI, where you need to move those massive data sets fast into the memory. But for those data that you do not need every day and that can run a bit slower, you can rely on SATA more for uh um I don't want to say archiving because because that will be the next level, but for data that can that can be there because you you don't need it actively very, very, very, very often. So for whatever needs to run fast every day, you can rely on MVME because you need that kind of that that kind of speed and then SATA for uh for the rest.

SPEAKER_02 22:09

One of the one of the questions that you and I spoke about before the video started was where SSDs and NVMe sit from an archiving perspective, and what people don't understand and is that you do have to periodically power these maybe once every year or once every two years, or you know, you just can't take an SSD and stick it on a shelf because it will it's like a it's like a battery over time it use loses charge, doesn't it?

SPEAKER_01 22:35

So I like this comparison with the battery, yeah. It's true. I mean it relies the flash technology in general, not just Kingston, but flash technology relies on electrical power to retain your data. And depending on the type of NAND, depending on how new the drive how new the drive is of or how worn out the drive is, whenever you write data into it, there will be electrical circuits to to to retain that data. And then it's good to refresh those data from time to time to make sure that they remain there and the electrical power does not just wear away.

SPEAKER_02 23:09

So they're not an archive medium, they could be, but you've got to do that power cycle, otherwise they will degrade.

SPEAKER_01 23:16

They can potentially I mean flash technology wears out, that's normal. It's not infallible and it wears out depending on the type of nun, depending on the product, depending on how you use it, depending on the temperature, on the humidity, on how much data you've written into it, how much how often you read off it. There are things that you can do keep to keep your flash fresh. Yes. We've seen USBs last for 10 years. We've seen SSDs retain that they retain their data for a lot of years without being used. So there's a lot of factors that play a role in there. It's not just about how often you refresh your data. But yes, refreshing can definitely help maintain the flesh, the the cells.

SPEAKER_02 23:52

So there you go, you've you've learned about archiving there. What's the next technological breakthrough from Kingston's point of view?

SPEAKER_01 24:00

Oh, this is an interesting question because everybody expects like the the the brightest of the newest of the fastest components. You know, I don't see as a next technological breakthrough a new shiny component. I see more an architectural change in the way organizations choose and manage their their devices. So it's a full architectural change. With more intelligence, how you want to how you want to manage your storage and memory, the decisions that you make, there is certainly a shift towards PCE Gem 5 into mainstream, but again, only for hot data. But fan fact, there are organizations that are not even fully utilizing Gem4 yet. So when you make that decision and you want to go towards newer technologies beforehand, you want to make sure that your architecture can actually support that. Yeah. Because otherwise you're overinvesting in something that you cannot use. It's like having a car, a sports car that can only run on second gear. You don't want that. So you need to make sure that everything works. And then I see CXL Compute Express Express Link, which allows for memory pooling and uh dynamic allocation of memory. I see this coming from time to time. I'm not, I don't want to say that this is this is gonna be the next technological breakthrough, but there is, I think, something there with allocating the memory dynamically whenever when you need it instead of having memory allocated in one server immutable that you cannot allocate whenever you need it. So this is good.

SPEAKER_02 25:35

Do you do you see a market for as an interface, uh, a silicon photonics interface? And and I don't know where we are on the roadmap. Let's say you had a pool full of Kingston memory in a shoebox, and it had a silicon photonics connector, and so in a data centre today, a data center could be like the size of a football pitch. All those components in that data centre are all cooled pretty much to the same temperature, regardless if it's a CPU or if it's a tape drive, they're all in the data centre, they're all cooled at the same temperature. So, what silicon photonics is is it's a it's a it's a terabit connection using optical, and basically what they're saying is that you could buy a shoe box of Kingston Memory, and that could go in a broom cupboard, and the broom cupboard's got, and all they do is add another box, another shoe box, and I'm sure and they connect it up and they connect it up. So, what you would have is you would effectively have throughout a building a vertical data center rather than a horizontal data center, which obviously, as you know now, you can get a petabyte of storage in that much space, you know it's crazy. So people don't need as much space as what they perceive to be, and what they're saying is that these connections, these terabit connections, are like micro latency performance. The operating systems and Microsoft actually have this supported already, and a few Linux operating systems. Whereas if I came along with my shoebox full of Kingston memory, I plug it in, I've got terabit connections, and literally the operating system knows straight away that I've plugged in another petabyte of storage, bang, and it's already there and it knows what to do with it.

SPEAKER_01 27:16

With more intelligence that I was mentioning at the beginning, so we need to rethink the way we want to allocate memory, we want to use storage, it's it's going to change because now we're we're with motherboards and CPU and memory and storage, but possibly there's going to be a big shift on how this is going.

SPEAKER_02 27:34

The idea was, and it to me it makes perfect sense. So you buy a box of CPUs, shoe box of CPUs, you buy a shoebox full of memory, a shoebox full of Kingston SSDs, and they all connect together in one seamless thing. So rather than building a flat server, flat system there, everything's stacked, so you stack it up and you buy what you need, and that's what they're working on. Dell's working on this as well, actually.

SPEAKER_01 28:02

So you wouldn't rely on a PCB that electrically and mechanically binds all together, but you would rely on uh on fiber, yeah. And you basically pay with your components depending on what on what you need.

SPEAKER_02 28:14

So so you know, from a from a virtualization point of view, you might need some extra CPUs. You buy just a box of CPUs, you plug it in, and the iOS automatically knows and it uses those CPUs. And to me, it makes really smart sense, and you get very dense, it's very dense, but it's super cooled. You know, that cabinet there is CPU, so that needs to be the cooled the most, and then you know, so your cooling costs come down massively as well.

SPEAKER_01 28:40

I see something in common with CXL, yeah, because you know you're you're dynamically allocating whatever you need and you can move on.

SPEAKER_02 28:48

And it does it so fast in the blink of an eye. You know, the performance is phenomenal because it's all travelling at the speed of light. And I saw this originally, oh god, probably seven years ago. So seven years they were talking about this, and I and I spoke to someone a couple of years ago, and it's still being developed. So it's it's called silicon photonics, you know, and it's uh really literally you get the cable like USB, plug it in, bang, we've got some extra memory, or we've got some GPUs in a in a box, or whatever it is. So everything's shrunk down rather than building a whole circuit board of components. You buy what you effectively you need.

SPEAKER_01 29:24

That would be that that would be fantastic to see. And I really want to really. But I really want to see that in action at this point because uh yeah, we're so the technology. You think that technology is very fast in developing, and there's the latest of the newest and the greatest every single day. But then to see something actually being implemented, that is a much slower process. So I don't we always don't want to overhype technology, big technological breakthroughs. It is as I as I've mentioned before, people are not use utilising Gen 4 yet in their server infrastructure. So this is going to take a lot of time. You might have that greatest new breakthrough there, but then when are you actually going to see it in action?

SPEAKER_02 30:09

Exactly, exactly. So that's just one for you to to watch, but it is coming, apparently it's coming. But I think part of the problem is is is to do with certifications and getting everyone on board, you know, to to buy into this. I think Lenovo were doing it as well, I think they said, didn't they, when we met them? So they're all all these manufacturers are developing this in the background, but it might not come out for for 10 years.

SPEAKER_01 30:34

Probably exactly, yeah. We're seeing now even new new uh memory form factor, and this has taken and this is taking long for implementation.

SPEAKER_02 30:44

Well, you talk about the u.2. The u dot two for SSD. U.2's been around six years, yeah. And it and it's only now starting to become a little bit more mainstream. Exactly.

SPEAKER_01 30:53

A little bit more mainstream, exactly. Also from memory, there's new new form new form factors coming. Basically, you need to rethink the way a memory module is sort of is um plugged into a motherboard. So instead of going there vertically in the socket, it's applied horizontally, like on the PCB. Oh, okay. And it connects with like an EPROM. Like an EPROM yes. It connects over the motherboard horizontally to save some power, to man to make a thermal management, and it's screwed onto the motherboard. So you save space. This is for for ragged devices. Laptops are very compact, compact systems. But yeah.

SPEAKER_02 31:34

Why is storage suddenly back on in the boardroom on the agenda of 2026?

SPEAKER_01 31:39

Storage used to be an IT line item in the past, but now it's not that anymore. Now it's directly linked to competitiveness and risk. Because first you have AI model training that you might want to do in your organization, and that is massively data hungry. So you need to make the right decision how many drives you want to have, the capacity for each drive, the capacity that you actually need, and you need to be ready for these data sets to grow. Because this is a constantly growing type of work. You pick, for example, a hundred SSDs, 50 gigabyte each, or you pick 50 drives, 100 gigabyte each, there's a difference there because um the 100 drives will all use their own power, even when in idle. So if you don't use some of those drives, they will still be using power. Whereas if you pick 50 drives with higher capacities, you can there's there's differences there in terms of power of power consumption. Explosive data growth in organizations that want that are producing more and more data more than ever, and they want to keep them closer to them for sovereignty matters. These data need to be cyber resilient. The boards now want to know if we're hit tomorrow by a cyber attack. Are we protected? How are we gonna manage this? Are we gonna have some backups? Are our drives self-encrypting, for example? Are we how are we gonna how are we gonna get through this? They want to know this, and so you need to implement drives that can be protected and strategies that can protect your data, like offline backups, for example, air-gapped backups that cannot be accessed online, so you know that your data is there and nobody can access those through the network. These these kinds of strategies. And then again, there's energy energy efficiency as well. Yes. So back to what I've mentioned before about the number of drives that you pick. Each drive has their own fixed idle power consumption. So you need to think you need to think of that when uh when you want to populate your slots, not just populate all the slots because they look better. You need to populate the right slots with the right capacities, and same goes with memory.

SPEAKER_02 33:55

Yeah, Mateo, thank you very much for coming in. Appreciate it. You take care, thank you again. Thank you, thank you everybody for watching. Take care. Thanks.