On this episode of Six Five On The Road, hosts Dave Nicholson and Lisa Martin are joined by Kalray’s CEO, Eric Baissus, for a conversation on how Kalray, along with its collaboration with Solidigm, is revolutionizing AI performance and efficiency through high-density storage solutions.
Their discussion covers:
- Kalray’s strategic partnership with Dell and the value it brings to AI innovators
- The critical role of data acceleration in AI and the blend of hardware and software to achieve it
- Exciting announcements from Kalray
- The debate on “Walled garden” vs. “Open” approaches in supporting AI solutions
- Kalray’s collaboration with Solidigm and its impact on the Dell-Kalray offering
Learn more at Kalray and Solidigm.com/AI.
Watch the video below, and be sure to subscribe to our YouTube channel, so you never miss an episode.
Or listen to the audio here:
Disclaimer: The Six Five Webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we ask that you do not treat us as such.
TRANSCRIPT
Lisa Martin: Hey everyone, welcome back to Six Five On The Road from Vegas. Lisa Martin here with Dave Nicholson. We’re covering Dell Technologies World 2024, or as they’re calling it the AI Edition. We’re excited to be talking with Dell, its ecosystem of partners, on all the things that they’re doing to enable customers to not only be successful, but to really embrace AI for all of the amazing benefits it’s already delivering. We’ve got Eric Baissus here with us, the CEO of Kalray. Eric, great to have you. Thank you so much for joining us.
Eric Baissus: Thank you. Very pleased.
Lisa Martin: Share with the audience a little bit about Kalray and its evolution to a data management platform. From what I understand, started as a semiconductor company, then you built a DPU, and now you are creating a data management platform. Tell us about that and why the evolution, the pivot.
Eric Baissus: Sure. It’s not really a pivot actually, because we are still a semiconductor company. Okay. So, we designed a new type of processor, industry call this type of processor, DPU for data processor unit. And so we’ve been around for 15 years. So, we are a spin-off of a well-known European lab. And the vision of this when we created the company was that the modern societies generate more and more data, and you need to have new type of processor to be much more efficient to analyze this data. It happens you know very well GPUs, right? So GPUs are what we call massively parallel processor, which can run a lot of calculation at same time, but they do basically the same operation.
Okay? So, what we invented with a DPU is the same mechanism, but where each, what we call core, can run independently. So, why is it important? Because if you look at the data center, especially in the AI space, there are workloads where you need to be very efficient to run the same operation. GPU, AI training, AI inference, okay? But also you need to feed these GPUs, right? And to do that, you need to manage packets. And packets are asynchronous. So, GPUs are not very good at all to manipulate packets. So you need new type architecture, these are the DPUs, and that’s where we are very strong.
Now, to come back on your question, so we are a semiconductor company. We designed this DPU, and then we realize that it’s not so easy for our customer to adopt this DPU. So, how do we make sure that we are going to leverage this DPU? It has been by where we acquired another company three years ago to be the complete platform, a complete solution. So, this platform is called Ngenea. It’s what we call a data expression platform. And the goal of Ngenea is really to help large enterprise, having a lot of data, to process that data as fast and as efficiently possible. And where does our DPU sit is that, so we, of course, provide our DPUs through accession cards.
It’s optional. So, our customers either they have a pure software solution, okay, can run on the cloud, can run on-premise, but they can accelerate, so to get much more performance thanks to our DPU. And we see that actually some of our competitors, they are going in this direction too. So, starting with a software solution, but if you are pure software, it’s very difficult to make big difference. Okay. So that’s one of the key differentiation we bring. We’ve got this software accession platform that you can accelerate with our DPUs.
Dave Nicholson: Can you give an example of maybe walking through, think of it in terms of a customer journey, someone who maybe thinks that their environment is fine software only? So, you come in and you deploy the software environment, only to find that, “Hey, we’ve taken a look at this. You could benefit from having DPU acceleration.” What do those workloads look like? What kinds of customers are those, to make it more concrete?
Eric Baissus: Of course.
Dave Nicholson: I understand when you talk about packets and talk about parallelism versus doing a bunch of different things at the same time. But give us some concrete examples of industries that typically need these.
Eric Baissus: We’ll give you a very, very obvious example, which is AI today. So, how does that work? So you’ve got to form of GPUs, plenty of GPUs. These GPUs, of course, they need to run the calculation and so you need to feed these GPUs. So on our side, so we’ve got our engineer, and you can provide… And we provide unique performance with the software version of engineer. Why? Because we will use what we call GPU direct type of technology to directly make the GPU accessing to the data, to the NVMe data.
But then if you want to go further, it happens that we have a partnership with Dell, and so we build what we call the NG boxes, which are NVMe boxes for scale-out storage, very fast storage with their servers. And then these boxes, you can integrate our DPUs to make the access to this data much faster. And so, a DPU can run the encryption, can make the cluster resilient, so basically perform all the workloads that are super data intensive. Also very energy hungry, and so that’s a way to bring very concrete use case with all of our customers. A lot of them, they want to have a DPU version because it brings immediate value.
Dave Nicholson: Are you typically engaging customers shoulder to shoulder with Dell? Is there a lot of education that has to happen around the concept of the DPU? As you mentioned, there’s a lot of XPUs around, people of course know CPU, and now you have TPU and NPU.
Eric Baissus: It’s very simple. We usually don’t talk about DPU, for the exact reason that you said. People are not so much educated. There are a lot of different type of processor. What is a DPU? What is a XPU? So we talk about our platform, data expression platform, and then when we go into the details, how does it work? Then our customer they realize, “Okay, oh, I can take benefit of this new type of processor. Okay. I want it.” But they don’t really care, is it a DPU? You call it whatever you want. They want the performance, “We’ve got this product, without this product, with we can see the difference in term of performance. Hey, I want this one.”
Dave Nicholson: So nobody wakes up in the morning thinking, “I really want A DPU today.” But they might wake up thinking, “Acceleration would be nice.”
Lisa Martin: Yeah.
Eric Baissus: Exactly. So, I wouldn’t say not yet.
Dave Nicholson: Okay.
Eric Baissus: Okay. Not yet, of course now… 10 years ago, maybe nobody, or not so many people knew about GPU. Right? Now, everybody, even if you talk during your Christmas party, GPU maybe everybody around the table may know, “Oh yeah, GPU, Nvidia, of course I know that.” So now, but I think we’re still in a kind of open bar situation where, “Okay, I want the GPU, I don’t care how much it costs, we will see.”
We’re at the beginning of the hype. But what we see little by little that our customers, they realize that it’s super expensive. Once again, it’s super power hungry. So, now we see coming the V8 where we need to optimize, how do we optimize? And that’s where DPU are very interesting. And that expression platform software, how do you manage faster, more efficiently your data, is something which is getting hotter and hotter right now.
Lisa Martin: Talk about how you’re presenting this to customers in a way that is digestible to them, and it’s what they want to hear. I want to talk about complexity, or specifically how Kalray is addressing reducing that. We know that AI technologies require more data-
Eric Baissus: Correct.
Lisa Martin: … much, much faster. So than the world of data management is inherently getting more complex. How is Kalray addressing that for your customers?
Eric Baissus: So, maybe let’s start by talking about the pain points. What are the pain points of our customers? So, as you know, AI now is the best, so a lot of large enterprise or small enterprise will come to us and say, “Okay, I want now to run AI workloads.” Okay, so what are the pain points? One of the first pain point of course, as we said, is about leveraging their GPU. They’re very expensive, so you need to feed them in the most efficient way. So, that’s pain point rule number one. It happens when engineers has said to you, one of the key benefits is to boost dramatically the performance of your overall store environment, that’s why we call it a data acceleration platform.
The point is not to replace your existing infrastructure, but by adding our module to accelerate it. The first one is performance. The second pain point is exactly what I said, is about legacy. Much have our customers have invested for years in their storage, they’ve got different type of storage, can be a tier zero, tier one, can be on premise, it can be on the cloud. So, I don’t want to put in the trash of the investment. Some of our competitor actually they’ve got a very easy answer, “Hey guys, take what we sell, and you rebuild everything.” We don’t believe that’s a proper solution for our customer. Okay? So, the second pain point is how do we basically accelerate and transform their existing infrastructure into an infrastructure which is relevant for AI workloads? And I believe it’s pretty unique. Our value proposition here is pretty unique.
The third one actually, and I think we will see that more and more, it’s pretty new in the mind of our customer, is about enabling new AI applications. Because I want to do AI, but actually the reality is that except a few company designing new models, if you’re OpenAI, if you’re Google, yes, you need GPU to create your model. Most of the enterprises they will not create a new model. What they’re going to do? They’re going to use a model. Maybe you fine train a model. So, what you want is to leverage tool, like a chatbot, to be much more than chatbot.
It could be creating a new video asset, it can be just entering to your customer. So, it will be a complete new ecosystem of AI powered tool that will appear on the market. And now the key pain point is how do you connect these AI powered tools with your data, and still control it? So, that’s where the data expression platform not only ease and speed up the infrastructure, leverage the existing infrastructure, but enable the connection of AI powered new services to your assets, which are your data.
Dave Nicholson: So, I think it’s interesting, often when you’re trying to understand a new company or a new acquaintance, you go to who their friends are, and you have a lot of really cool friends in this industry. So, Dell and Solidigm recommended that we talk to you. And you take Dell, who’s in the business of, yes, efficiency, openness. Solidigm in the business of driving density, efficiency, power efficiency, and all of these things, pointed to you. Kalray’s not a ginormous company.
Eric Baissus: No, we’re not.
Dave Nicholson: So you must be doing something right. You’re here at Dell Tech World. Tell us more about the partnership with Dell. How did it start? How did that relationship start?
Eric Baissus: Well, I think it’s back to what I said, it’s about pain points, and they were facing the pain point of their customer, and it was exactly what I said. It was about performance. It was about a lot of complex requirements.
Dave Nicholson: So, very practical requirements.
Lisa Martin: Right.
Eric Baissus: I think Dell is a very practical company, right?
Dave Nicholson: Yes, we’ve noticed.
Eric Baissus: So they realized, “Okay, guys, we’ve got something which is very complementary to what you have.” And once again, this was also, I think, that’s the basis of a strong partnership. Of course, you need to be complementary to compete, and I think Kalray positioning is very complementary to Dell. Of course, Dell is a huge company having a huge, awesome portfolio in terms of storage and so on. So I think what they realize is that we bring the additional small pieces of the complete system that was missing. And it was once again, how do you maybe improve a little bit the performance when it’s about super intensive workloads, without changing your traditional type of strategy you have?
And then it’s about, okay, how do I make this data much easier to manipulate? And so, I think that’s the reason why we’ve got this very interesting partnership, and we’re very, very proud to be here today, and very, very happy with the relationship we’ve got with Dell. We’ve got many customers together, some of them of course now more and more in the AI space. It depends that one of them actually has been a Dell award, Duos Technology was doing an awesome use case analyzing train and for safety reasons. So, I think it’s super…
Lisa Martin: Is that Duos?
Eric Baissus: Yes, Duos, yeah.
Lisa Martin: I saw that, my dad worked in the rail industry. So, any rail car story and voice of the customer-
Eric Baissus: It’s amazing, right?
Lisa Martin: … it’s like, oh my gosh.
Eric Baissus: It’s amazing use case.
Lisa Martin: But what I saw there is that the number and metrics always jump out at me that what Kalray is helping Duos do is reduce rail car inspection times by more than 100X. That’s a big impact.
Eric Baissus: It’s a quite big number, right? It’s not one or two times, it’s more than 100.
Lisa Martin: 100X for AI image-based processing. That’s huge.
Eric Baissus: So, that’s what we can bring, right? By just manipulating the data in a different way, by parallelizing everything, by boosting the performance with your hardware, you can really bring a huge value in this market.
Lisa Martin: And that’s what customers are expecting.
Eric Baissus: Of course.
Lisa Martin: Customer, regardless of industry. We have this expectation that as consumers or in our business labs, we can do any transaction right now, we’re going to have connectivity, it’s going to be real time, it’s going to be personalized, and it’s going to be far more efficient than it was yesterday. Last question, Eric, for you from my perspective, is with all the announcements that Dell has made in just the last, I don’t know, 36 hours, with an AI edition of Dell Tech World, that five AI PCs, what they’re doing with servers and storage and more, what do you see from a futures perspective with Kalray Dell going forward, helping customers become AI ready, and really start to be able to leverage that as a powerhouse for their business?
Eric Baissus: Yeah, I think really it’s back to what I said previously. I think it’s this small piece which basically in a way connect the dots. There are a lot of initiatives in Dell, all very, very valuable. And now is maybe, I think, the small salt we add into the recipe is about making this access to the data easier, but easier doesn’t mean slower.
Lisa Martin: No.
Eric Baissus: Okay. So, how to keep being at the best performance, but by simplifying automatizing basically the whole data pipe.
Lisa Martin: Yeah. Nobody wants less data slower.
Eric Baissus: Nope.
Lisa Martin: Nope. Eric, it’s been such a pleasure having you on the program.
Eric Baissus: My pleasure.
Lisa Martin: Thank you for joining us, talking about what Kalray is doing, how you’re enabling customers to really move the needle in big ways, the partnership with Dell and Solidigm, and we can’t wait to just see what’s on the horizon for you guys. Thank you for sharing your insights.
Eric Baissus: Thank you.
Lisa Martin: Our pleasure. For our guests and for Dave Nicholson, I’m Lisa Martin. You’re watching Six Five On The Road from Vegas. This is Dell Tech World 2024. Dave and I are going to be right back with our next guest, so don’t go anywhere.
Author Information
Lisa Martin is a technology correspondent and former NASA scientist who has made a significant impact in the tech industry. After earning a masters in cell and molecular biology, she worked on high-profile NASA projects that flew in space before further exploring her artistic side as a tech storyteller. As a respected marketer and broadcaster, she's interviewed industry giants and thought leaders like Michael Dell, Pat Gelsinger, Suze Orman and Deepak Chopra, as she has a talent for making complex technical concepts accessible to both insiders and laypeople. With her unique blend of science, marketing, and broadcasting experience, Lisa provides insightful analysis on the latest tech trends and innovations. Today, she's a prominent figure in the tech media landscape, appearing on platforms like "The Watch List" and iHeartRadio, sharing her expertise and passion for science and technology with a wide audience.
David Nicholson is Chief Research Officer at The Futurum Group, a host and contributor for Six Five Media, and an Instructor and Success Coach at Wharton’s CTO and Digital Transformation academies, out of the University of Pennsylvania’s Wharton School of Business’s Arresty Institute for Executive Education.
David interprets the world of Information Technology from the perspective of a Chief Technology Officer mindset, answering the question, “How is the latest technology best leveraged in service of an organization’s mission?” This is the subject of much of his advisory work with clients, as well as his academic focus.
Prior to joining The Futurum Group, David held technical leadership positions at EMC, Oracle, and Dell. He is also the founder of DNA Consulting, providing actionable insights to a wide variety of clients seeking to better understand the intersection of technology and business.
