Cofounder, DataDirect Networks
Alex Bouzari, DDN | theCUBE + NYSE Wired: AI Factories - Data Centers of the Future
Alex Bouzari, DDN | theCUBE + NYSE Wired: AI Factories - Data Centers of the
Future
00:00 - Exploring AI: Insights and Innovations with DDN
02:51 - Enterprise AI: Adoption and Value Creation
05:15 - Monetization of AI in Various Industries
08:38 - AI Strategies in Enterprise Operations: Insights from Elon Musk
12:18 - Title: "Driving Innovation: NVIDIA's Influence and DDN's Collaborative Impact
16:12 - Towards Tomorrow: Merging HPC, AI, and Quantum with Agentic Evolution
About Alex Bouzari
Alex Bouzari, cofounder and CEO of DataDirect Networks (DDN), appeared on theCUBE in June 2025 and at Google Cloud Next in April 2026. In the June interview, Bouzari said DDN was powering a million GPUs across hyperscalers, neoclouds, and enterprises. He described the company as a "data engine that powers the AI economy" and stated that DDN helps enterprises deploy AI in a non-disruptive way while lowering cost per token. Bouzari argued that the AI economy has split into two groups: one where GPUs generate value and profit, and another where GPUs sit idle, with DDN being the differentiator. He also discussed agentic AI, noting that an agent sends 30 requests compared to a chatbot's one, requiring highly efficient compute and data infrastructure. At Google Cloud Next 2026, Bouzari appeared alongside Asad Khan, Google Cloud's Senior Director of Storage. Bouzari stated that DDN and Google together help customers optimize where data should be processed and delivered. Khan noted that the partnership has achieved over 95% TPU utilization. Bouzari said that organizations are seeing employee AI spends of $10,000 to $20,000 per month, and that agentic AI requires cost containment and economic value delivery. He described DDN as an enabler of AI that helps drive business and financial outcomes for enterprises, nations, and consumers.
Transcript (30 segments)
✨ AI-enhanced transcript with speaker attribution
J
John Furrier0:00
Palo Alto studio connecting Silicon Valley and Wall Street. I'm John Furrier here with Dave Vellante, my co-host.
D
Dave Vellante0:17
Hi, everybody. We're here at theCUBE's NYSE Wired studio overlooking the Options Exchange, the famous Buttonwood Podium. CUBE alum, Alex Bouzari is here. He's the CEO of DataDirect Networks, DDN. Alex, great to see you, my friend.
A
Alex Bouzari0:31
Great to see you, as always. Thanks so much for taking some time.
D
Dave Vellante0:33
Always wonderful. I heard you were coming in. I said, 'Oh, I'm going to stay.' Usually, I'd be on my way back, but you always have the best questions and the best insight on the industries.
A
Alex Bouzari0:41
Thank you, sir. Always a joy.
D
Dave Vellante0:42
So I appreciate that. So I went back to our interview at SuperCompute, which seems like eons ago, right? And at the time you said, well, last year, this is 2024, you were powering 500,000 GPUs. By November, you were powering a million GPUs, hyperscalers, neoclouds, enterprises across geographies. You guys have been established for quite some time. Give us the update. What's up with you?
A
Alex Bouzari1:09
So it's continuously increasing. I think, look, the main theme now, I think Jensen spoke to it, is NVIDIA is now an AI infrastructure company. So it's all about building infrastructure, and it's creating value out of that infrastructure. And I think what DDN does is that's exactly what the data layer does. I think the AI economy has now bifurcated into two universes. One where the GPUs are being utilized and are generating value and are generating profit and business and financial outcomes. And the other one where the GPUs are sitting idle, and the difference is really DDN. We're now powering sovereign initiatives in three continents in the US, in Europe, in Asia Pacific, in the Middle East. We're powering seven out of the 10 largest, biggest AI implementations out there, hundreds and hundreds of enterprise customers, and that's really what's going on. I think the value creation and monetization, making GPUs productive, making GPUs profitable is what it's all about. That's what we're razor-sharp focused on.
D
Dave Vellante2:09
And sovereign is not just overseas. It's actually companies here want sovereign infrastructure as well; do they not?
A
Alex Bouzari2:18
And it's not just sovereign in the context of US, US defense, US government. It's every state needs a sovereign implementation. We're working on some organizations here where it's being deployed at the level of the state. How do we deliver products and services and capabilities to the constituents of a state, and how do we help them leverage AI and create value out of AI? So sovereign is going to be absolutely massive, I think, for the US and globally, and it's something that will continue, I think, for a very long time to come.
D
Dave Vellante2:51
You've always made the point, you come back to value. I mean, I like that because that's what it's all about. So let's talk about where the value is. When the AI factory term first came out, it's a very catchy term. Okay, we're manufacturing intelligence in the form of tokens. Okay, but then you start to track, well, how does that turn into value? Obviously in the scientific community with your HPC heritage, there's clear value there for scientific discovery, but now it's starting to seep into businesses. What are you seeing governments and enterprises, where are they getting value out of all this intelligence?
A
Alex Bouzari3:24
Sure, sure. So the value really comes from monetizing AI, monetizing intelligence. How do you gain insight from AI? And the only way we can get insight from AI is data. You take AI, you apply it on top of data. From that data, you train models, and then you gain insight, and that insight creates value for enterprises, for nations and for consumers. So we're increasingly seeing enterprises benefiting from AI. Salesforce, for instance, has been able to increase the training capabilities by 70%. That means that people say SaaS is dead. SaaS is over. No, SaaS is transforming itself into leveraging the AI tools. But for the ROI to pencil out, you need a data plan. You need a data layer that is making that possible. That is the value that we're providing. 85 out of the Fortune 500 are now using DDN in their data centers in the cloud to derive business outcomes and financial outcomes. Likewise, in sovereign, it is really preserving and enhancing the cultural heritage of a nation, of a state, of a certain geography, as well as providing safety and security. So these are the areas where things are happening. It's really tokens, but tokens in the context of ROI and making sure that the ROI pencils out. That's what DDN delivers. We're the data engine which powers the AI economy. And by doing so, we deliver true value, financial value, business value.
D
Dave Vellante4:56
Interesting. You talk about your penetration into the enterprise. It's deeper than I realized. But for years you've had aspirations and designs with your architecture to get into the enterprise. It's almost as though the enterprise wasn't ready for what DDN is capable of. But now it's like there's a huge tailwind for you.
A
Alex Bouzari5:15
Totally, because I think enterprises are finally coming to the realization that AI is not just an OpEx improvement. It is value creation for the products and services. It's giving them the ability to develop better products, better services that generate more value, and therefore their top line increases, their bottom line increases. So it's really an alignment, I think, between earnings per share in the world of finance where we are today and token monetization. So I think the reason why our adoption in the Fortune 500 is skyrocketing is we're able to align the token economics and the ROI associated with that with the business and financial outcomes that CEOs, CIOs and CFOs are seeking across industries from financial services to life sciences, automotive, manufacturing. We're able to articulate it in a way that they go, uh-huh, I see how this is helping my company, and it's aligning token economics with earnings per share. I think that is really the light bulb.
D
Dave Vellante6:24
So let's click in on that. So the coordination activities that AI is able to absorb, I call it paper cuts. It's the mundane tasks that nobody wants to do. It's not the job that you want. Let AI do that. That's clear, and that's driving productivity. It seems we're not at the human judgment level yet. AI is not doing that judgment for humans, so we're safe there for a while. But how about actually developing and monetizing products from that intelligence, that tokens turning into products, are you seeing that at this point?
A
Alex Bouzari7:00
Absolutely. I mean, look, if you're in the life sciences sector, you're developing a new drug that you're bringing to market to save lives. Well, if you can bring that new drug to market more effectively, faster with a higher likelihood of FDA approval, you're delivering true value to your organization. I think that's an example. Autonomous driving, one of the things that is required for the adoption of autonomous driving is make these autonomous cars safer so that we can deploy them in cities and states all over the world. Well, with AI, you're able to link what is happening at the level of the car, which is the edge, to what is happening in the data center to the cloud. Connecting all that is delivering true financial value. So I think it really all comes to the realization that organizations have invested in AI infrastructure, that is the compute, and NVIDIA is by far the best, the best compute in the world. And now the monetization and the value creation comes through data. You connect compute and data together, you connect NVIDIA and DDN together and the value is instantaneous.
D
Dave Vellante8:05
So we're talking before about, you were mentioning the SaaS is not dead. It is evolving. It's changing. And so what our observation is we're bringing the deterministic world of SaaS together with the stochastic world, the non-deterministic world of LLMs. But you can't just take an LLM and stick it into a vector database and expect to get value. The value, to your point, comes from data. And the hard part is harmonizing that data and then learning from the interactions of the humans in that closed loop. So what are you seeing in your customer base in terms of that closed loop? Because that's really where the value is.
A
Alex Bouzari8:48
Sure. So I think what is really resonating with our customers, and that is resulting in just massive massive growth for DDN, is make it easy for me to deploy AI in my enterprise in a non-disruptive way. That's what the CIO is most worried about. Okay, I have an infrastructure. My job is to keep the lights on and to serve all of the people in my organization. Bring AI into my world in a non-disruptive way. And we're able to articulate that saying, hey, it's non-disruptive and value add. So the CIO who was viewed as, well, he or she keeps the lights on, now, all of a sudden, is becoming a strategic element within the organization. The CEO is now looking at the CIO as you are strategic. That's one part. The second part of it, the economics have to pencil out, and that's significantly lowering the cost per token. NVIDIA is talking about we're improving cost per token by a factory of 10, by a factory of 20. We're doing it day in, day out across industries and customers. You connect these two things together and that's how the rapid adoption of enterprise AI happens. I mean, when Elon talks about the addressable market of SpaceX being close to $30 trillion, that requires enterprise adoption of AI. And for that, you need the data layer to be enabling, and that's the job that DDN does.
D
Dave Vellante10:14
So Elon goes out and pays 60 billion for a cursor because he doesn't have a harness. And in order to compete with the frontier vendors models, you got to have a harness. You got to have a coding agent. And to train these models, you have to have compute. And yet, he's now selling his compute to Anthropic. You saw the Google deal the other day. Why do you think that is? Is it because he's tearing down and rebuilding? He has to have so much excess capacity. Is he the only guy on the planet who has excess capacity?
A
Alex Bouzari10:46
You're right on. I think what Elon did, basically built up significant capacity. And because he was able to deploy it faster than anybody else, he ended up with that significant capacity in a very short period of time. I mean, I know that. We were there alongside them building it, deploying it, running across 200,000 GPUs on Grok. So we were there. So he built it so fast that what he has outweighs what he's using. And then on the other side of it, everybody in the world needs more capacity. So the thought process is, well, why don't I monetize that? I mean, the biggest source of revenue today, I mean, look at the Google announcement, the Anthropic announcement with Elon. He's going to monetize tens of billions of dollars in SpaceX just through that. So the SpaceX story is not putting a million people on Mars. The monetization of SpaceX and the valuation of SpaceX is the Google deal and the Anthropic deal while everything else is being built up.
D
Dave Vellante11:50
Specifically, he's rethinking Grok, and then he'll have the capacity, the compute capacity to apply to Grok. In the meantime, he's monetized it. It's just yet another funding mechanism along with an IPO.
A
Alex Bouzari12:03
You're talking about a billion dollars a month with Anthropic, a billion dollars a month with Google. Two billion times 12 months, it's more than $20 billion in revenue creation, just like that.
D
Dave Vellante12:18
And those guys, they definitely need the compute. It's interesting. I mean, Google, tremendous respect for Google, what they've done with the TPUs, but to your earlier point, NVIDIA, to me, is so far ahead, and they're on an annual cadence. It's interesting when Google announced their latest generation of TPUs, they said it's not an ASIC. Okay, meaning it's more of a general purpose GPU, I guess, but NVIDIA has been there for a long time. And as you well know, they've got the best GPU. They also have, I would say, the best CPU in Vera, and they got the best networking, and they announced new storage architecture. And so that's important to you all. The STX is what they announced. Ever since we've been in the industry, there's been a pyramid, a hierarchy, and now it's just much more memory intensive. So your optimization has to take that into consideration, helping improve KV cache utilizations, but how does it change the way DDN thinks about storage?
A
Alex Bouzari13:24
So increasingly our focus and the value that we deliver is in the data engine, the software layer that makes all of these things run more efficiently. Again, you have idle GPUs, and that translates into non-profitable AI initiatives. And you have GPUs that are running at 99% efficiency, which is what DDN delivers day in, day out. So that is very important. You mentioned Google. So our data plan is embedded into Google Cloud. We have more than 50 enterprise customers using DDN software technology in Google, in the cloud and benefited from it in their AI initiatives. But what NVIDIA is doing, which I think is remarkable, is exactly to your point, it's connecting GPUs to compute to storage to now LPUs, the LPX. I mean, that acquisition of Grok by NVIDIA is now helping connect inference enablement, which is the LPX, to model training enablement, which is the GPUs. You connect VR, Vera Rubin, on the one side into LPX. Now you have a framework that is serving the needs of the AI pipelines, pre, post-model training, analytics all the way to inference. So they're putting the compute in there, the AI infrastructure compute. We're adding the data layer on top of it. They're accelerating from a compute standpoint, the AI infrastructure. We're accelerating from a data standpoint. You connect the two together and that's how your AI initiatives become incredibly valuable, profitable, enabling, again, across all organizations, all industries.
D
Dave Vellante15:01
Did you have to... I guess you didn't. You don't really have to rethink your architecture for accelerated computing because of your HPC heritage. It's perfectly suited for accelerated computing.
A
Alex Bouzari15:15
Absolutely, absolutely. Again, I think the reason why DDN and NVIDIA get along so well on the engineering side, I mean, our engineers talk to NVIDIA's engineers day in, day out across all aspects of NVIDIA development, is because they come from the same world. They come from the world of high-performance computing, accelerated computing, highest levels of efficiency. We come from the same world. It just so happens that to deliver value to enterprises, to sovereign, you need to marry these two pieces together. I mean, NVIDIA is using DDN, has been using DDN internally for almost a decade now. They could do anything they want. Somehow the value that we deliver and continue to deliver in all aspects of the compute AI infrastructure enablements aligns with what their desires are, accelerates the adoption of AI and results in better value creation for AI. That's what it's all about. Accelerate the adoption of AI through value creation in business and financial outcomes. It's not about the technology. Of course, you have to develop the best technology in the world, but you have to deliver business and financial outcomes. That's how enterprises adopt. That's how sovereign adopts. That's how consumer benefits from the tools, the super apps that are getting deployed on mobile devices.
D
Dave Vellante16:28
I was at Oak Ridge National Labs a couple months ago on a quantum fact-finding mission, speaking of value, and they were putting forth the scenario that our HPC infrastructure is going to be a hybrid with AI infrastructure and GPUs and ultimately with quantum. They were building out quantum capabilities. And the value that they were getting out of that is they could model such incredibly much more sophisticated capabilities with quantum, but certain things that AI couldn't do and that classical computing couldn't do. And as a result, they forecast this hybrid world, which is interesting. You've got some visibility into major labs. So how do you see that coming up?
A
Alex Bouzari17:17
So I think what's very interesting is up until now we've been living in a world where it's chatbots. A chatbot sends something out, it's one query, it is one call. Now we're pivoting into the world of agentic agent enablement. Again, that's what Jensen says. Vera Rubin was developed specifically for agentic use cases. Well, a chatbot sends one request. An agent sends 30 requests. That's 30 times the power, 30 times the compute, 30 times the data. So unless you have a framework and AI infrastructure where the compute is highly efficient and the data is highly efficient, well, your agents cannot function. And that's why in the world of agentic, the value that we're delivering at DDN is something so significant because those 30 calls that are going out there, well, you'll go out of business if you try to do it the old-fashioned way. So you have to bring in novel approaches, novel architectures developed specifically for this kind of scale in order for AI to shift into this new world of agentic enablement.
D
Dave Vellante18:23
I know you got other appointments. They're giving us the hook. Alex, he's a friend of mine. We're going to hang out for a while. My buddy. I really appreciate you coming in.
A
Alex Bouzari18:32
Thank you so much. Thank you. I appreciate it.
D
Dave Vellante18:33
I know you've been super busy and look forward to seeing you around the block. All right, Alex Bouzari. Thanks again. This is Dave Vellante for theCUBE Wired here at New York Stock Exchange. Thanks for watching.