Connection Silicon Valley and Wall Street. I'm John Fost here with Dave Volante, my co-host.
Welcome to the Raise Summit coverage brought to you by Solidine with support from Tensidine, Datrix, and Argentum.
Welcome back everyone to the Cube live stream here in Paris, France for the Ray Summit. My name is John Furry, your host of the Cube. We have the straight talking veteran entrepreneur here, Alex Bouzari, cube alumni, CEO and co-founder of DDN. Really, the giant of AI data, looking at the large scale. Alex, great to see you.
Great to be here. Thank you so much for having me. It's always a pleasure.
As always. You're dressed to the nines. We saw each other. You came to our day zero curated private dinner. All the players were there at Alan Samasi's house under the Eiffel Tower and the conversations were great. I mean, are we, is there compute available? Is China beating us? It was a wide ranging Chatham House rules intimate leaders getting together. I loved your perspectives because you see a lot because you're in the middle of all the action.
Absolutely. We're at the center of all of it.
You're in the center of the value proposition, but it's not storage anymore. It's memory. Hard disc drives had PCs and servers. Now you have storage as a fabric around the element of chips that make up the computing apparatus of this new architecture of AI factories, large scale, whatever you want to call it.
The new architecture of providing intellectual capabilities tokens is about data. Yep.
Data moving fast into the processing units or systems producing results driving value which is money.
Not hey how much throughput and IOPS do you have? Remember those days? We have shifted, we have crossed the threshold.
What is your current view now raise here? I mean the conversation ranged from time to money, sovereignty, everything in between.
Well, I mean look absolutely. Rays is very interesting because one of the trends as you pointed out is it's all about data. The value creation from AI is you need to lay out the infrastructure. You need the land and power and the data center builds. You need to populate them with GPUs, TPUs, LPUs, and what have you. But at the end of the day, the monetization, the value creation comes from data. You're taking data, collecting that data, training your models, adding analytics on top of it, and then it's the inference, the insights that are gained from that data which helps organizations, enterprises, and nations deliver value. It's the outcomes being delivered. The world on the one hand has shifted from model training into inference and agentic enablement. DDN delivers a 100x better monetization on tokens, whether it's tokens per watt or tokens per unit of time. At this conference, Raise in Paris, one of the key elements is sovereignty. We live in a world now where every country needs to put in place AI infrastructures, AI factories that are sovereign in nature. It has to be air gapped, controlled, safe and secure. At the conference, we're seeing all nations want to do that. They want to leverage available technologies but in the context of their own nation. We're going to see significant growth in years to come. DDN is involved in a dozen sovereign projects right now. The consistent message is make it easy to implement, make it truly sovereign, give us the independence where data stays in our country and does not trickle out. So that's one of the themes, and then obviously the monetization.
The data in the country is great because you brought up staying in country. That's certainly important, it's an infrastructure element. But product market fit is transitioning to product economic fit. The economies of these countries will be dependent upon that data staying there. By the way, the applications we're seeing emerge, whether it's Replet or Lovable or whatever, they're going to be running in the country. So if the e-commerce of AI is happening in say France, you keep the money there. We were talking to a large enterprise customer based in France with globally distributed stores. They said we want data in the Asia-Pacific region managed and controlled in APAC, and the ones in Europe managed and controlled in Europe. We want to put walls around these things to protect data per country, per region. That's a trend we'll see increasingly. The value creation for enterprises is business outcome, financial outcome driven by lowering cost of token, lowering power footprint, combined with sovereignty. You've been doing extreme storage and data for over 20 years, self-funded, well documented. You're a zillionaire. But you got BlackRock, put some money in, make it public. We'll see. The question is more about this AI gold rush. What are people getting wrong and right about how to view the storage layer as its role changed?
Yes. So from our vantage point, the world is bifurcating into those enterprises and nations where GPUs are fully utilized at highest efficiency and those where GPUs sit idle. The difference in AI delivering value so that CEOs say let's double down is utilization of the GPUs and infrastructure. DDN provides significantly better efficiency and cost at very high levels of scale and security to these enterprises, nations, and consumers. The ones getting it right are banking on companies making AI easy to integrate and cost-effective in delivering business and financial outcomes. The ones not getting it right are cobbling solutions together, wasting capital. Data is at the center. Data optimization needs to be enabled, and DDN is the best partner to make that happen.
You guys are doing hundreds of thousands of GPUs for XAI. I think that's public information. When you're at that scale, what breaks?
Yeah.
At any level of scale you have additional challenges that creep in. When you start to scale, enterprises deploy AI and are very successful, so requirements scale. Things break: networking, number of GPUs in racks, footprint in the data center, software layers. The most important thing is a software stack that delivers optimized value on compute (Nvidia) and data (DDN). If you get this right, GPUs hum along and you meet SLAs, token per watt requirements, and ROI pencils out. We've done it for enterprises like Salesforce, increasing GPU productivity by 70%. Now it's very profitable. Scale across sovereign, pharmaceutical, autonomous driving, and so on.
But other rivals say they do that too. Is that true?
At scale I believe we are the best suited. Nvidia has used DDN internally for eight years. Jensen says without DDN, Nvidia supercomputers would not be possible. That's a black and white statement. Yes, there are lots of claims in the market.
He does put a lot of those logo slots, storage logos on those slides.
He puts lots of logos, but what does he use internally? Nvidia has to be Switzerland to foster AI industry growth. But what does Nvidia use internally? That tells your story. They could use anything.
Exactly. What does a dentist use to brush his teeth? Of course.
Exactly. But they keep coming back because we're optimized. It's the optimization.
