Welcome back to Google Cloud Next '26. We are streaming live right here in Las Vegas. I'm Alison Kosik alongside John Furrier, and we're about to get into AI infrastructure kind of is the new competitive advantage.
Yeah, and storage is the number one thing that everyone keeps talking about because storage, network, and compute are the three main ingredients in these large scale clusters, and they're being reconfigured differently. And every day the story's still the same, get everything closer to the chips, to the processing. And storage is moving closer. We got two great guests, Google and DDN, to unpack that.
All right, let me introduce the guests here. We've got Alex Bouzari, the CEO and cofounder of DDN, and Asad Khan, the Senior Director of Storage at Google Cloud. Welcome back to theCUBE to both of you.
Oh, you're welcome. Let's start with DDN, for those who don't know what DDN is in this picture of AI infrastructure.
Sure. So, it's very easy. So, DDN is a data engine that powers the AI economy. We are to data what NVIDIA and Google are to compute. You need both to do AI successfully. We help drive business outcome, financial outcomes for enterprises, for nations, for consumers. That's what we do. We're the enabler of AI, the AI revolution. Bring it home, make it happen, deliver value from it.
Yeah. So, I talk about Google Cloud's news here because one of the things that's jumping out at everyone is the Gemini story. Obviously, it's well-documented, nice and tight, very clean positioning, but the real story is the data cloud and the agentic platform, that's got to be powered by all the storage. So, how is that playing out from an infrastructure standpoint? Because you've got the new TPUs, got NVIDIA leaning in with you guys, so you got a lot of action from the infrastructure, a lot of power. What's going on with the data storage layer? What's actually happening this year?
Yeah. So, I think as you rightly pointed out, there's a lot going on on the Google side. And I think the one thing which truly differentiates is that it is not just about us building something for the customers, Google itself is using the same exact infrastructure to train Gemini and serve those capabilities to the customers. So, all the way from the bottom, if you look at the innovations we have at the TPU, the GPUs, the networking stack that we have, all the way to the Kubernetes clusters and the investments we have in the application framework, that is what differentiates. Now, when it comes to the storage, what you will find interesting is that that is becoming over time the differentiator. And the reason is that we ourselves saw 4X more growth on data in just the last three years because everyone is bringing more and more data because, yes, you can have the processing chips, but what will they process? They will process the data, and that is where the differentiation is.
The scale conversation, I talk to practitioners all the time on theCUBE, and the consistent message is, and these are my words, just paraphrasing, 'I want to run production workloads at scale. That's our North Star. Everything we do drives towards that.' How is that working out in terms of how they think about storage? Because the data growth is massive. And these systems are getting it done. What does that North Star mean for production workloads at scale for AI specifically? What's changed in the storage equation with AI workloads at scale?
I think it comes down to a very simple thing, how saturated your GPUs and TPUs are, because those are super expensive, hard to find, and that drives a lot of the TCO. And one of the complaints you will hear from a lot of the companies that we work with is like, 'I'm not getting the return on my investment because they are overspending when it comes to the infrastructure.' And when you go and double-click that thing, you realize that they are not fully saturating the most expensive part of the infrastructure, which is the GPUs and the TPUs. To give you the example, and I'm sure we'll get into the more details, but we were working with Harmonic and they were saying that, 'We are not getting the ROI.' And they started using the Managed Lustre and the saturation was 6X more, which is crazy. And now you can imagine how much of the money they were investing by not getting the ROI. So, how saturated is your GPU and TPU? It comes down to that. And it is about the network and the storage you are using.
Alex, weigh in here, because at GTC, we talked about this on theCUBE and at Supercomputing last year. It's kind of going faster than we see it happening. The data engine, you actually use that word, data engine multiple times, that's a positioning, but it's storage with software. What's different this year as the cloud is continuing to thunder away? There's demand. There's clearly demand.
Absolutely. Demand is skyrocketing. I think what's happening is finally the world is evolving from experimenting with AI, trying to figure out what to do, how to do it, how to drive value, to production scale. That is really the pivot that is taking place. Agentic AI I think is making a huge difference. We have customers now who are utilizing internally trillions of tokens per month. Well, that's fantastic. It's helping drive value, but the economics have to pencil out. And so you need a data engine that delivers a number of tokens per hour where the ROI works out. And I think that's what we're doing. That's what our partnership with Google is enabling. We're able to deliver a data engine that is powered by Google Cloud and it's delivering real value to enterprises, to banks, to hedge funds, to pharmaceuticals companies, manufacturing companies that are finally realizing the benefits of AI and are able to make money out of it, create value out of it because the monetization isn't there. If it's just an investment phase, who cares? That investment has to translate into business value, financial value. And I think that's what we're able to accomplish.
Talk about the relationship with Google Cloud because I think this is where we're seeing a lot of action because there's a lot of co-designing going on, a lot of partnerships, teamwork. Talk about the relationship with DDN and Google.
So, I think it's an optimization. Google came to us last year and they said, 'Well, we have a business opportunity. We have a problem to solve. We believe that the DDN data engine, the DDN data plane can be an enabler.' We deployed it together. We optimized it together. We started delivering value through that together to organizations, and that business has grown phenomenally. I think it's one of the fastest growing partnerships that Google has had, definitely in the world of data. So, it's resonating with customers because the customers are getting true business value out of it. It's not just, 'Oh, I'm trying new things.' They're using it in production. It's making banks more compliant. It's helping hedge funds generate better returns for their investors. Pharmaceuticals companies are able to bring new drugs to market faster, FDA approval. It's all of that. It's business enablement.
Asad, weigh in on your side. You guys are intentional with the full stack here. It's clear on this event. Talk about the partnership from how that.
Yeah, yeah. So, I think as Alex is saying that we saw the customer demand coming in. We wanted to make sure we have the best tech, and that is where the Lustre could differentiate. But as Alex is saying, it's not about just taking some IP or something. It is co-designing. It is ensuring we are using the right network, we are using the right VMs. There is a lot of discussion like, 'Oh, C4X is the new VM type coming out. How do we become the first one using that capability?' And that's why even if you look at the announcement, one of the announcements was that we are now delivering 10 terabytes per second, which is a lot of throughput, right?
Which is 10 times when it was just very recently.
Which is 80 terabits per second of throughput. That is four to 20X of any hyper cloud is offering because we did not just pick some open source software, we partnered with DDN who has been the primary company this project. So, the other thing which was really good partnership is that if you look at our pipeline for the customer, it is a very nice split where Google is bringing customer, but DDN is also bringing customer because these customers walk in, and they say, 'Oh, we are already familiar with them. We were working with them in on-prem, and now it's great that you have this partnership, and we would love to move to the cloud.'
Customers need the flexibility, right? They're doing things on-prem, but then being able to utilize these capabilities in the cloud, in a trusted cloud, such as the one Google is bringing in with the DDN data engine.
The other thing is the economics aspect of it. TPUs, TPUs are expensive. And so for these services to work out for Google to make money, for customers to be enabled, TPU utilization has to go up. Together, we've been able to achieve 95% plus TPU utilization, which is huge. It's amazing.
Yeah. First of all, great to lay that out. I love the performance numbers. Let's dig into that because the machines are only as good as the data feeding them. But you guys bring up a good point: if I'm checking on the weather forecast, I don't want to use the most expensive machinery possible. I wanted my hardest problems being worked by the best Pareto curve, by the best storage speed. So how is that looking? Is that going to be software driven? I'm seeing software-enabled scenarios like policy. How does.
Well, I'd say, look, software I think is the enabler because in order for AI to function, in order for AI to deliver value, the data is coming in from everywhere, which means you have to be able to do inference at the edge; autonomous driving, cameras, sensor data, robots. Then that needs to tie into data centers. Some customers have data centers. Then it needs to be enabled in the cloud. All of that needs to work together seamlessly, and so you need a data plane. You need a software layer that is not only driving efficiency, but is delivering the value. You have to be able to plug into the customer's IT infrastructures. Google and DDN together, what we've created is a capability that ties straight into the customer's infrastructure, so it's seamless. You have to be able to please the CIO, non-disruptive, and you have to be able to deliver value to the business owners and the CFO and the CEO. And together, we're able to achieve that better than anybody has done in the past.
So, you guys have the intelligence then to figure out where to put the data? That's a key thing.
