Hello, this is John Furrier from theCUBE. We're here in theCUBE studios in New York, overlooking the New York Stock Exchange. And of course, we have theCUBE studios in Palo Alto, connecting Silicon Valley with Wall Street, where technology meets money. Again, we're here as part of the opening series 'AI Factory' as part of our Media Week. An ongoing series where we host leaders of AI factories and the future of data centers as distributed computing continues to expand. Large-scale systems are what's driving all of this. Daniel Roberts is here, Co-Founder and Co-CEO of IREN, which is performing wonderfully in the public markets right now. The company's stock is at its 52-week high this week. Congratulations, Daniel. Thank you for coming to theCUBE.
Thanks for having us, John. You're truly a great model for AI factory businesses. And you're also a great model for data center businesses because you have a lot of them, and a lot of power. You've done some Bitcoin applications, but now the AI factory and energy story has become clear; stock prices have risen sharply over the past eight months. Business is going well. Everyone is talking about more data center and energy stories, as they're making the front pages of the top news agencies. Billions are going to this state, and billions more to that, and 100 billion from Nvidia to OpenAI. I mean, this highlights the fact that demand for data centers is at an all-time high. You must love that.
Absolutely, it's a great time for this sector. Definitely. And I think it's two-sided. There's the demand side. We see demand continuing to surge, but also from the supply side, because what we're dealing with in AI factories is a completely new asset class. Traditional data centers in urban areas are not equipped for the level of power required, nor for rack density or infrastructure. Yes. And we see the institutional market preparing, although it hasn't fully opened up as we expected this year. We think maybe next year, but we're seeing tremendous growth among the hyperscalers, and the emergence of new clouds or 'AI as a service' and 'GPU as a service', where anyone can get the computing power needed for these training and inference applications. Those are the key applications. How has your AI factory changed? What was the moment you realized that Bitcoin is great, but Bitcoin miners are power-constrained? AI is also power-constrained. What was the moment you said, 'Okay, AI has arrived'? And when did the AI factory concept become clear to you? Can you walk me through that mindset? Because this is a huge business decision.
Yeah, look, it really is. But 7 years ago when my brother and I founded this company, it was based on the premise that the digitization of society was coming. We were moving to the cloud as humans and we saw all these accelerating digital adoption curves where things go from zero to one overnight. So, Bitcoin wasn't worth anything a few years ago. And today it's a two-trillion-dollar asset class. Two years ago, AI was confined to the halls of PhD researchers. Today, it's the next step that defines humanity. And I think the crux of the matter is that it's really hard to build data centers at scale. And yet, these end target markets are seeing vertical growth in terms of their appetite for compute, and that's where we come in. We have a huge amount of data center capacity.
Dan, you know, it's funny. We just moved into a new studio here, and opened a new center in New York, and obviously we have a point of presence in New York. We hear things in the halls here at the New York Stock Exchange. Things like maybe we should turn the data center into a tradable commodity. The power is there. Also. So, it seems like an asset class. I mean, there are already discussions about maybe if you bought power in advance now from the data center or bought GPU cycles now, I can resell them. This doesn't really mean... I mean, this is New York thinking, you're there with what looks like options. It's options pricing.
Yeah, it is because we see that option being exercised now. We're witnessing an almost insatiable appetite for GPU cloud capacity. We've grown from 1,900 servers at the end of June and we now have 23,000 servers running or being installed in the coming months, and we have the ability to triple that in a short time as well.
Talk about data center topology. How many data centers? Can you share a number? And what are you doing that others aren't? Because the old data center model was a bunch of real estate investment trusts, real estate players, and hosting providers, then the cloud comes, then they need capacity, then you start to see the actual design of data center systems — not just a building with power and racks of servers on it. It was more like: 'Wow, let's design this as a supercomputer.' That's now obvious, but it wasn't seven years ago. You saw that. What are you doing? Explain how your company works.
Yeah, I think the first thing we did was secure a large amount of power. We now have nearly three gigawatts of land and secured power, and we currently operate data centers of approximately 800 megawatts, and by operating these centers, we are continuously replacing Bitcoin mining ASICs with NVIDIA graphics cards to meet the needs of AI training and inference. And as I mentioned earlier, we're starting to see demand emerging from enterprises through our partners.
That's interesting. You did all the hard work in the beginning. Now everyone is spending a lot of money to do it. But you already own it.
It's been tough years, but yes, we're in a very good position now.
'Tailwinds' doesn't do it justice. What is the role of Dell's 'AI Factory'? Because one of the things we see, first, Jensen Huang coined the term 'AI Factory' two years ago, I believe, at the GTC conference. Then this year you said the KV cache is the operating system, and that was a really expressive signal. Networks are part of this system. So it's getting into the fine details there. Then Dell adopted that concept. Dell used to sell servers. They still do, but the 'AI Factory' now consists of a lot of servers with integration operations. And there are all kinds of other things they've done. What is the role of Nvidia and Dell Technologies in your deployments and your success?
Look, Dell's role has become much broader than just supplying servers these days. Yes, they've been an extremely valuable partner on the 'AI Factory' side. Their air-cooled XC 9680 has many practical advantages compared to other manufacturers. So, we've had great success using it. But as we move to liquid-cooled systems, we find that their view of the data center is exactly like ours. The data center is the compute unit, and it's about how the entire data center interconnects together. It's about the servers, the network, the cooling, the GPU, and how it all comes together in terms of reliability and flexibility. We're now in the phase of deploying the NVL 72 system. These are Nvidia's Grace Blackwell 300 chips in partnership with Dell right now.
Yes. Okay, we've interviewed Michael Dell. We've covered Dell news for a long time, and for two years now, EAE and John Roese have seen this well, and they talk about it the same way you do. It's like the factory, it's a system with a variety of software on scalable servers, plus networking and storage systems. What's interesting is their reliance on OCP standards, and I've noticed in your work how components are swapped seamlessly in your videos. It looks very easy. I know it's not entirely easy, but it's not a complicated matter. I mean, you don't need to demolish the building completely or make radical infrastructure changes. So, let's talk about the cost of transition, where I think AI factories excel here in their ability to integrate directly. I know you're already doing it. Tell us about the scope of that plan and the transition process to AI factories.
We've been building multi-purpose data centers designed for AI factories from the beginning, in terms of rack density. So, today we have 160 megawatts of data centers fully ready to run these AI factories. We are currently replacing mining platforms by installing Blackwell chips and some AMD chips as well.
And regarding the customer side. Okay, the factories are ready and running now. I think they strip away technical complexities. When you hear the word 'factory', you immediately think of production and value. What are the value points customers get from AI cloud service? What are the main benefits? Obviously there is very high demand for your services. Can you mention some examples of that?
Absolutely. What I like most about the term 'factories' is that every previous industrial revolution was known by its factories and the workers inside them, right? And here we are in the fourth industrial revolution, centered on augmenting human intelligence. And in these AI factories, GPUs are the workers. They're the compute clusters, right? They don't stop working. They don't get tired. They work continuously. And they can scale into the millions. And so we're scaling human intelligence. And the ability to do that is very exciting.
And on the performance side, you have to commit to a service level agreement (SLA). You have to ensure high and stable performance. How do these factories actually work? And what results have you achieved? Can you share some stories or examples?
Yes, and this is where our relative uniqueness lies. We have full vertical integration. We own the land. We own the substations to deliver power. We own the electrical interconnect grid. And we own the buildings down to the server level. And this is what allows us to offer a flexible and low-cost service to customers because we don't deal with intermediaries. We don't have external hosting fees to pay. We don't have to contact a hosting partner to commit to their service level agreement. We have field teams, people present in the data centers. Our customers have direct access to these people, which creates a truly reliable and seamless service.
And they get what they want. They get everything they need.
And ultimately, that's the most important thing.
Um, regarding the business model, one of the things I love about your logo is that it's green. Green means money. Stock prices are doing well. You're in a healthy position on the commercial side. Great momentum. So, congratulations. But the sustainability side is huge. Can you explain this part of your business model?
Look, 100% of all the power we consumed over the past seven years came from renewable energy sources. So, that has been a fundamental and defining part of our business. And when you look at the forecasts for the data center industry, where McKinsey expects an additional 100 gigawatts of data center demand over the next five years, you need to do it sustainably. And our vision is to go to low-cost renewable energy sources and convert that into GPU cloud.
