IREN FY25 Earnings Q&A with Daniel Roberts | AI & HPC Growth Strategy | IREN Stock News | IREN News

Thanks Ross. Good to see you guys.

Yeah, absolutely. So we're a 30 June year end. So, our FY25 revenue hit $501 million, which was up 168% year-on-year. But what was also exciting was Q4 as a quarter standalone was $187 million, which is indicative, I guess, of the late growth and the achievement of 50 exahash late in FY25. Again reporting net income, our EBITDA was about 10 times up to $278 million. Again, showing that profitable growth, which we think is very important. Operationally, as you alluded to, we expanded our contracted grid connected power to almost 3 gigawatt. We've got a multi-gigawatt pipeline behind that. We achieved 50 exahash. But importantly, sitting here today, we're doing run rate revenue of about a billion dollars per year on the Bitcoin mining business. And we announced procurement of 10,900 GPU servers from Nvidia, which should take our AI cloud services business up to around 200 to 250 million annualized revenue by December.

Yeah, I think July, August is always a tougher time of year for traditional Bitcoin miners. Because when you're racking these ASICs up in shipping containers and tents, the heat gets to them, right? And this has been the whole value proposition of Iron. We've never subscribed to that business model. We've built proper data centers to house these ASICs. So in summer when the temperatures spike, you know, our ventilation, our cooling works fine and we don't see the drop off that others see. So yeah, like it goes to plan in July, August and that's what allows us to operate GPUs for AI in exactly the same data centers.

So, Canada, we're obviously outside the US tariff regime and that's where we're importing the GPUs today. The US, look, it's still a work in process. There's semiconductor exemptions that have been finalized. The bottom line is we've got the capacity. We're seeing the demand. The supply is going to come online. There is going to be a market-based solution for this. And it's clear that the US administration wants to be a leader in AI. So we're not concerned about it. There might be a few bumps in the road around tariffs on, off. But at the end of the day, we've just got to make good commercial decisions and make sure that we're making good economics off the purchases that we make.

Absolutely. So we had a follow-up call with one of our largest institutional shareholders last night actually. They said, 'You guys have just established a reputation for creating and curating the most optionality that we've ever seen in listed markets.' And I think that's been an underpinning of Will and I's approach to this business when you're dealing with such volatility in terms of the macro and these huge tailwinds in terms of how it might play out. Preserving that optionality is key. So the ability to, I guess, straddle Bitcoin mining, AI colocation, AI cloud services, it gives you enormous optionality and the ability to pursue different markets based on where the return and the risk profile sit at any point in time. Today we see the biggest opportunity in AI cloud services when you can buy these chips on a three-year payback and that's not just the chip, that's the entire data center. When you compare that to colocation and some of the risks involved there, sitting here right now AI cloud looks super compelling and part of the reason why those returns exist is not everyone can do it. There are a lot of technical requirements and expertise that you need to stand up an AI cloud service and we've been building that capability for the last couple of years and it probably goes back further in terms of the multi-tenancy nature of our digital data center infrastructure. So that just gives us huge optionality to pursue the area of the market where we see the best risk return.

Yeah, it's interesting. This popped up a few times yesterday in the analyst follow-ups and this concept of utilization and I was a bit confused because the concept of owning these GPUs and not having them almost fully utilized is something that hasn't really crossed our minds but it's becoming apparent that not every cloud business operates this way and that there is a natural level of underutilization and it's relatively high. So, it hasn't been relevant for us. The vast majority of our current GPU fleet is contracted. In fact, I think the only ones that aren't are some internal nodes that we're using for R&D. We've got access to multi-year agreements. In reality, we stagger maturities. Everything from 1 month through to three years. And part of that's portfolio diversification for us. Part of it just meeting customer demands and their needs. And look, while reserve provides predictable revenue, having access to some of those or the ability to offer shorter term or even on-demand where the economics can be attractive, you know, is a good way to add to the profile of what we're able to offer the market.

Sure. So rack density for the listeners refers to a 6-ft cabinet in a data center and how much power goes into that 6ft high cabinet. So 80% of global data centers have rack density of 5 kW or less. What we're seeing is an increase in density as a result of all these GPU and HPC workloads. So we announced 200 kilowatt liquid cooled rack density for Horizon 1, our 50 megawatt DLC facility in Texas. But as it's playing out, actual rack densities that are able to be delivered today are less than that. So we've incorporated additional flexibility in that data center to accommodate less than 200 whilst also accommodating the ability to house 200 kW rack density in the future as those additional generations are built. Now PUE essentially this is a metric of how much ancillary power you use in a data center that isn't used for the IT load. So independent of the compute. So if you have a PUE of 1.5, then out of every 1.5 units of energy that you consume, 0.5 is for ancillary such as cooling, comms etc. The one is the compute, so the actual usable load. So in British Columbia, we're operating at 1.1 PUE or lower, which is obviously super competitive. And then our Horizon 1 DLC facility is looking like an average of around 1.2. So super competitive. And this allows us to keep costs down for customers and utilize more of, I guess, that precious power from the grid for proper IT compute load rather than ancillary cooling infrastructure.

Yeah, look, Anthony's a fantastic hire. Most recently he was group treasurer at one of Australia's largest banks and in the investment banking side there. So, we've known him for a number of years and we're very pleased to bring him on board. And even though he's only been here what feels like a few weeks, he's already had a very large impact. We've seen with some of those hardware financing deals that he's secured where he's managed to secure 100% lease financing against batches of GPUs at single-digit interest rates. So they're not deals that we're seeing widely spread in the market. And it's a testament to the value that he's already adding. And I guess when you look at a business like ours that's growing so quickly, value creation for shareholders is about two things. One is the end market and the opportunity to deploy capital to earn the returns that we are in things like AI cloud, but how you fund it and how you finance it is directly related to the value that you could deliver for shareholders. There's no point having all these fantastic investment opportunities if you're not able to finance it on an accretive basis. And with Anthony joining us as chief capital officer, we feel like we're dedicating some really good resources and momentum into that area of the business. And the first two deals over the last week are great, but like everything we've done, we're very much focused on the next ones.

Absolutely. I mean, you guys have known us for a few years now. We don't tend to do small and think small. And it's funny, we join these podcasts and we have our earnings calls and we're talking a lot about what we've achieved to date and what we're doing today. But 90% of our headspace, once I jump off this call, I'm not talking or thinking about any of the stuff that we've just discussed. It's all about what's next. How do we grow beyond Horizon 1? How do we fill that up with GPUs, whether it's our own, whether it's a third party through colocation? How do we finance different structures? So, we've got a team set up that's very much focused on that. It's great to look in the rearview mirror sometimes and celebrate the quarter and the year that we've had. But it's not our nature to sit still. So, onwards and upwards.

Sure. And this is part of the strategy. We've built out 800 megawatts of high performance compute data centers. We've bootstrapped them with Bitcoin as we said we would, and we're layering in higher and better use cases over time as they emerge in the market. So the playbook from seven years ago when we launched our seed fundraising round is just being played out. We don't know what the future holds but we do believe in a digital future. We do believe humanity is going online and it's going to drive insatiable appetite for compute. But we can't prescribe what that's going to look like. We just want to be positioned to monetize it. So having 800 megawatts of data centers operating today, let alone the additional 2.1 GW of powered land we haven't built out into provides enormous runway for both near-term immediate growth as well as kind of staggered growth over the coming few years. And I think when you look at competitive advantage, vertical integration, super important. I know we've harped on about this for many, many years and it becomes just a bit of a tagline without much substance, but the issue is you've got full control when you own the dirt, the transmission line connection, the substations, the data centers, the layers of the compute. Particularly in a founder business where Will and I, we are super involved, we are thinking long term, the ability to drive good decisions in the moment rather than having a misalignment of incentives where we're focused on short-term, whether that's to optimize different objectives rather than I guess long-term shareholder value, it does mean that we're able to make good decisions. But importantly, as you alluded to, we don't pay colocation fees, so we've immediately got an advantage over other AI cloud service providers because of that lack of fee. But the vertical integration also goes further. Unlike a Neocloud or an AI cloud, we don't sit in the middle of a customer and a data center when there's an issue, and issues always pop up. That's life. We are on site. We're immediately out to rectify. And in fact, we're often quite proactive in terms of telling the client there's an issue before they've even spotted it. And you compare that to a traditional Neocloud model where all they can do is be beholden to SLAs with their data center contract because they're just a middleman. And if there's an issue with the GPUs, they just lodge a ticket and it's like, okay, well, we'll come back to you between 9 and 5 on a business day, provide 3 days notice, and then we can get in there. So, you can see naturally how the service provision becomes a lot more compelling. And then on a staff side, I mean, people see this stuff and your sales guys sitting inside some of these other companies, they're getting paid bonuses even though they can't sell GPUs just as a retention mechanism. But eventually, human nature, people want to do stuff. And then they see us with this really short-term ability to scale and the fact that we are executing on it now and saying, 'Wow, I've got customer demand. I really want to be selling GPUs. I can't where I am now because we don't have access to data center capacity or GPUs.' So the ability to, I guess, attract and retain good people is starting to feed on itself because of all those factors.

Absolutely. We don't intend for the facility to sit there empty and we could have signed contracts already just like we could have signed contracts for Sweetwater and we've been very transparent over the last 12, 15 months around that and this goes for the optionality we're trying to create, understanding the value, understanding the opportunity cost and that's not to suggest that we're going to try and over-optimize to the nth degree like you do need to get on with it, make decisions, which I think we've also demonstrated an ability to do so. Look, we still have time and time is our friend. We've got months until that facility is commissioned. As you say, we've now created additional optionality through being able to provide that through a cloud service. So working through conversations with a very large range of prospective partners across both cloud and colocation over the coming month or so is exciting and I don't know where it's going to land. All I can do is to look at what the options are and then analyze it through a risk-return lens to make the right decision.