Wes Cummins: We’ve procured all the transformers. They will all be on-site within the next few weeks and then it will just be the work connecting these. I don’t have to get into the details too much, but there’s been a certain connector component that actually has been the delay, not the transformers on connecting and energizing these, but we’ve already installed several of them and they should just continue daily, as we ramp this back up all indications from a performance perspective. The new transformers we procured are working extraordinarily well. And so we expect that to proceed fairly smoothly. But procuring transformers in this market is not easy. We were really happy with the team able to find the amount of transformers we found and the timeframe we found them.
Like I said, already shipped to site, it’s painful for us with that site being down just from an economic perspective and for our customer by the way. The faster we can get that back online the better for all of us.
Operator: Our next question comes from the line of Mike Grondahl with Northland Securities.
Mike Grondahl: Thanks. Wes, you said the contract with the hyperscaler, the 400 megawatts, was like 60 to 90 days from when you started. When roughly did those discussions start? Just trying to figure out that start date.
Wes Cummins: I’m trying to think on it, but it’s been going for maybe three or four weeks on the discussions. Just to make it clear, there’s no contract. There’s a letter of intent, which is kind of the standard process here.
Mike Grondahl: But I think you are saying from when you started three or four weeks ago, 60 to 90 days from that time, you might have a contract and financing in place?
Wes Cummins: Yes, I think that’s the right way to think about it.
Mike Grondahl: Okay. And then on the cloud services GPU side, how many GPUs did you own at the end of February and how many were generating revenue? And then what’s your kind of estimate for the same, how many you’ll own and it’ll be generating revenue at the end of May?
Wes Cummins: Yeah, so we owned, I believe, 5,000, 5,120 or 5,120. The H100 class GPUs, so there’s 4,000. I’m having trouble doing the math in my head to round it to the exact number. But, so round it to 4,000. There’s 4,000 in revenue generation now. And then there’s 2000 that are being brought up to that stage. And we should have more before the end of the quarter.
Mike Grondahl: Got it. So 4,000 as of today are generating revenue and another two to 4,000 by the end of May.
Wes Cummins: Yeah. That’s our goal.
Mike Grondahl: Okay. Okay. And roughly how much did the transformers cost that you needed to put in to Ellendale the new ones?
Wes Cummins: I have David here.
David Rench: 300,000 a piece.
Wes Cummins: Yeah, a piece.
Mike Grondahl: And how many total did you need?
Wes Cummins: We needed about 45 of those. There’s some of the ones that the other model that we had that are still working technically, but we’re replacing all of them.
Mike Grondahl: Got it. Okay. And last question from me. Do you guys have a rough, kind of committed CapEx number for the rest of calendar ‘24?
Wes Cummins: Well, we have seven more weeks of oh, calendar ‘24. I’m sorry. Let me come back to you on that, Mike. I don’t have that in front of me. We didn’t have it here for the call.
Mike Grondahl: Fair enough. Okay.
Wes Cummins: Mike, I did want to make a point on the GPU business. We’ve been adding people, we’ve been accelerating or working to accelerate, from receiving to revenue generating. But there’s one piece that I mentioned on the last call, and we’re much more focused on it now, which is, we started seeing demand from enterprise customers and large enterprise customers, which we’ve really been focused on. And so pushing because we can continue to deploy with the current customer base kind of as aggressively as we want to, but we’d really like to transition up to the enterprise customers. And we’re close. I think we have a few of those in process right now. But that’s one of the reasons that there’s some slowness in the deployment through the end of May, just because I like to diversify our customer base outside of just the startups, even though I love our customers there, but we would like to diversify the customer base. We’re working pretty hard on that.
Operator: Our next question comes from the line of Darren Aftahi with ROTH.
Darren Aftahi: Hey guys, thanks for taking my questions. Wes, if I could follow-up on that last comment you made. So I think in the prior call you guys guided to 10,000 as a bogey for GPU number exiting May. Can you sort of speak to that goal, if it’s still one, and then embedded in that your comment about slowing down to diversify away from more VC-backed clients is the achievement of getting that 10,000 less important, but more important to be diversified going into fiscal ‘25?
Wes Cummins: Yes. I think you hit it right on the second one. The types of customers diversifying away from the group of — they are all doing different things obviously, different products for the start-up. It’s diversified through those customers, but it’s all similar and they’re all start-ups and mostly VC-backed. I think it’s more important for us just to think about diversification in that business over the longer-term instead of kind of rushing to make a single date. Outside of that, I wanted to make a correction, $200,000 on the transformers not $300,000 on the prior question.
