APLD - Applied Digital Corporation (APLD) Q2 2024 Earnings Call Transcript
2024-01-16 12:24:09 ET
Applied Digital Corporation (APLD)
Q2 2024 Earnings Call Transcript
January 16, 2024, 09:00 AM ET
Company Participants
Alex Kovtun - IR, Gateway Group
Wes Cummins - Chairman and CEO
David Rench - CFO
Conference Call Participants
Lucas Pipes - B. Riley Securities
George Sutton - Craig-Hallum Capital Group
Darren Aftahi - ROTH MKM
Rob Brown - Lake Street Capital Markets
John Todaro - Needham & Company
Mike Grondahl - Northland Securities
Kevin Dede - H.C. Wainwright
Presentation
Operator
Good morning, and welcome to Applied Digital Fiscal Second Quarter 2024 Conference Call. My name is Sheri and I will be your operator today. Before this call, Applied Digital issued a financial result for the fiscal second quarter ended November 30, 2023, in a press release, a copy of which will be furnished in a report on a Form 8-K filed with the SEC and will be available in the Investor Relations section of the company's website. Joining us on today's call are Applied Digital's Chairman and CEO, Wes Cummins, and CFO, David Rench. Following their remarks, we will open the call for questions.
Before we begin, Alex Kovtun from Gateway Group will make a brief introductory statement. Mr. Kovtun, please proceed.
Alex Kovtun
Thank you, operator. Good morning, everyone, and welcome to Applied Digital's fiscal second quarter 2024 conference call.
Before management begins formal remarks, we would like to remind everyone that some statements we're making today may be considered forward-looking statements under securities laws and involve a number of risks and uncertainties. As a result, we caution you that there are number of factors, many of which are beyond our control, which could cause actual results and events to differ materially from those described in the forward-looking statements. For more detailed risks, uncertainties, and assumptions relating to our forward-looking statements, please see the disclosures in our earnings release and public filings made with the Securities and Exchange Commission. We disclaim any obligation or any undertaking to update forward-looking statements to reflect circumstances or events that occur after the date the forward-looking statements are made, except as required by law. We will also discuss non-GAAP financial metrics and encourage you to read our disclosures and the reconciliation tables, the applicable GAAP measures in our earnings release carefully as you consider these metrics. We refer you to our filings with the Securities and Exchange Commission for detailed disclosures and descriptions of our business, as well as uncertainties, and other variable circumstances, including but not limited to, risks and uncertainties identified under the caption Risk Factors in our quarterly report on Form 10-Q.
You may get Applied Digital's Securities and Exchange Commission filings for free by visiting the SEC website at www.sec.gov. I would also like to remind everyone that this call is being recorded and will be made available for replay via a link available in the Investor Relations section of Applied Digital's website.
Now, I will turn the call over to Applied Digital's Chairman and CEO, Wes Cummins. Wes?
Wes Cummins
Thanks, Alex, and good morning, everyone. Thank you for joining our fiscal second quarter 2024 conference call. I want to start by thanking our employees for their ongoing hard work and service in supporting our mission of providing digital infrastructure solutions to the rapidly growing high-performance computing industry. Before turning the call over to our CFO, David Rench, for a detailed review of our financial results, I'd like to discuss some recent developments across our business.
Let's start with our data center hosting operations. Our 100-megawatt Jamestown facility continues to perform as expected and operated at full capacity with consistent uptime throughout the quarter. This marks the fifth consecutive quarter in which the Jamestown facility has operated at full capacity. Our 180-megawatt Ellendale facility in North Dakota also operated at full capacity with consistent uptime during the quarter, bringing our total hosting capacity to 280 megawatts across our North Dakota facilities. Both facilities are contracted out to customers on multiyear terms.
During the quarter, we announced the initial energization of our 200-megawatt Garden City facility in Texas. This is a significant milestone in Applied Digital's ongoing efforts to meet the growing demand for low cost scalable digital infrastructure. The Garden City facility had a small contribution to our results this quarter and is currently operating at approximately 132 megawatts, with the remainder of the capacity expected to come online in the next several months. As we brought on the facility, we realized there were additional infrastructure improvements needed for the grid. We expect these improvements to be made no later than April. Our customers continue to send miners to the facility and we are actively installing them. With the increase in the cost of bitcoin, we are seeing demand increase significantly for hosting services. As a reminder, our Garden City facility is fully contracted with fixed prices, so we are not exposed to volatility in the crypto markets heading into the halving event this year. Once our Garden City facility becomes fully energized, we will have approximately 500 megawatts of hosting capacity across our three data center hosting facilities. We expect our three sites to deliver up to $300 million in revenue and $100 million of adjusted EBITDA on an annualized basis. Operating cash flow from data center hosting services will ramp up significantly in March as the majority of our prepayments burn off in February.
Let's move on to cloud services, which provide high-performance computing power for primarily AI applications. It continues to grow quickly as we progress further in supporting our existing contracts and pursue additional opportunities in our pipeline. Since our last earnings announcement, we have added an additional cloud customer, which brings our total annual contract value of cloud service contracts at full capacity to approximately $398 million. We tailor our agreements to our customers so that they -- as they raise money, we can exercise options embedded in the contract to deploy GPUs and ramp host -- ramp-up hosting capacity over time. While the typical customers for our cloud service have been private VC-backed companies, we are now also seeing strong demand from the enterprise market for large amounts of GPU compute capacity. We are excited to see demand increasing from this important segment of the market and have plans to hire sales talent to enhance our outreach efforts.
We continue to secure access to GPUs, however, there have been some delays in installations attributable to pending deliveries of networking components. We believe it's prudent to receive GPU deliveries only when all associated equipment is on-site and ready for installation, which is how we structure our client deposits. Additionally, we continue to actively explore vendor financing and other tailored financing options to support the capital requirements for the 34,000 H100 GPUs we have on order to support our current customer demand. To date, we have four 1,024 clusters installed and are planning to ship an additional four in the next two weeks. These clusters, as they are currently configured put us in an elite class of next-generation supercomputers in terms of raw compute power or petaflops for the most demanding AI applications. We expect to reach a minimum of 10 before the end of the fiscal year with the Jamestown cluster representing the opportunity to put us in the top 10 supercomputers for AI workloads. The fully commissioned clusters are expected to generate over $200 million of annualized revenue.
Lastly, let me provide an update on our purpose-built HPC data centers. During the quarter, we broke ground on our first 100-megawatt high-performance compute facility in Ellendale, North Dakota. This facility will offer low cost, high-efficiency liquid-cooled infrastructure designed for HPC applications. Construction is proceeding as expected. Our unique proprietary architecture and design implementation together with the strategic placement of the Ellendale facility near sources of abundant and renewable power will offer scalable infrastructure for these workloads. It will offer significant cost reduction to our customers and deliver best-in-class performance that maximizes high power density compute. We believe that this advantage is sustainable in this emerging market for data centers specialized in running AI workloads. Our contracted power and adjoining land at our facilities will become valuable assets over the next 18 months. We believe there will be a significant supply constraint for power in the data center market. We have already seen a robust demand for our data centers, which driven by the burgeoning AI landscape has exceeded our initial expectations. We believe we'll be in a strong competitive position to support this demand. As a reminder, we have 400 megawatts of capacity in development across North Dakota and Utah. This does not include the current 9 megawatts of capacity we have at our standalone facility in Jamestown to support cloud service customers.
As we enter the second half of fiscal 2024, we're well-positioned to capitalize on the demand we're seeing across both our cloud service and HPC data center business and we will continue to allocate our capital appropriately to the highest risk-adjusted returns to maximize shareholder value.
With that, I'll now turn the call over to our CFO, David Rench, to walk you through our financials and provide an update on guidance. David?
David Rench
Thanks, Wes, and good morning, everyone. Revenues for the fiscal second quarter of 2024 were $42.2 million compared to $12.3 million for the fiscal second quarter of 2023. The increase was driven primarily by the full quarter of revenue generation from the Ellendale facility, the Garden City facility beginning revenue generation during the fiscal second quarter of fiscal year 2024, and additional revenue from the Jamestown facility due to increased uptime. In addition, the company recognized a full quarter of revenue from the first cloud service contract during the fiscal second quarter of 2024.
Cost of revenues for the fiscal second quarter of 2024 was $29.2 million compared to $11.8 million for the fiscal second quarter of 2023. The increase in cost of revenues was attributable to higher energy costs used to generate hosting revenues, depreciation and amortization expense, and additional personnel expenses, driven by the growth of the business as more facilities were energized.
Selling, general, and administrative expenses for the fiscal second quarter 2024 were $21.1 million compared to $27.2 million in the prior year comparable period. The decrease was primarily due to lower stock-based compensation expense and was partially offset by increases in depreciation, amortization, and personnel costs. Net loss for the fiscal second quarter of 2024 was $10.5 million or a loss of $0.10 per basic and diluted share based on a weighted average share count during the quarter of approximately 109.7 million. This compares to a net loss of $26.8 million or a loss of $0.28 per basic and diluted share in the fiscal second quarter of 2023 based on a weighted average share count during the quarter of approximately 93.4 million.
Adjusted net loss, a non-GAAP measure for the fiscal second quarter of 2024, was $5.2 million or adjusted net loss per basic and diluted share of $0.05 based on a weighted average share count during the quarter of approximately 109.7 million. This compares to an adjusted net loss of $3.8 million or $0.04 per basic and diluted share for the fiscal second quarter of 2023 based on a weighted average share count of approximately 93.4 million during the quarter.
A significant headwind we faced during the fiscal second quarter of 2024 was amortization and occupancy charges for leases of computing equipment and data center space that have been assessed -- accessed by the company but are not yet supporting revenue. The lease expense for the colocation sites not supporting revenue totaled $1.5 million and were not added back to -- into adjusted EBITDA or adjusted earnings. Amortization of GPUs not supporting revenue was $3.7 million and was not added back to adjusted earnings. We expect this impact to decrease in future quarters as we resolve supply chain delays and are able to stand up full computing clusters that support revenue. Adjusted EBITDA, a non-GAAP measure for the fiscal second quarter of 2024, was $10.6 million compared to an adjusted EBITDA loss for the fiscal second quarter of 2023 of $2.2 million.
Lastly, on our balance sheet, we ended the fiscal second quarter with $34.6 million in cash equivalents -- cash, cash equivalents, and restricted cash, and $42.8 million in debt. During the first two quarters of 2024, we received $81.8 million in customer payments due to the structure of our commercial arrangements with our customers that incorporate upfront deposits and prepayments. In certain contracts, the prepayments are credited back to the customers over the term of the contract. This has no impact on revenue recognition, but the upfront cash flow is a major benefit for the company as it helps with our CapEx funding as we build out our data centers. Since the quarter closed, we have received an additional $11.1 million in customer prepayments and $23.1 in net proceeds from the ATM offering. The ATM offering is now complete.
Now, turning to guidance, due to the delayed delivery of certain networking components for our GPU clusters, we now expect our revenue and EBITDA to be below the low end of our previously guided range for the fiscal year 2024. Network component deliveries improved in recent weeks, but did have a significant impact on the timing of commissioning clusters and our revenue and EBITDA. We now expect to exit the fiscal year 2024 at an annual revenue run rate of approximately $500 million and an annualized adjusted EBITDA run rate of $250 million.
Now, I turn the call over to Wes for closing remarks.
Wes Cummins
Thank you, David. We're well-positioned to capitalize on the growing opportunities across our business and look forward to continuing our momentum in the second half of the year. I'd like to thank all of our team members for their dedication in making Applied what it is today and our shareholders for your continued trust on our mission and execution.
We are now happy to take questions. Operator?
Question-and-Answer Session
Operator
Thank you. [Operator Instructions] Our first question is from Lucas Pipes with B. Riley Securities. Please proceed.
Lucas Pipes
Thank you very much, operator. Good morning, everyone. My first question is on the HPC hosting side and the conditional agreement that you announced and a few ones. First, in terms of the total value you cite there, should we kind of think of $220 million of revenue per year? And then are your margin expectations for the segment still around 40% and on the capital cost side? I've been working with the $5 million per megawatt assumption. I wondered if that's still a good number to use. Thank you very much for your details.
Wes Cummins
Good morning, Lucas. Thank you. So there's not a lot more detail that I can give versus what we announced, but let's talk about the cost. I think we talked about in our shareholder -- in Analyst Day, and the Shareholder Day, that cost moving towards $6 million to $7 million per megawatt versus the $5 million as we've worked through the new design. So the $5 million was more for the previous design and this is the three story design that we're working for. So that's what we're looking at. But as far as economics, this fits in the economics that we've talked about previously, right, which is about $2 million per megawatt in revenue and $1 million of EBITDA per megawatt.
Lucas Pipes
Got it. That's very helpful. Thank you for that. And turning to guidance for a moment, in terms of the components that have been delayed, what exactly has been the bottleneck? Could you add a little bit more color on that? And then I think previously you provided some color on GPUs online for the average in physical Q3 and Q4. I think you mentioned it in your prepared remarks but I was trying to take notes and couldn't quite keep up. So if you have maybe an expectation around kind of Q3, Q4 GPUs, would appreciate the color around all of this. Thank you.
Wes Cummins
Sure. So the components that are the issue for delivery, It's not the GPUs themselves. It's specific networking components related to the InfiniBand networking piece of the cluster. And I've talked about this several times already, and this has been the bottleneck, I would say for the last kind of three or four months. We are getting delivery of those. It's a matter of making sure you get delivery of every component because you need all of the components to stand up the cluster, commission it, and get it operating for customers. So we're seeing improvement in the delivery of the InfiniBand. It's been specifically on the transceiver side of the InfiniBand deployments. So we had one cluster up and running last quarter. We have four deployed now. As I said in my prepared remarks, we expect to receive another four. And on the cluster, just as a reminder, Luke, is the cluster for us. I know it can be a little bit confusing, but it's 1,024 GPUs per cluster is how we refer to it. So we'll receive another four in the next two weeks is our expectation. And so also think of where pricing has gone for us on these clusters. You should think about $20 million of annual revenue per cluster deployed. So the difficulty we have is just a few weeks, because right now, let's say we're running about eight weeks behind our original expectation. But when you think about the revenue ramp and the revenue generation on a week by week basis, we go from one cluster, so we're generating $20 million of revenue per year for that cluster to our business goes to four clusters, which is $80 million of revenue per year. And then in a few weeks, our business goes to eight clusters, which is $160 million of revenue per year. So when we look at our guidance, we're just assuming 10 clusters deployed by the end of our fiscal year, which is May, about 4.5 months away. And so that 10 clusters plus our blockchain hosting solutions, that gets us to that $500 million run rate. I think we can do better than that, but that's the number that I think is a very conservative number for us to hit by the end of the year.
Lucas Pipes
That's very helpful. Thank you, Wes, for all the color. One quick one, the $45.8 million in property, equipment and other assets that have been purchased year-to-date, are you able to provide a breakdown between HPC and GPUs in that number? Thank you.
Wes Cummins
I'm sorry, Lucas, which number was that? The purchase year-to-date?
Lucas Pipes
Yes, that's the number of property and equipment that's been purchased to date.
Wes Cummins
So the majority of that goes into HPC. So when you look on our balance sheet -- so as you look at our balance sheet, where the GPUs are showing up because of how we're financing the GPUs is the lease asset, the right to use asset.
Lucas Pipes
Okay.
Wes Cummins
And then we'll have a -- and then on the liability side you'll see a capital lease. One of the things I would call out with this is the -- on the leases, we deploy the GPUs, the entire right to use asset goes into long-term assets, whereas on the lease liability, it's split about half and half between long-term liability and short-term liabilities. So when you look through the balance sheet, the leases right now, because we're doing what I always refer to as equipment finance. These are capital leases. That's how we're financing the GPUs. So when you look through CapEx, the vast majority of what you'll see is CapEx on equipment, is the HPC facility, the data center. And then on the lease right to use and the capital leases and the liabilities is where you'll see the GPUs.
Lucas Pipes
That is very helpful. Thank you, Wes, for all the color and best of luck.
Wes Cummins
Yeah. Thanks, Lucas.
Operator
Our next question is from George Sutton with Craig-Hallum Capital Group. Please proceed.
George Sutton
Thank you. It's great to see the conditional agreement. I just wondered if you could walk us through the project level financing side of this and sort of how do these ultimately come together in your mind? Is it a combination of construction loan financing and project financing? And any sense of the market dynamics there that we should know about?
Wes Cummins
Yeah. Thanks, George. So I've talked about this publicly before. So the way these agreements, we've been marketing this since mid-September. And when you think about this, you should think about marketing. People do -- our potential customers do a lot of due diligence on the site. You answer a significant number of questions, site visits, all of those things that you would expect in the due diligence. And then you'll typically go into a ROFR period, a right of first refusal period where someone gets exclusivity that you won't sell it outside of anyone. And I'm not talking about our specific agreement now, I'm just talking about the way that we've experienced this working. And then you work to get to a contract. And then post the contract, you go to project level finance. We are engaged with multiple parties on the project level finance side. We have been for a while. We haven't -- we're not waiting. We weren't waiting for an agreement to go to project level finance. But on the project level finance side, you'll get in the neighborhood of 65% to 80% loan to cost at the project level. So this won't go at the corporate level. It goes down at the site level, just like we've done with all of our Bitcoin sites. And then there's what we call the equity component, which I always look at as we work through this is more of a -- people would refer to in our industry as like mezz debt. So you have the construction finance, that debt runs kind of in the 7.5% to 8% type cost range. And then you have the mezz piece, the equity piece. And then you have our contribution to it and our contribution, we can have continued construction, we broke ground continued construction on the site in Ellendale, and we've put a significant amount of money into that already. So I think we're close to where we will need to be on the equity portion of that. And the remainder will come in from project level finance and then this mezz debt piece where typically someone will get kind of a high -- mid to high teens return on their capital and it's generally first money out and then maybe retains a small piece of ownership, call it [4%, 5% -- 3%, 5%] (ph) in the site itself. And so that's the process that we're in now.
George Sutton
So I wondered if you could walk through the 400 megawatts that you're ultimately marketing, obviously 100 megawatts now effectively spoken for. Just, it's very clear to us that the demand side of the equation here is going to be pretty significant. Just curious what you're seeing as you're going to market with the other 300 megawatts of opportunity.
Wes Cummins
Yeah, so demand was robust. We had two parties very deep in diligence last year. As of the thing that we have seen, which has been interesting in January, as we kicked the year off, we've had several more parties show up, three more in the last week. And it feels almost like kind of panic looking for capacity just in the last couple of weeks. So we're seeing a lot of interest and the parties that are involved would easily take more than the capacity that we have. So it's nice to get the first one close to over the finish line, but the expectation for me is that over the next month or two months, we're going to have the full 400 booked out. And as a reminder, I think when we think about this, the biggest issue that we face, which is a high quality issue, is how much do we carve out for ourselves because we want to carve some of this out for ourselves for our own cloud solution. And I think that's really the question mark, what's the highest and best value for these, for our assets, because we want to carve some out for ourselves, but we have a massive amount of demand for the capacity that we have. And, George, the reason we have that demand is, the 400 megawatts that we have is 400 megawatts that could come online over the next 18 months. Right? The power is available, the land is there, permitting, we're in a really good position in a market that is already short capacity and I think it's going to get worse over the next few months.
George Sutton
So last question for me on the Sai Computing side. Obviously we understand the supply chain challenge. Looking past that, I'm just curious on the demand side. So you mentioned you've got another four clusters that could ship here soon. Is there any demand challenge that you're seeing or has anything changed there or is it really just a limitation on the supply chain side?
Wes Cummins
So on the demand side, the only thing that has changed on the demand side is we're seeing a new group come into the market as I mentioned in my prepared remarks. So the demand remains extremely robust on the VC-backed startup companies. But what we're seeing in the market is what we're referring to as enterprise customers. So these, and just so I can define enterprise customers, these are companies that generally are very large companies, typically publicly traded, typically north of $50 billion type of market cap that have a business. They already have a business. And now to me, what it feels like is they have been working on their AI strategy, they have landed on what they plan to do with their AI, and now they're looking for significant amounts of GPU capacity. So that's a new element for us. We have one of those customers that we've been working with for about two months now, and they've moved into what we call -- we all call it a proof of concept, which is basically a test drive of our infrastructure. And I think we could get that customer into contracting here in the next two to three weeks. But that's been the only change, which is increased demand, but from a different segment of the market. The way I've looked at this market, it is a bit of a barbell, right? It's -- on one side you have the hyperscalers, so this was last year, you have the hyperscalers on one side and then you have the VC startups on the other side and then kind of there was nothing in the middle and now we're seeing that piece in the middle start to show up.
George Sutton
Perfect. Thank you for the details.
Wes Cummins
Absolutely. Thanks, George.
Operator
Our next question is from Darren Aftahi with ROTH MKM. Please proceed.
Darren Aftahi
Hey, good morning. Thanks for taking my questions. Just two, if I may. I think if I heard you correctly, pricing has kind of gone up on the GPU side. I think, Wes, maybe [say] (ph) on an annualized basis with the 10,000, you'd be over $200 million. I'm just kind of curious. I think at the Analyst Day, which is not too long ago, you talked about the $1.5 million monthly run rate. I guess what's changed? And then on the GPU side, I know you are targeting 10,000 by the end of your fiscal year. I guess given there's some uncontrollables on components, what's your level of confidence in that 10,000 number?
Wes Cummins
Sure. So, yeah, Darren, you're right. The pricing has gone up some. So what we're seeing -- the number I gave before is with one of our largest customers, and we were pricing that at around $2 an hour on the GPU capacity and our reserve contract. What we're seeing now is for most of our customers, we're signing contracts and what it looks like in the marketplace is kind of in the $2.20, $2.25 range for two-year reserve contracts and somewhere between 20% and 30% prepayments on the contracts. So that's the color on the kind of the update on pricing. And then on the 10,000, that's a number -- we took that down from the 26,000. So the original for us was 26,000. We had co-location capacity for 26,000. And where we are now is just with the slowness on the component side for InfiniBand. The 10,000 is a number that we feel really comfortable with hitting. We feel really comfortable in multiple ways, both on delivery of the GPUs, but also on the financing of the GPUs without going outside to a larger debt piece that has been done by some of the players in the market. So we feel really comfortable on both sides of that. But that's the reason we gave that guidance. However, what I would say about that is I think there's more we can do on the GPU side. So we'll have the 4,000 plus 4,000 shortly. So call it mid-February, we're at 8,000 of those clusters of the 10,000 that we're guiding for. So I think that leaves us plenty of room between there and the end of May.
Darren Aftahi
Well, I'm going to squeeze one more in. On Garden City, just looks like it obviously didn't ramp as fast as everyone expected. With the grid components, can you just maybe talk a little bit about maybe what is needed, how quickly those can get procured? And then I'm just kind of curious, your propensity to continue to do business maybe with the next site in somewhere like Texas, like, how would you kind of grade that in terms of wanting to do business in a place like Texas? Thanks.
Wes Cummins
Yeah, I mean, I live in Texas. I love it there. So I don't want to say anything bad about Texas, so I'll leave that. But the sites we're doing in the future right now are North Dakota and Utah. So those are the two areas we're working on. I think we've talked about this in the past. We're out looking for more capacity because of the demand we see in the market. We have a pipeline of additional capacity that's north of 1 gigawatt. So we're working through that. And then, specific to the Texas site, there's some improvements. I'm going to butcher this a little bit, but I think we need a little resiliency, which is a capacitor bank put in, not specifically in our location, but a substation that's in the area to get fully up to the 200. There's two ways to go there. There's getting approved for wind plus grid is one route, and the other is this, I think it's a capacitor bank that needs to be installed, and not a huge expense on that, by the way. But we've been working on that since either late November or December. And the guidance that we gave is what we view as the worst case scenario, which would be the April time frame for that -- the remainder of that to come on. So we're basically waiting for the last 65 megawatts to come on there.
Darren Aftahi
Great. Thank you.
Operator
Our next question is from Rob Brown with Lake Street Capital Markets. Please proceed.
Rob Brown
Hi, good morning.
Wes Cummins
Good morning, Rob.
Rob Brown
Just following up on the new anchor customer, could you give us a sense of sort of what vertical that customer is in, I think you mentioned, I guess, enterprise or the VC-backed side. Which sort of group is that customer in?
Wes Cummins
So we can't do that, but what I can tell you is the customer set that we are seeing of all the people that are looking at the site, there's not a lot of companies that exist in the world that are going to take down 100 megawatts or 200 megawatts or 300 megawatts themselves. So it's a very small group and all of the companies that are in the mix for us in North Dakota are names that everyone would easily recognize. They're looking for high-power density hosting, high-power density data center capacity, but it's all companies that you would recognize the name instantly.
Rob Brown
Okay. Okay. Thank you. And then I just wanted to follow up on the GPU discussion around, I guess, do you still have commitments to buy the 26,000 GPUs and deploy them after May or is that still to be determined on contract activity?
Wes Cummins
Yeah, those orders are still valid, even up to the 34,000. And so we just expect to continue. By the way, Rob, just to clarify on those orders, it's 34,000 right now for H100. We can still change those orders, whether it's for H200, GH200, right? These are fluid for us. So as the market evolves, we're able to react to that. But we still have those in queue and the ability to bring those when all of the components are available. So the issue in the quarter we had just now, as David mentioned in his prepared remarks, just to give clarity on that, so we took delivery of a second cluster during the quarter. We were paying for that cluster. Those were the expenses. In total, we're just under $4 million. So paying for that cluster without that cluster generating revenue because we didn't have the InfiniBand components to fully commission that cluster and turn it over to our customer. So we basically paused the GPU deliveries ourselves because we don't want to be paying for these GPUs while we can't offer them to our customers and generate revenue for them. So that's really what's happened for us, but we still have all those orders in place and expect to deploy those. We just need to be more careful, I guess, about when we expect those to be deployed.
Rob Brown
Got it. Great. Thank you for the color. I'll turn it over.
Operator
Our next question is from John Todaro with Needham & Company. Please proceed.
John Todaro
Great. Thanks for taking my question. A couple of ones here. One, so as you called out before, I think it was supposed to be 20,000 GPUs by December 23. Remember on the last call or the Analyst Day, maybe you did talk about some possible delays in the InfiniBand. Just kind of curious, did that situation get worse than you expected or were those delays kind of on your mind and maybe you guys just mis kind of analyzed it?
Wes Cummins
So, John, specifically on those, we're getting delivery of certain -- of the vast majority of the networking equipment that we needed. There was one particular component, but you need that component to make it work. So in an instance where in one of our clusters, right, we have the entire InfiniBand setup besides like 28 transceivers, just as an example, but you can't fully commission that cluster and generate revenue. So it's been just squeezing in those transceivers is specifically what it is. So like I said, we could have taken delivery of a lot more GPUs, but I see no point, and I think it's detrimental to us to take delivery and not be able to generate revenue and pay for the GPUs. So we just held that off and made that decision in December. We could have taken a significant number of GPUs in December, but again, no reason to do that. So as those come available, I think that you could see us speed that up again significantly and we've started to see that loosen up in the marketplace in late December and January but I'm not ready to say that we'll be able to speed that up to meet kind of the 20,000 and then 26,000 and 34,000 deployment.
John Todaro
Got it. Okay. And then that's helpful. Another question I had, so you had mentioned this new customer contract in the Sai compute side, enterprise customer, how you could move to a proof of concept and start maybe delivering on that contract shortly. Just curious, with the delay, wouldn't the previously existing contract come before this one, or did anything change with those contracts?
Wes Cummins
Nothing has changed with those contracts. And just to clarify, John, we don't have a contract with the enterprise customer. It's in proof of concept and then we'd be moving to contracting. So just want to be clear on that. But nothing's changed with our previous customers. But it's exciting to see an entire new group show up in the marketplace looking for significant amount of GPU compute. And these are established companies that make money and have a product. And it's just calling out that that's kind of a new area of the market that we've seen develop over the last -- it really started kind of in late November.
John Todaro
Okay, got it. And just to clarify though, so even with the delays though, customers are still kind of lining up, knocking down doors, it sounds like?
Wes Cummins
Yeah, we have had no issue with that.
John Todaro
Okay, got it. Thanks, guys.
Wes Cummins
I would say, and I said this earlier, John, but I would say we've seen -- with the entrance of the enterprise customer, I would say that overall we've seen demand increase from our last conference call. We haven't seen anything slow down. We've seen it increase.
John Todaro
Got it. Thank you. Appreciate it.
Operator
[Operator Instructions] Our next question is from Mike Grondahl with Northland Securities. Please proceed.
Mike Grondahl
Hey guys. What is a rough estimate of Applied’s contribution to complete financing of the construction, just the project side of it for the anchor tenant? What do you guys have to pitch into that, roughly, just a number? How much have you done so far?
Wes Cummins
So -- good question, Mike. When you think about -- we say it's 7 million megawatts. Let's go to the high end of that. So it's 7 million megawatts, right, we're going to land anywhere from 65% to 85% -- sorry, 65% to 80% on construction finance, on project level finance. And then we expect to have what we in the industry would call an equity partner. However, as I explained earlier, that generally looks more like mezz debt. And so at the end, you're looking at us, our expectation is that we'll contribute somewhere between 5% and 10% of the project in cash for the equity portion of that. And we've already spent north of…
David Rench
$25 million.
Wes Cummins
…$25 million on that currently for Ellendale. And then, Mike, when we look at that too, I think David mentioned this in his prepared remarks is for us on the cash flow for the company, I think it's important to note that when we hit March 1, the cash flow from our Bitcoin data centers improves dramatically. Right? We burned through the vast majority of the prepayments at that point and so the cash flow from that portion of our business improves dramatically.
Mike Grondahl
Got it, got it, that's helpful. And I'm assuming that the anchor tenant will be supplying the GPUs, but could you clarify there?
Wes Cummins
Yeah. So on the data center business, we are just providing space. Think of this as an Equinix or DLR style business. We're providing the space, they select the equipment, and they buy the equipment and then we're just hosting. Very similar to what we do on the Bitcoin side.
Mike Grondahl
Got it. Got it. And roughly, when would you begin to recognize revenue with this anchor customer? And hey, I'm not going to hold you to it, but roughly as this plays out, when would you expect that revenue to start?
Wes Cummins
So, right now, the expectation is the revenue would be the very early part of the second half of this calendar year.
Mike Grondahl
So, like July, August would be kind of…
Wes Cummins
Yeah.
Mike Grondahl
Okay. Great. Then just one last question. The Bitcoin hosting business, any contract renewals or extensions, any updated kind of terms, or is everything kind of locked down in that business?
Wes Cummins
Yeah, it's locked down. The things of note, our largest customer there, we have four years, maybe a little over four years on most of the capacity for our largest customer there on the contract. The only thing of note, and this probably shouldn't surprise you, but we're seeing -- we’re getting a lot more calls about hosting capacity. We are never out marketing because we don't have any to offer, but we are getting a lot more calls about hosting capacity over the last you don't call it six weeks, eight weeks as the price of Bitcoin has went up significantly.
Mike Grondahl
Got it. Okay. Hey, thank you.
Wes Cummins
Absolutely.
Operator
Our next question is from Kevin Dede with H.C. Wainwright. Please proceed.
Kevin Dede
Hi, Wes. [Technical Difficulty] question. I'm curious about the number of facilities that you're running or leasing to supply the power you need for your cloud service at the moment.
Wes Cummins
Sure. So we have third-party facilities in Denver, in Minnesota, and in Salt Lake City. And then we have our own Jamestown facility in Jamestown, North Dakota.
Kevin Dede
Yeah. And Jamestown, you have maxed out at 8 megawatts, if I understand correctly.
Wes Cummins
So Jamestown will hold 5,000 GPUs for us. I can do it on megawatts, or I can tell you on the GPU capacity, however you prefer.
Kevin Dede
Oh, whatever you're used to is fine. I guess the real genesis of the question is in meeting a cloud build-out, right? The amount -- given the market's really tight, how are you securing the power that you need to meet your customer demands on the cloud service side?
Wes Cummins
Yeah, so we secured this power back in the summer of last year. So middle of the year last year, we secured these power sites. We saw the demand and we ran out and grabbed the capacity to be able to service it. So we secured that at that point. And so we have the capacity both from third party and our own facilities for the 26,000 GPUs that we talked about. And then as we go beyond that, our expectation is that it will go into our own facilities post that 26,000 as we continue to grow.
Kevin Dede
Right, so [26,000 to 30,000 of that Delta will go] (ph) into the Ellendale facility that's under construction now.
Wes Cummins
Yeah.
Kevin Dede
Okay. I think I correctly heard you mention a marketing initiative. And I was wondering if you could be more specific about the direction that you're going to take that. Do you think that goes toward that enterprise market that you're seeing starting to develop?
Wes Cummins
Yeah. So it's specifically for the enterprise market. So, as a reminder, Kevin, I think we've talked about this before, but we've never had a single salesperson at our company. And we hired our first salesperson a few months ago. And we're going to add more to that capability specifically to go after this enterprise market that we see developing now. So that's our expectation. But previous to that, we've never had a salesperson. We've had no sales force whatsoever in the company.
Kevin Dede
Understood. And congrats on that. I guess I'm just still a little shaky in how you see that market developing. And I guess sort of your competitive positioning and if demand is so strong, why would a marketing initiative be necessary at all?
Wes Cummins
So, we've hired one salesperson, maybe we'll hire another one, but there's an idea about going out and making sure people know who we are and what we do because we've never done that before in the history of the company. So that's really the idea. So I don't think you should be thinking that we're hiring 20 salespeople, but I do think it's prudent to get a few salespeople on board, develop that sales organization. And the idea here that we've discussed a lot internally is we don't need it now, but we don't want to wait until we need it to build it.
Kevin Dede
That makes sense. Okay. Well, thanks for that. [indiscernible] any offense in kind of looking for clarification. Thanks a lot.
Wes Cummins
No, no, no. That's it though. But yeah, Kevin, we're discussing it on the call here, but we've debated a lot internally where there's a group of people saying, why would we have sales people, we don't have anything to sell. And another side that goes, the things that we should develop this before we need it. And I have landed in the latter camp, which is I think it's a good idea to develop a sales team before you actually need it.
Kevin Dede
I don't disagree. Right marketing makes it all happen. Well, congrats again on the results, Wes. Thank you very much for taking the questions.
Wes Cummins
Thanks, Kevin.
Operator
[Operator Instructions] And there are no further questions at this time. I would like to hand the conference back over to Wes for some closing comments.
Wes Cummins
Thanks and thanks everyone for joining. And my last comment is just I want to say thank you to the team in Ellendale. We're sitting here in January. It's not the most pleasant climate for them to be continuing construction on our site. So we really appreciate the team and everyone who makes it happen for us. And I look forward to speaking to you next quarter.
Operator
Thank you. That will conclude today's conference. You may disconnect your lines at this time and thank you for your participation.
For further details see:
Applied Digital Corporation (APLD) Q2 2024 Earnings Call Transcript