Applied Digital Corp (APLD) 2024 Q1 法說會逐字稿

Good morning, and welcome to Applied Digital's Fiscal First Quarter 2024 Conference Call. My name is Robin, I'll be your operator for today.

早安，歡迎參加 Applied Digital 2024 年第一季財報電話會議。我叫羅賓，今天我是你們的接線生。

Before this call, Applied Digital issued its financial results for the fiscal first quarter ended August 31, 2023, in a press release, a copy of which will be furnished in a report on Form 8-K filed with the SEC and will be available in the Investor Relations section of the company's website.

在此電話會議之前，Applied Digital 在一份新聞稿中發布了截至2023 年8 月31 日的第一財季財務業績，該新聞稿的副本將在向SEC 提交的8-K 表格報告中提供，並將於公司網站的投資者關係部分。

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.

參加今天電話會議的還有 Applied Digital 董事長兼執行長 Wes Cummins；和財務長大衛·倫奇。在他們發表講話後，我們將開始提問。在我們開始之前，Gateway Group 的 Alex Kovtun 將做簡短的介紹性演講。科夫通先生，請繼續。

Great. Thank you, operator. Good morning, everyone, and welcome to Applied Digital's Fiscal First Quarter 2024 Conference Call. Before management begins their 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 that there are a 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.

偉大的。謝謝你，接線生。大家早安，歡迎參加 Applied Digital 2024 年第一季財報電話會議。在管理階層開始正式發言之前，我們想提醒大家，我們今天發表的一些聲明可能被視為證券法下的前瞻性聲明，並涉及許多風險和不確定性。因此，我們警告稱，有許多因素（其中許​​多因素超出了我們的控制範圍）可能導致實際結果和事件與前瞻性陳述中描述的結果和事件有重大差異。

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 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.

有關我們前瞻性陳述的更詳細的風險、不確定性和假設，請參閱我們向美國證券交易委員會提交的收益報告和公開文件中的揭露內容。我們不承擔任何更新前瞻性陳述以反映前瞻性陳述做出之日後發生的情況或事件的義務或承諾，除非法律要求。

We will also discuss non-GAAP financial metrics and encourage you to read our disclosures and the reconciliation tables, 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 filing for free, by visiting the SEC website at www.sec.gov.

我們還將討論非 GAAP 財務指標，並鼓勵您在考慮這些指標時仔細閱讀我們的揭露和調整表以及收益發布中適用的 GAAP 衡量指標。我們請您參閱我們向美國證券交易委員會提交的文件，以了解我們業務的詳細披露和描述以及不確定性和其他可變情況，包括但不限於我們季度報告中風險因素標題下確定的風險和不確定性。表 10-Q。您可以透過造訪 SEC 網站 www.sec.gov 免費取得 Applied Digital 的證券交易委員會備案。

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.

我還想提醒大家，本次電話會議正在錄音，並將透過 Applied Digital 網站投資者關係部分的連結進行重播。

Now I will turn the call over to Applied Digital's Chairman and CEO, Wes Cummins. Wes?

現在我將把電話轉給 Applied Digital 董事長兼執行長 Wes Cummins。韋斯？

Thanks, Alex, and good morning, everyone. Thank you for joining our fiscal first quarter 2024 conference call. I want to start by thanking our employees for their ongoing hard work and service and advancing our mission of providing digital infrastructure solutions to the rapidly growing high-performance computing industry.

謝謝，亞歷克斯，大家早安。感謝您參加我們的 2024 年第一財季電話會議。首先，我要感謝我們的員工持續的辛勤工作和服務，並推進我們為快速發展的高效能運算產業提供數位基礎設施解決方案的使命。

Before turning the call over to our CFO, David Rench, for a detailed review of our financial results, I'd like to briefly discuss some recent developments across our business.

在將電話轉給我們的財務長 David Rench 詳細審查我們的財務業績之前，我想簡要討論一下我們業務的一些最新發展。

Let's start with our existing blockchain hosting operations. We aim to have all 3 of our blockchain hosting facilities fully online shortly with high reliability and performance for our customers. Our 100-megawatt Jamestown facility continues to perform as expected and operates at full capacity with consistent uptime throughout the quarter. This marks the fourth consecutive quarter in which the Jamestown facilities has operated at full capacity.

讓我們從現有的區塊鏈託管操作開始。我們的目標是讓我們的所有 3 個區塊鏈託管設施盡快全面上線，為我們的客戶提供高可靠性和高效能。我們的 100 兆瓦詹姆斯敦工廠繼續按預期運行，滿載運行，整個季度的正常運行時間保持一致。這標誌著詹姆斯敦設施連續第四個季度滿載運作。

Our 180-megawatt Ellendale facility in North Dakota was fully energized and became fully operational during the first quarter of fiscal year 2024 and contributed to our results this quarter, the facility is fully online and operating with consistent uptime during the second quarter, bringing our total hosting capacity to 280 megawatts across our North Dakota facilities, all of which are contracted out to customers on multiyear terms.

我們位於北達科他州的180 兆瓦Ellendale 設施在2024 財年第一季度完全通電並全面投入運營，為本季度的業績做出了貢獻，該設施在第二季度全面上線並保持穩定的正常運行時間，使我們的總發電量達到了我們北達科他州設施的託管容量高達 280 兆瓦，所有設施均按多年期合約外包給客戶。

In September, we entered into a facility extension agreement with Oncor for the transmission and metering of power to our Garden City, Texas facility. With this in place, metering and telemetry equipment owned by the power provider will be installed and once complete, the site will be energized. This installation is expected to be completed by October 23rd. Once our Garden City facility becomes fully energized, we will have approximately 500 megawatts of hosting capacity across our 3 facilities.

9 月，我們與 Oncor 簽訂了一項設施擴展協議，為我們位於德克薩斯州花園城的設施提供電力傳輸和計量。完成後，將安裝電力提供者擁有的計量和遙測設備，完成後，該站點將通電。該安裝預計將於10月23日完成。一旦我們的花園城設施完全通電，我們的 3 個設施將擁有約 500 兆瓦的託管容量。

We expect our 3 sites to produce around $300 million in revenue and $100 million in EBITDA on an annualized basis. Presence of all 3 operational facilities with high uptime will provide us with consistent cash flow, supporting our capital requirements for the build-out of our HPC data centers and purchasing GPUs to service our AI cloud customers.

我們預計我們的 3 個基地每年將產生約 3 億美元的收入和 1 億美元的 EBITDA。所有 3 個營運設施的正常運作時間較長，將為我們提供穩定的現金流，支援我們擴建 HPC 資料中心的資本需求，並購買 GPU 來為我們的 AI 雲端客戶提供服務。

Let's move to our AI cloud services, which launched this calendar year to provide accelerated computing power for AI applications. Our AI cloud service continues to ramp up as we make further progress in supporting our existing contracts and pursue additional opportunities in our pipeline. In July, we activated the first cluster of GPUs for character AI and since then have made meaningful progress receiving our second cluster of GPUs in September with the expectation of receiving additional GPUs this month.

讓我們轉向今年推出的人工智慧雲端服務，為人工智慧應用程式提供加速運算能力。隨著我們在支援現有合約方面取得進一步進展並在管道中尋求更多機會，我們的人工智慧雲端服務繼續增強。 7 月，我們為角色 AI 激活了第一個 GPU 集群，此後在 9 月接收第二個 GPU 集群方面取得了有意義的進展，預計本月將收到更多 GPU。

Since our last earnings announcement, we have added two additional AI cloud customers. Both customers have an established user base and are growing quickly. These customer agreements have a similar structure to our first two. They also include significant prepayments to fund a large portion of the capital requirements for purchasing the GPUs.

自從我們上次發布財報以來，我們又增加了兩個人工智慧雲端客戶。兩家客戶都擁有成熟的用戶群，並且正在快速成長。這些客戶協議的結構與我們的前兩個協議類似。它們還包括大量預付款，以滿足購買 GPU 的大部分資本需求。

This brings our total annual contract value of AI Cloud services contracts at full capacity to approximately $378 million. In addition to substantial prepayments we received from customers, we are using vendor financing and actively exploring other tailored financing options to support the capital requirements for the 34,000 GPUs we have on order to support our cloud service. We remain on track for delivery of the majority of these GPUs by April of next year.

這使得我們滿載的人工智慧雲端服務合約年度合約總價值達到約3.78億美元。除了從客戶收到的大量預付款之外，我們還利用供應商融資並積極探索其他客製化融資方案，以支援我們訂購的 34,000 個 GPU 的資本需求，以支援我們的雲端服務。我們仍有望在明年 4 月之前交付大部分 GPU。

Our established partnerships with leading OEMs like Supermicro, Hewlett Packard Enterprise and Dell, combined with our recent elite partner status in NVIDIA's partner network provide us with visibility into the delivery time line, ensuring timely receipt of these GPUs. As previously mentioned, we will initially provide this service from our 9-megawatt HPC Jamestown facility along with third-party colocation space as we continue to execute on the element of our dedicated next-gen HPC data centers.

我們與Supermicro、Hewlett Packard Enterprise 和Dell 等領先OEM 廠商建立了合作夥伴關係，再加上我們最近在NVIDIA 合作夥伴網絡中的精英合作夥伴地位，使我們能夠了解交付時間線，確保及時收到這些GPU 。如前所述，我們最初將透過我們的 9 兆瓦 HPC 詹姆斯敦設施以及第三方主機託管空間提供這項服務，同時我們將繼續執行我們專用的下一代 HPC 資料中心的元素。

The pipeline of opportunities for our AI cloud service business remains robust. We look forward to capitalizing further on this opportunity and providing further updates on our newly signed customers going forward.

我們的人工智慧雲端服務業務的機會依然強勁。我們期待進一步利用這一機會，並提供有關我們新簽約客戶的更多最新資訊。

Lastly, let me provide a quick update on our purposeful HPC data centers. We have 300 megawatts of capacity in development and have begun initial groundwork for our Ellendale facility. In order to proceed with the construction of these facilities and obtain the necessary financing, we're in the process of securing a credit-rated anchor tenant. We have been actively engaged in ongoing discussions with several potential anchor tenants for our Ellendale and Utah facilities.

最後，讓我快速介紹一下我們有針對性的 HPC 資料中心的最新情況。我們擁有 300 兆瓦的開發容量，並已開始為埃倫代爾工廠進行初步基礎工作。為了繼續建造這些設施並獲得必要的融資，我們正在尋找一個具有信用評級的主力租戶。我們一直在積極與埃倫代爾和猶他州工廠的幾個潛在主力租戶進行持續討論。

Similar to our new customers for our AI cloud service, we will provide more information once available. We aim to secure an anchor tenant customer for each facility and have these facilities fully energized and operational within the next 24 months. With that, I now turn the call over to our CFO, David Rench, to walk you through our financials and provide an update on guidance before providing my closing remarks. David?

與我們的人工智慧雲端服務的新客戶類似，我們將在可用後提供更多資訊。我們的目標是為每個設施爭取一個主要租戶客戶，並在未來 24 個月內讓這些設施完全通電並投入營運。現在，我將電話轉給我們的財務長 David Rench，在我發表結束語之前，他將向您介紹我們的財務狀況並提供最新的指導意見。大衛？

Thanks, Wes, and good morning, everyone. Revenues for the fiscal first quarter of 2024 were $36.3 million compared to $6.9 million for the fiscal first quarter of 2023. The increase in hosting revenues was driven by an increase in online capacity due to Ellendale, North Dakota site being operational and revenue from the company's first AI cloud service contract, which began during the three months ended August 31, 2023.

謝謝，韋斯，大家早安。 2024 年第一財季的收入為 3,630 萬美元，而 2023 年第一財季的收入為 690 萬美元。託管收入的增長是由於北達科他州埃倫代爾網站運營導致在線容量增加以及公司的收入第一份人工智慧雲端服務合同，於截至2023 年8 月31 日的三個月內開始。

Cost of revenues for the fiscal first quarter of 2024 was $24.4 million compared to $6.1 million for the fiscal first quarter of 2023. The increase in costs was attributable to higher energy costs used to generate hosting revenues, depreciation, amortization expense and personnel expenses for employees directly working on our Jamestown and Ellendale hosting facilities.

2024 年第一財季的收入成本為2,440 萬美元，而2023 年第一財季的收入成本為610 萬美元。成本增加的原因是用於產生託管收入、折舊、攤提費用和員工人事費用的能源成本增加直接在我們的詹姆斯敦和埃倫代爾託管設施上工作。

Operating expenses for the fiscal first quarter of 2024 were $17.1 million compared to $5 million in the prior year comparable period. The increase was primarily due to personnel-related costs as a result of the increase in head count as well as an increase in depreciation. Net loss for the fiscal first quarter of 2024 was $9.6 million or a loss of $0.10 per basic and diluted share based on a weighted average share count during the quarter of approximately 100.5 million.

2024 年第一財季的營運支出為 1,710 萬美元，而上年同期營運支出為 500 萬美元。這一增長主要是由於人員數量增加以及折舊增加導致的人員相關成本。 2024 年第一季的淨虧損為 960 萬美元，根據該季度約 1.005 億股的加權平均股數計算，每股基本股和稀釋股虧損 0.10 美元。

This compares to a net loss of $4.7 million or a loss of $0.05 per basic and diluted share in the fiscal first quarter of 2023 based on a weighted average share count during the quarter of approximately 93.1 million.

相較之下，根據該季度約 9,310 萬股的加權平均股數，2023 年第一財季的淨虧損為 470 萬美元，或每股基本股和稀釋股虧損 0.05 美元。

Adjusted net income, a non-GAAP measure for the fiscal first quarter of 2024 was $0.1 million or adjusted net income per basic and diluted share of less than $0.01. Based on a weighted average share count during the quarter of approximately 100.5 million. This compares to an adjusted net loss, a non-GAAP measure of $3.4 million or a loss of $0.04 per basic and diluted share for the fiscal first quarter of 2023. Based on our weighted average share count during the quarter of approximately 93.1 million.

2024 年第一季的調整後淨利（非 GAAP 衡量標準）為 10 萬美元，即每股基本股和稀釋股調整後淨利低於 0.01 美元。基於本季約 1.005 億股的加權平均股票數。相較之下，2023 年第一財季的調整後淨虧損（非GAAP 衡量標準為340 萬美元）或每股基本股和稀釋股虧損0.04 美元。基於我們本季約9,310 萬股的加權平均股數。

Adjusted EBITDA, a non-GAAP measure for the fiscal first quarter of 2024 was $10 million compared to an adjusted EBITDA loss for the fiscal first quarter of 2023 of $1.9 million.

2024 年第一財季調整後 EBITDA（非 GAAP 衡量標準）為 1,000 萬美元，而 2023 年第一季調整後 EBITDA 虧損為 190 萬美元。

Lastly, on the balance sheet, we ended the fiscal first quarter with $31.2 million in cash, cash equivalents and restricted cash and $44 million in debt. During the fiscal first quarter of 2024, we received $39.5 million in customer payments due to the structure of our commercial arrangements with our customers that incorporate upfront deposits and prepayments.

最後，在資產負債表上，第一財季結束時，我們的現金、現金等價物和限制性現金為 3,120 萬美元，債務為 4,400 萬美元。在 2024 年第一財季，由於我們與客戶的商業安排結構包含預付款和預付款，我們收到了 3,950 萬美元的客戶付款。

In certain contracts, the prepayments are amortized back to the customer over the first year of their contract with no impact on revenue recognition, but the timing of cash flow with upfront cash to us is a major benefit for the company and that it helps our CapEx funding needs as we build out our data centers.

在某些合約中，預付款在合約第一年分攤給客戶，對收入確認沒有影響，但預付現金給我們的現金流的時間安排對公司來說是一個主要好處，它有助於我們的資本支出我們建設資料中心時需要資金。

Since the quarter closed, we have received an additional $15 million in customer prepayments and are expecting an additional $23 million this week.

自本季結束以來，我們已額外收到 1,500 萬美元的客戶預付款，預計本週將額外收到 2,300 萬美元。

Now turning to guidance for the full year fiscal 2024. We are reaffirming our expectations for revenue in the range of $385 million to $405 million and adjusted EBITDA in the range of $195 million to $205 million. Now I'll turn the call over to Wes for closing remarks.

現在轉向 2024 財年全年指引。我們重申營收預期在 3.85 億美元至 4.05 億美元之間，調整後 EBITDA 在 1.95 億美元至 2.05 億美元之間。現在我將把電話轉給韋斯做總結發言。

Thank you, David. As we look ahead, we remain confident in our competitive advantages and differentiated capabilities to meet the sophisticated and demanding requirements for businesses and enterprises to run AI workloads and other emerging HPC applications. I remain optimistic about the future of Applied Digital as we solidify our leadership in next-generation digital infrastructure for both blockchain and non-blockchain HPC use cases during this era of digital transformation.

謝謝你，大衛。展望未來，我們對自己的競爭優勢和差異化能力仍然充滿信心，能夠滿足企業和企業運行人工智慧工作負載和其他新興高效能運算應用程式的複雜而苛刻的要求。我對 Applied Digital 的未來保持樂觀，因為我們在這個數位轉型時代鞏固了在區塊鏈和非區塊鏈 HPC 用例的下一代數位基礎設施方面的領導地位。

I'd like to thank all of our team members for their dedication in making Applied what it is today and our shareholders for trusting us and our mission and execution. We're now happy to take questions. Operator?

我要感謝我們所有團隊成員的奉獻精神，使應用材料公司取得了今天的成就，並感謝我們的股東對我們以及我們的使命和執行力的信任。我們現在很樂意回答問題。操作員？

(Operator Instructions) Thank you. And our first question comes from the line of Rob Brown with Lake Street Capital.

（操作員說明）謝謝。我們的第一個問題來自 Rob Brown 與 Lake Street Capital 的對話。

Good morning. First question on the -- on your kind of efforts for anchored customers in the AI business. I can't give too much detail, but maybe a sense of kind of the sizing of the center. Do they need to fill out kind of a whole center before you kick off? And how does that sort of kind of scope out to get -- to make the decision to kick off a data center build?

早安.第一個問題是關於您為人工智慧業務中的固定客戶所做的努力。我無法提供太多細節，但也許可以了解中心的大小。在你開始之前他們需要填寫整個中心嗎？這種範圍如何做出啟動資料中心建置的決定？

Yes. I think we've shown in the past, we've talked about this, and it remains the same, which is we think we need to contract roughly 70% of the capacity. Doesn't have to be one customer, but we think we need to contract that on a reasonable, say, 7- to 10-year contracts with renewals to kick off the construction, and we're in process of securing that.

是的。我認為我們過去已經表明，我們已經討論過這一點，而且它保持不變，即我們認為我們需要收縮大約 70% 的產能。不一定是一個客戶，但我們認為我們需要簽訂一份合理的合同，例如 7 到 10 年的合同，並續簽才能開始建設，我們正在確保這一點。

But Rob, the way I think about these is kicking off roughly 100 megawatts in North Dakota shortly and then followed up with 100 in Utah and then come back kind of mid next year for the second 100 in North Dakota.

但羅布，我對這些問題的看法是，不久之後將在北達科他州啟動約 100 兆瓦，然後在猶他州啟動 100 兆瓦，然後在明年中期在北達科他州啟動第二個 100 兆瓦。

Okay. Great. And then on the ramp of the AI business, it depends on getting the GPUs sort of delivered. How do you sort of see that ramping of the GPU's you've order? How do you see those coming in, in terms of the ability to ramp up that business?

好的。偉大的。然後，隨著人工智慧業務的發展，這取決於 GPU 的交付。您如何看待您所訂購的 GPU 的提升？就擴大業務的能力而言，您如何看待這些進來的人？

Yes. So we -- I think at this point, we're getting pretty good visibility on delivery schedules. It's gotten better since our last call. And as I mentioned in the prepared remarks that we received our second cluster in September. Expect to receive more this month and really expect to start to receive large volumes in November, December and January.

是的。因此，我認為目前我們對交付時間表有了很好的了解。自從我們上次通話以來，情況已經好轉了。正如我在準備好的發言中提到的，我們在 9 月收到了第二個集群。預計本月將收到更多貨物，並且真正預計在 11 月、12 月和 1 月開始收到大量貨物。

Our next question is from the line of George Sutton with Craig Hallum.

我們的下一個問題來自喬治·薩頓和克雷格·哈勒姆。

This is Adam on for George. Wes, starting with Garden City, could you provide a little more detail on the pace of energization once Oncor has installed their equipment?

這是亞當替喬治上場。 Wes，從 Garden City 開始，您能否提供有關 Oncor 安裝設備後通電進度的更多詳細資訊？

Yes. Sure. So the date that we mentioned in the prepared remarks and in the press release is the expectation that everything will be ready to go for energization. We should start energizing that day. And then the ramp should be faster than we've seen at our North Dakota facilities where we were continuing to finish buildings and energize them.

是的。當然。因此，我們在準備好的演講和新聞稿中提到的日期是期望一切準備就緒，為通電做好準備。我們應該從那天開始充滿活力。然後坡道應該比我們在北達科他州工廠看到的要快，我們在那裡繼續完成建築物並為其供電。

It will still take weeks to energize, but it won't be the months that we've seen in both Jamestown and Ellendale because construction is complete, the miners are racked and ready to turn on. And so it should go much faster. So at some point in kind of the November -- mid- to late November should be fully online from that late October start, the 23rd.

仍然需要幾週的時間才能通電，但這不會是我們在詹姆斯敦和埃倫代爾看到的幾個月，因為建設已經完成，礦工們已經準備好開機。所以它應該會更快。因此，在 11 月的某個時間點——11 月中下旬應該會從 10 月下旬開始（即 23 日）全面上線。

Great, one more follow-up for me. With respect to current construction efforts, are there any milestones that you need to hit before you face more adverse weather?

太好了，我還有一個後續行動。就目前的施工工作而言，在面臨更惡劣的天氣之前，您需要達到哪些里程碑？

Yes. In North Dakota, we're -- we started on the groundwork there, and it's kind of the same thing we did last year, which is it will be a little bit of a rush to get foundations poured. To be able to work through the winter to enclose the facility and work through the winter there. And so that's what we're doing right now.

是的。在北達科他州，我們開始了基礎工作，這和我們去年做的事情有點像，就是澆地基會有點倉促。能夠在整個冬天工作，封閉設施並在那裡工作整個冬天。這就是我們現在正在做的事情。

Our next question is from the line of John Todaro with Needham & Company.

我們的下一個問題來自李約瑟公司的約翰·托達羅。

One here on -- so the site computing business, just trying to understand -- first off, congrats on adding a few more contracts. But kind of what is the capacity out there? Can -- is there more contracts we can add now? Or are you looking to get those HPC sites on and anchor tenants there? Just kind of wondering really how much capacity if we can expect any more contracts to come online?

這裡的一個 - 所以站點計算業務，只是想了解 - 首先，恭喜你添加了更多合約。但那裡的容量是多少呢？我們現在可以添加更多合約嗎？或者您是否希望建立這些 HPC 網站並在那裡固定租戶？只是想知道如果我們可以期待更多的合約上線，到底有多少容量？

Yes. We have some more capacity, not a lot until we start to bring our own facilities online. So John, if we walk through kind of the math on the site compute side. So roughly every 1,000 GPUs generates approximately $1.5 million of revenue per month. And we've talked about before about getting to 26,000 on by April. And so that gets you to, call it, $460 million plus annual revenue business. And then when we go from April on, we'll be looking to put it in our own facilities. So through that ramp in April, we have our Jamestown facility and then we've secured third-party colo to support that ramp and then we're looking to put further deployments in our own facilities.

是的。在我們開始將自己的設施上線之前，我們還有更多的容量，但不會很多。約翰，如果我們在站點計算方面進行一些數學計算。因此，大約每 1,000 個 GPU 每月可產生約 150 萬美元的收入。我們之前已經討論過 4 月要達到 26,000 人。這樣一來，您的業務年收入就達到了 4.6 億美元。從四月開始，我們將考慮將其放入我們自己的設施中。因此，透過四月份的斜坡，我們擁有了詹姆斯敦設施，然後我們已經獲得了第三方託管來支援該斜坡，然後我們希望在我們自己的設施中進行進一步的部署。

Got it. Okay. Great. That's helpful. And then I guess just on the timeline, too. So that's when you guys increased from 26,000 to 34,000 GPUs, just any kind of color from the suppliers that it seemed like certainly doable? Or is that -- what is the difficulty in kind of get that? I know you guys said you're tracking, but any more color there would be helpful.

知道了。好的。偉大的。這很有幫助。然後我想也只是在時間線上。那麼，當你們將 GPU 從 26,000 個增加到 34,000 個時，供應商提供的任何一種顏色似乎確實可行？或者說——要實現這一點有什麼困難？我知道你們說你們正在追踪，但更多的顏色會有幫助。

Yes. So those are the -- what we expect to deploy post April of next year. And so we've added those in because of, again, the pipeline of demand that we're seeing. But you should think about those being added after April. So the 26 that we talked about on the last call being added through April and then the additional 8 being added post, but I would continue to think about it's 8 right now, but we'll see where the demand comes through. So if we need to expand that, we should have our own data center capacity coming online to support that.

是的。這些就是我們預計在明年四月後部署的內容。因此，我們再次添加這些內容是因為我們看到了需求的成長。但你應該考慮四月之後添加的內容。因此，我們在上次電話會議上討論的 26 個將在 4 月之前添加，然後在發布後添加額外的 8 個，但我會繼續考慮現在是 8 個，但我們將看看需求來自哪裡。因此，如果我們需要擴展，我們應該擁有自己的資料中心容量來支援這一點。

Next question comes from the line of Lucas Pipes with B. Riley Securities.

下一個問題來自 Lucas Pipes 與 B. Riley Securities 的關係。

My first question is on guidance. And with the delays at Garden City should we think kind of more of the EBITDA and revenue contribution having shifted to AI Cloud. Thank you very much for walking me through those changes.

我的第一個問題是關於指導。隨著 Garden City 的延誤，我們是否應該認為更多的 EBITDA 和收入貢獻已轉移到人工智慧雲端。非常感謝您帶我經歷這些改變。

Yes. Thanks, Lucas, for the question. So the -- given what I just said about the AI cloud, you can kind of walk through as GPUs are deployed, how that steps up throughout the year, but it will be a fairly steep ramp, especially in our last 2 quarters as these come online, we'll get the deliveries. We got the delivery in September. We'll get additional in October.

是的。謝謝盧卡斯提出的問題。因此，考慮到我剛才所說的關於人工智慧雲端的內容，您可以了解一下GPU 的部署情況，以及全年的進展情況，但這將是一個相當陡峭的斜坡，特別是在我們最後兩個季度，因為這些上線，我們就會收到貨。我們九月就收到貨了。我們將在 10 月獲得更多。

And then I think the deliveries really ramp for us in November, December and somewhat into January. So as you see those come online, you'll see a pretty steep revenue ramp on the cloud business through the remainder of the year. That should make up for the delay that we've seen in Garden City.

然後我認為我們的交付量在 11 月、12 月和 1 月會增加。因此，當您看到這些產品上線時，您將看到今年剩餘時間內雲端業務的收入大幅成長。這應該可以彌補我們在花園城看到的延誤。

On the HPC side, can you walk us through the capital intensity with many of the planning completed from what I understand. And in terms of financing, what are your current targets for debt to equity? Is it project financing that would cover the majority of the capital needs would appreciate your thoughts on that.

在 HPC 方面，您能否向我們介紹一下資本密集度以及根據我的理解完成的許多規劃。在融資方面，你們目前的債轉股目標是多少？專案融資是否可以滿足大部分資本需求，我們將不勝感激您的想法。

Yes, sure. So on the data center business, and we'll go through this in great detail on Thursday. But we -- I think a lot of people have seen. So we redesigned the data center with the knowledge we have now. So what we built in Jamestown, which was the initial build that we talked about, it was single-level horizontal because that's the cheapest way to do it. And we have plenty of land in North Dakota, and so it's going to be just build it as far as you wanted to horizontally.

是的，當然。關於資料中心業務，我們將在周四詳細討論這一點。但我們——我想很多人都看到了。因此，我們利用現有的知識重新設計了資料中心。所以我們在詹姆斯敦建造的，這是我們談論的最初的建築，它是單層水平的，因為這是最便宜的方法。我們在北達科他州有足夠的土地，所以只要你想水平建造就可以了。

But now that we know how these workloads work and the necessity of them being all much higher density closer to the network core. We redesigned our structure to a 3-story structure that has network core that runs through the middle so that you can get much more density for -- specifically for training and then somewhat for inference. But with that redesign, you should be thinking, we talked before about kind of $4 million, $4.5 million, we're looking more around $6 million per megawatt to build those facilities.

但現在我們知道這些工作負載是如何運作的，以及它們必須以更高的密度靠近網路核心。我們將結構重新設計為 3 層結構，其網路核心貫穿中間，這樣您就可以獲得更高的密度——特別是用於訓練，然後用於推理。但透過重新設計，你應該想，我們之前談到大約 400 萬美元、450 萬美元，我們正在尋找每兆瓦 600 萬美元左右的資金來建造這些設施。

We think with the work that we have done that we can get roughly an 80% loan to cost for construction, this is -- looks much more like data centers, so we get -- secure that tenant or tenants that are credit-rated, then we can put the construction financing in place. And then there's an equity component, but we call it site level equity and you can get financing partners on the equity component of that as well.

我們認為，透過我們所做的工作，我們可以獲得大約 80% 的建築成本貸款，這看起來更像是資料中心，所以我們得到了——確保租戶獲得信用評級，然後我們就可以落實建設融資。然後還有股權部分，但我們稱之為站點級股權，您也可以就其中的股權部分獲得融資合作夥伴。

That we're -- we've spent a lot of time talking to there, and you should -- it's usually in the data center industry is called the equity financing, again, it's at a site level, but I think of it more as what we would see in the capital markets as like Mezz debt. So we're working through all those pieces of the capital structure to finalize and start the initial build in North Dakota. I guess technically, Lucas, not our initial build. It's the initial build of this design.

我們花了很多時間與那裡交談，你應該——它通常在數據中心行業被稱為股權融資，同樣，它是在站點級別，但我認為它更多就像我們在資本市場上看到的夾層債務一樣。因此，我們正在研究資本結構的所有這些部分，以最終確定並開始在北達科他州的初步建設。我想從技術上講，盧卡斯，不是我們最初的構建。這是該設計的初始版本。

Understood. I'll try to squeeze one more in. In terms of the 34,000 GPUs today versus 26,000 GPUs previously, would those be related to the additional two customers you announced or -- are there other moving pieces there?

明白了。我會嘗試再擠進去一個。就今天的 34,000 個 GPU 與之前的 26,000 個 GPU 而言，這些是否與您宣布的另外兩個客戶有關，或者 - 是否還有其他移動部件？

There's other moving pieces. It's really more related to the demand we continue to see. We haven't seen demand in this area slow down. I think if you go look through the capital raising in the industry hasn't slowed down. There was a large deal, I think, announced with Anthropic, was that last week with AWS. But we're still seeing a robust funding environment for the companies that will be customers for us we see good demand, really strong demand in our pipeline. And so it's really just a decision based on what we're seeing from a demand perspective.

還有其他活動。這確實與我們繼續看到的需求更相關。我們沒有看到該領域的需求放緩。我認為，如果你仔細觀察一下，該行業的融資並沒有放緩。我認為，上週與 Anthropic 宣布的一項重大交易就是與 AWS 的交易。但對於那些將成為我們客戶的公司來說，我們仍然看到了一個強勁的融資環境，我們看到了良好的需求，我們的管道中的需求非常強勁。因此，這實際上只是基於我們從需求角度看到的情況而做出的決定。

Okay. Very helpful. I appreciate it. And to you and the team, best of luck.

好的。很有幫助。我很感激。祝您和團隊好運。

Thanks, Lucas.

謝謝，盧卡斯。

Our next question comes from the line of Darren Aftahi with ROTH MKM.

我們的下一個問題來自 Darren Aftahi 和 ROTH MKM 的對話。

Wes, could you speak a little bit more to the cadence on the GPU orders? I'm just kind of curious if you can talk month-to-month in September and in October. Are those numbers getting bigger relative to the initial 1,000 order? And then I guess, why is the cadence kind of hockey ticked up into November, December and January for your comments. Is that just a function of backlog? And I guess what's your confidence that, that number isn't kind of pushed up further into '24?

Wes，可以多談談 GPU 訂單的節奏嗎？我只是有點好奇你是否可以在 9 月和 10 月逐月討論。相對於最初的 1,000 個訂單，這些數字是否越來越大？然後我想，為什麼曲棍球的節奏會在 11 月、12 月和 1 月進行，以徵求您的評論。這只是積壓的功能嗎？我猜你對這個數字不會再推高到 24 年有什麼信心？

Yes. So it's a good question, Darren. So what we saw in September is our second 1024 cluster, 1,024 GPUs being delivered. And in October, we could see that be doubled and then get significantly bigger in November and December and into January. And it's just the our order book that we put in back in the May time frame, starting to be delivered, you get small pieces. And then it's the schedule that we've been given as far as when deliveries will happen really between now and the end of the calendar year.

是的。所以這是一個很好的問題，達倫。所以我們在 9 月看到的是我們的第二個 1024 集群，1,024 個 GPU 正在交付。在 10 月份，我們可以看到這一數字翻了一番，然後在 11 月、12 月和 1 月份顯著增加。這只是我們在五月的時間範圍內放入的訂單簿，開始交付，你會得到小件。然後是我們收到的時間表，從現在到年底之間真正進行交付的時間。

So that's what gives us that confidence in the deliveries, can things be pushed. I mean that it could always happen, I suppose, but it seems that we've -- this is probably trying to think when it was maybe 3 weeks ago, we're given or a little bit longer than that, kind of a firmer delivery schedule from our suppliers. So feel pretty good about the deliveries for us. And I think it's just a matter of when we ordered those and when they're being shipped out. And kind of the cadence of that ramping up over the next 3 months.

這就是讓我們對交付充滿信心的原因，事情是否可以推進。我的意思是，我想這總是可能發生，但似乎我們已經——這可能是在嘗試思考，當它可能是三週前時，我們已經給出了或比這更長一點，有點更堅定我們的供應商的交貨時間表。所以對我們的交付感到非常滿意。我認為這只是我們何時訂購以及何時發貨的問題。在接下來的三個月裡，這種節奏會加快。

I don't know. I don't think it's for me to say whether there's better supply in the industry or not or if it's just related specifically to us that I don't know.

我不知道。我不認為我應該說這個行業是否有更好的供應，或者它是否只是與我們具體相關，我不知道。

That's helpful. And then just one last one for me. On the anchor tenant with HBC, you have multiple tenants you're talking to? And I guess, in terms of slotting people in, I mean, how close are we to North Dakota vis-a-vis Utah? And then I assume your data center financing is probably going to be right behind that based on kind of your prior comments. So if you just kind of walk through -- I'm just trying to understand kind of anchor tenant demand relative to the capacity you have and then kind of a longer-term plan.

這很有幫助。然後是我的最後一件事。在 HBC 的主力租戶中，您有多個正在交談的租戶嗎？我想，就安置人員而言，我的意思是，我們離北達科他州與猶他州有多近？然後，根據您先前的評論，我認為您的資料中心融資可能會緊隨其後。因此，如果你只是走一走——我只是想了解相對於你擁有的容量的主要租戶需求，然後是一種長期計劃。

Sure. So we're seeing a lot of interest. We've been having these conversations for a few months now. And then in the first week, the end of the first week of September, we finalized our design. And so we've been in a, what I would call, a formal marketing process from that point. until now, and we'll move that into an LOI stage and into a contracting stage over the next few weeks or a month or so that's the expectation, but that's kind of how it's come together as it really started a formal process in, call it, mid-September, and we hope to conclude that in the coming weeks.

當然。所以我們看到了很多興趣。這些對話我們已經進行了幾個月了。然後在第一周，九月的第一周結束時，我們完成了我們的設計。因此，從那時起我們就進入了我所說的正式行銷流程。到目前為止，我們將在接下來的幾週或一個月左右將其進入意向書階段並進入合約階段，這就是我們的預期，但這就是它的整合方式，因為它真正啟動了一個正式流程，請致電九月中旬，我們希望在未來幾週內得出結論。

But the interest is high, and we'll go through this in more detail on Thursday. But it's -- the demand for this style of data center with this type of density because we've been both in out getting colocation space for ourselves on the cloud side and then talking to potential customers on the data center side for building our own. We're seeing a huge amount of demand and really demand for power that is available in capacity that can come online over the next 24 or 36 months.

但人們的興趣很高，我們將在周四更詳細地討論這個問題。但這是對這種密度的這種類型的資料中心的需求，因為我們一直在為自己在雲端獲得託管空間，然後與資料中心端的潛在客戶交談以建立我們自己的資料中心。我們看到了大量的需求，而且是對未來 24 或 36 個月內可以上線的電力的實際需求。

Next question is from the line of Mike Grondahl with Northland Securities.

下一個問題來自 Northland Securities 的 Mike Grondahl。

A couple of questions. The first on potential anchor tenants, would you say the potential anchor that you're going to announce near term, is that still sort of wide open, meaning there's still multiple five, six, seven big potential anchors out there you're talking to? Or have one or two kind of made it way down the funnel and you're just kind of finalizing who it might be out of a very small group?

有幾個問題。第一個關於潛在主力租戶的問題，您是否會說您近期要宣布的潛在主力仍然處於開放狀態，這意味著仍然有多個五、六、七個大的潛在主力正在與您交談？或者有一兩種人進入了漏斗，而你只是從一個很小的群體中最終確定了可能是誰？

So Mike, I would say that we're right in the, I would say, in the middle of the kind of the two scenarios that you described. So the funnel has gotten smaller, but we're not right at the end yet. Does that makes sense.

所以麥克，我想說，我們正處於您所描述的兩種情況的中間。所以漏斗已經變小了，但我們還沒到終點。這樣有道理嗎？

Got it. Yes. Just trying to understand. And then on the prepayments, you mentioned $39.5 million in the August quarter. And then I think a $15 million prepayment and a $23 million prepayment you expect this week. Can you say are the $15 million in the $23 million, are those from customers three and four? Or do those relate back to customer one and two?

知道了。是的。只是想了解一下。關於預付款，您提到 8 月季度的預付款為 3,950 萬美元。然後我認為本週您預計將預付款 1500 萬美元和 2300 萬美元。您能說 2300 萬美元中的 1500 萬美元是來自第三個還是第四個客戶的嗎？或是這些與客戶一和客戶二有關嗎？

It's both for those prepayments. The ones that we talked about that we've received already and then the one we expect to receive this week.

這都是預付款。我們討論過的我們已經收到的，然後是我們預計本週收到的。

Got it. Got it. And then lastly, the 1,024 GPUs in August that you put to work, did they all go to customer one?

知道了。知道了。最後，您在 8 月投入使用的 1,024 個 GPU 是否全部都交給了客戶 1？

So those came in September, Mike. And they are for customer one.

所以這些是在九月到來的，麥克。它們是為客戶一準備的。

Our next question is from the line of Kevin Dede with H.C. Wainwright.

我們的下一個問題來自 Kevin Dede 和 H.C.溫賴特。

Maybe you could just help me understand a little bit better how you're thinking about AI cloud versus AI host and how that might figure in your calculus your construction calculus.

也許你可以幫助我更好地理解你如何看待人工智慧雲端與人工智慧主機，以及它們如何在你的微積分和建立微積分中發揮作用。

In what way, Kevin?

凱文，用什麼方式？

Well, I guess what I'm wondering is when you go look for your anchor tenants, are you looking at them purely from a cloud customer perspective? Or are you looking at some perhaps from a host perspective, where they're bringing her own GPUs.

好吧，我想我想知道的是，當您尋找主力租戶時，您是否純粹從雲端客戶的角度來看待他們？或者你是否從主機的角度來看一些，他們帶來了自己的 GPU。

Yes, yes. So the anchor tenants are absolutely a host hosting business for us. where they will bring their own equipment. It's just a data center hosting business for us. When you think about anchor tenants and for these data centers, we've talked about the ideas, 70% goes to the colocation hosting style business and then we keep 30% for our own cloud capacity.

是的是的。因此，主力租戶對我們來說絕對是主機託管業務。他們將攜帶自己的設備。對我們來說這只是一個資料中心託管業務。當你考慮主力租戶和這些資料中心時，我們已經討論過這些想法，70% 用於託管託管類型業務，然後我們保留 30% 用於我們自己的雲端容量。

Then the you said you're comfortable in colocation that you need for -- to support the 26,000 to 34,000 GPUs you have coming in, how do you figure moving them once they're at their colocation to your own facilities once they're ready.

然後，您說您對所需的主機託管感到滿意——支持您擁有的 26,000 至 34,000 個 GPU，一旦它們準備就緒，您如何考慮將它們遷移到您自己的設施。

So we won't ever move those installations. And the way that I think about it from a cloud service perspective, is a lot of that colocation is in what I would call kind of traditional cloud regions. And as we ramp up capacity in North Dakota, it's really being built for very large training clusters. And so I think about the training portion of the business that's being deployed in smaller training clusters in these cloud regions moving over to the infrastructure that gets put in place in North Dakota and then using the smaller clusters that we're building out now is inferencing, which I think the way this market splits is there will be training batch inference, and some inferencing done in these large facilities like we're building in North Dakota.

所以我們永遠不會移動這些設施。我從雲端服務的角度來看，很多託管都在我所說的傳統雲端區域。當我們提高北達科他州的容量時，它實際上是為非常大的訓練集群而建造的。因此，我認為業務的培訓部分部署在這些雲端區域的較小培訓集群中，轉移到北達科他州部署的基礎設施，然後使用我們現在正在構建的較小集群進行推斷，我認為這個市場分裂的方式是訓練批量推理，以及在這些大型設施中完成一些推理，例如我們在北達科他州建造的設施。

And then a lot of the inference portion of the market will be more in what I would call traditional cloud regions. And so it works well for our cloud business over time to have that type of architecture of just being spread into more cloud regions for the inferencing portion of the business, while a lot of the training will move into North Dakota.

然後，市場的許多推理部分將更多地集中在我所說的傳統雲端區域。因此，隨著時間的推移，將這種類型的架構擴展到更多雲端區域以用於業務的推理部分，這對於我們的雲端業務來說效果很好，而大量培訓將轉移到北達科他州。

Okay. The redesign that you did in September, Wes for your new Ellendale facility, did you have to rethink latency? Or did you have to be more concerned about power backup on those? Or do you still think you can operate under the conditions that you built for in Jamestown?

好的。 Wes，您在 9 月為您的新 Ellendale 設施進行了重新設計，您是否需要重新考慮延遲？或者您需要更關心這些設備的電源備援？或者你仍然認為你可以在詹姆斯敦建立的條件下運作嗎？

So it's not about latency or necessarily power backup. It's about -- it's a redesigned specifically for density. So it's designed to be able to basically take a network core and go through multiple floors. And put all of the GPU. So the design here for the new North Dakota facility, it's really around kind of this magic number of, I call it, 30 meters being the magic number of how far -- what rack can be away from the network core and still be in the same cluster, same spine for networking.

因此，這與延遲或備用電源無關。它是關於——它是專門針對密度而重新設計的。所以它的設計基本上能夠佔據一個網路核心並穿過多個樓層。並且把所有的GPU。因此，北達科他州新設施的設計實際上圍繞著這個神奇的數字，我稱之為 30 米，這是一個神奇的數字，即什麼機架可以遠離網絡核心，但仍然位於網絡中。相同的集群，相同的網路骨幹。

You technically, Kevin can go out to 50 meters away from the network core, but you have to use a single mode transceiver on optics instead of multimode, so it gets more expensive, a little more difficult. So it's really designed around how many racks and how many servers can we get within 30 meters of the network core. And so that's why it goes through multiple levels in the building instead of the single level. That's the primary piece of the redesign.

從技術上講，Kevin 可以到達距離網路核心 50 公尺的地方，但您必須使用光學上的單模收發器而不是多模，因此它會變得更昂貴，也更困難。因此，它實際上是圍繞我們在網路核心 30 公尺範圍內可以容納多少個機架和多少台伺服器而設計的。這就是為什麼它在建築物中經歷多個樓層而不是單一樓層。這是重新設計的主要部分。

And what are the heat implications, though, if you've got multiple racks stacked, right, and heat's tendency to rise?

但是，如果堆疊了多個機架，並且熱量有上升的趨勢，那麼對熱量的影響是什麼？

Yes. So this is a liquid cool facility. That's the other part of the redesign. So it was air cool. You can effectively do air cool, our opinion up to, call it, roughly 50 kW per rack. But as you start to go above that, you really need to move to a liquid cool solution. And so the new facility is designed for liquid cool. The new facility will go to 150 kW per rack. It can still -- it still has the floor space to do it at 45 kW, which is what we're doing in Jamestown, but we've -- this will be built for liquid cooling.

是的。所以這是一個液體冷卻設施。這是重新設計的另一部分。所以空氣很涼爽。您可以有效地進行空氣冷卻，我們認為每個機架約 50 kW。但當您開始超越這一點時，您確實需要轉向液體冷卻解決方案。因此，新設施專為液體冷卻而設計。新設施的每個機架功率將達到 150 kW。它仍然可以——它仍然有足夠的佔地面積來實現 45 kW 的功率，這就是我們在詹姆斯敦所做的事情，但我們——這將是為液體冷卻而建造的。

(Operator Instructions) At this time, this concludes our question-and-answer session. I'd now like to turn the call back over to Wes Cummins.

（操作員說明） 至此，我們的問答環節就結束了。我現在想把電話轉回韋斯康明斯。

Thank you, operator, and thanks, everyone, for joining our call. I look forward to speaking with you on Thursday at our Investor Day, which will be held in Midtown Manhattan. And again, thanks to all of our employees for the efforts they put in the last quarter and this quarter to date. I look forward to speaking with you on our next quarterly call.

謝謝接線員，也謝謝大家加入我們的通話。我期待著週四在曼哈頓中城舉行的投資者日與您交談。再次感謝我們所有員工在上個季度和本季迄今所付出的努力。我期待在下一次季度電話會議上與您交談。

Thank you for joining us today for Applied Digital's conference call. You may now disconnect.

感謝您今天參加 Applied Digital 的電話會議。您現在可以斷開連線。