Ocelote
Trading at around $7, Applied Digital Corporation (NASDAQ:APLD) has soared more than 200% since November 2022, outperforming Nvidia (NVDA) last year and nearly catching up earlier this year before dropping back down, as shown in the chart below. This price movement could suggest that some investors are viewing the HPC (high-performance computing) data center business as an alternative to get in on the AI story.
Data by YCharts
Indeed, Applied Digital Corporation’s greenfield approach compared to other data centers means the paper aims to show that the company should play a key role in building the infrastructure needed for AI models that will be trained in optimal conditions on the semiconductor giant’s advanced GPUs.
So while Applied Digital Corporation shares are a buy, you should consider not only the company’s role in the artificial intelligence value chain, but also that it is undervalued, especially compared to Nvidia. First, some thoughts on why the company’s previous role as a blockchain hosting provider makes it particularly well suited for hosting AI workloads.
The move from blockchain to AI data centers
Initially known as Applied Blockchain, the company changed to a new name in November 2022 to coincide with the launch of OpenAI’s ChatGPT, which is at the root of the company’s goal of moving into artificial intelligence. Currently, there are already so many players, such as Equinix (EQIX) and Digital Realty Trust (DLR), just to name the two most important public companies in the industry. And let’s not forget that hyperscalers such as Amazon (AMZN) are strengthening their in-house capabilities. So, some may wonder about the possibility of Applied Digital being able to compete in this space.
Possibly, since Nvidia sells billions of dollars worth of GPUs that need to be installed in data centers as part of the AI value chain. Like CPUs before them, these GPUs are typically intended for data centers. But environmental conditions vary, AI racks can consume twice as much power as traditional equipment, and power-hungry processors require advanced cooling.
One way to address this challenge is to retrofit old data centers with new power bars and cooling systems, but this is not an ideal solution. We recommend thinking carefully before embarking on such a strategy. Moreover, most incumbents are currently concentrated in metropolitan areas such as North Virginia, and some counties are set to impose restrictions on data center development. In this regard, Dominion Energy (D) is seeing a surge in demand due to the need to power electricity-hungry AI workloads, putting pressure on renewable energy as part of its energy mix.
For both of these constraints, Applied Digital appears to have a solution. First, the company, which has a history of acting as a blockchain colocation service provider to find cheaper renewable energy sources across the U.S., has built an AI data center (HPC hosting segment) in North Dakota. Second, it is a greenfield facility, so they won’t have to go through the pain of retrofitting an older facility.
Generating revenue but incurring losses
Additionally, the company owns existing brick-and-mortar facilities (its data center hosting division), which generated the majority of its revenue, $37.8 million, of its $43.3 million in the third quarter (FQ3) of 2024, which ended in February. The remainder is made up of its cloud services division, with HPC hosting not yet generating any revenue.
But as you can see below, the company is loss-making, with losses compounded by a $4.5 million revenue shortfall caused by a power outage at its 180MW datacenter hosting business in January.
SEC filings (seekingalpha.com)
Additionally, there was a $21.7 million loss related to its blockchain mining facility in Garden City, classified as “held for sale.” The site was sold to Marathon Digital (MARA) for $87.3 million (net purchase price) as part of its diversification from Bitcoin into HPC, with the transaction expected to close in the second quarter of 2024. The sale also entitles the company to receive $12 million in restricted cash related to a secured loan related to the site.
Significant investment in the fast-growing Gen AI market
This also represents an inflow of cash to its balance sheet, which had net debt of $290 million at the end of the first quarter. However, the company needs more capital to pursue its growth initiatives in cloud for HPC, which is why it secured a $200 million private debt facility, which includes an initial commitment of $125 million and $15 million already drawn.
Now, depending on whether or not they issue equity, this method of borrowing would only increase their debt-to-equity ratio to 248%, which is much higher than other companies that operate data center facilities, as you can see below.
Comparison with peers APLD, EQIX, DLR, DBRG (seekingalpha.com)
Digging deeper, for those familiar with investing in data center-focused REITs like Equinix and Digital Realty Trust, these REITs have been in the industry for many years and generate relatively high and stable cash flows from operations. For comparison, we’ve also included DigitalBridge (DBRG), which has investments in both cell towers and data centers.
Currently, only Applied Digital has a negative FCF, but this should change, as we’ll explain in more detail later. Furthermore, looking at the trends in the chart below, with the exception of DLR, free cash flow has been trending downward since the start of 2023, indicating investments in AI.
Data by YCharts
Valuing the company
Given the high investment required, Applied Digital’s approach of providing purpose-built infrastructure to capitalize on the Gen AI market, which is expected to grow by $20 billion per year between 2023 and 2030, makes sense.
Digging deeper into cash, the company expects to be FCF positive this year due to the previously mentioned $200 million secured mortgage loan dedicated to HPC at the Ellendale facility (previously funded internally). Shortly after the June 7 financing deal, the stock price rose about 50%, which may indicate market confidence in an operating model based on asset-level financing.
Despite the uptrend, the company trades at a 6% discount to the IT sector, trading at 18.92x trailing-year price-to-cash flow. However, from a strategic perspective, the company deserves a higher valuation. Specifically, as shown below, it is positioning itself as one of the first cloud service providers to use GPUs based on Nvidia’s Blackwell platform for HPC applications. This approach, which buys compute directly from Nvidia rather than just prepaying for power and cooling, differentiates it from other data center companies, whose strategy is to install GPUs on behalf of their customers.
So for investors, Applied Digital is an alternative to gain exposure to the AI story as opposed to Nvidia, which is trading at a historical price-to-cash flow of 49.13x. Therefore, due to its position in the AI infrastructure business, Applied Digital’s target price is 24.6x, or only half that of Nvidia, making it likely undervalued by about 30% or (24.6-18.92)/18.92. A 30% increase on the current share price of $5.67 would give a price target of $7.4.
ir.applieddigital.com
However, this is a long-term goal due to supply chain constraints and there have been some short-term setbacks that need to be highlighted.
Risks associated with delays
First, North Dakota is far from major supply chains on the East and West coasts, and it’s experiencing a surge in AI-driven demand for items like transformers used to electrify data centers. So while the energy source is there, it can take time to convert it into electricity and power the computing equipment. Second, building out an AI infrastructure requires specific equipment when deploying NVIDIA’s complete cluster solution, which can cause delays. This was the case when transceivers for InfiniBand networking, a proprietary networking standard for interconnecting GPU clusters, didn’t arrive in time, delaying the project’s delivery by eight weeks.
Additionally, whereas existing data centers already have the power in their facilities and can record revenue as soon as their customers’ racks are powered up, this is not the case at Applied Digital due to the upfront costs and costs of bringing the facility’s power online. As a result, costs were eating into profits.
This was the case in Q3, where start-up costs for the Cloud Services business were primarily responsible for $30.4 million in additional operating expenses, up approximately 200% year-over-year. So, depending on execution, we expect the same for the HPC business. Additionally, interest expenses related to the $200 million loan will further increase and impact the income statement, so unless revenues grow sharply, operating income may continue to trend below zero, as shown in the chart below.
Chart created using data from (seekingalpha.com).
GPUs come online, revenue surges, but risk of instability
Well, this is possible considering the eight H100 GPU clusters, which, when fully deployed, could generate $20 million in revenue per cluster per year. Each cluster has around 1,000 GPUs, with 4,000 already in revenue generating mode in April, and the number is expected to grow to 6,000-8,000 by the end of May. Now, with all 8,000 GPUs in the eight clusters online, the company could earn $160 million in the HPC segment, thereby boosting its overall revenue and causing the orange graph above to spike.
These revenues are primarily coming from startups with the intention of diversifying to larger enterprise customers. Diversifying the customer base to include larger enterprises is important to make revenue more predictable. We believe that this can be achieved this year, as IT spending is expected to grow 8% from 2023, according to Gartner. More importantly, the portion of spending on data centers is expected to accelerate by 10% compared to 4% in 2023, driven by Generative AI.
Finally, for those looking for an alternative to Nvidia, Applied Digital as an AI infrastructure startup has the potential to see its shares rise by 30%. However, it is a highly leveraged and interest-rate sensitive stock, which was volatile on Tuesday, losing nearly 1.6% in value. This comes after the chairman of the US central bank acknowledged progress in fighting inflation but reiterated his 2% target, meaning interest rates could remain elevated for an extended period of time, or even above 5%. Therefore, investors in Applied Digital Corporation shares should be able to tolerate the risk of volatility.
