The Nvidia-backed IPO grew 737% last year and is about to be made public. What investors need to know about CoreWeave

Although there have been no many early public offerings (IPOs) in recent years, not to mention interesting and well-known companies that could take on top tech stocks in the market. However, one company in the middle of the AI ​​revolution may be about to make its public debut. Not only is this inventory growing at an impressive level, but it also counts nvidia (NVDA) 1.92%)) As a large investor and Microsoft (msft -0.90%)) As a huge customer.

Here’s what you need to know about CoreWeave before your IPO:

“Neocloud” supported by Nvidia

CoreWeave, who will be listing it on the Nasdaq under the symbol CRWV, finds himself at the heart of the AI ​​revolution. However, the company didn’t actually start that way. In 2017, CoreWeave was founded by three executives from Hudson Ridge Asset Management, a natural gas-centric hedge fund with its first mission to mining Cryptocurrencies.

The cryptocurrency experience has refined CoreWeave’s skills in deploying Nvidia Graphic Processing Units (GPUs) that were used to mine crypto and used to manage energy-intensive computing clusters. Additionally, these skill sets have proven to be extremely important in AI computing.

In 2020, the company pivoted to build the CoreWeave Cloud Platform, and in April 2023 Nvidia invested in the company. Today, AI Chip Giant owns only 5% of its stock. Additionally, NVIDIA is a customer and may use CoreWeave to run software products and possibly test AI applications.

Nvidia’s investments have come at a very interesting time. As you may remember, May 2023 was the first blown Nvidia revenue report, leading to the overgrowth stage of AI buildouts.

Competitive Benefits of CoreWeave

Some people wonder what CoreWeave offers, which sets it apart from other major cloud infrastructure platforms. Reading Form S-1, it appears that the company is doing some very well.

One of the key points that CoreWeave highlights is that its cluster is built from the ground up as an A-Optimized GPU cluster. This contrasts with generalized cloud platforms that require building both AI and traditional cloud infrastructure across their footprint.

CoreWeave believes there are important benefits to running these clusters through its proprietary orchestration and observability software that allows for more efficient use of GPUs. AI GPU workloads are large, complex and computationally intense, making it difficult to efficiently coordinate the entire data center for multiple clients.

To put some numbers on it, CoreWeave explains a metric called Model Flops (MFLOPS) Utilization. This basically measures the utilization of AI clusters compared to its total theoretical computing power. The industry average may surprise you. Due to the complexity of AI, CoreWeave says that the typical MFLOPS rate is only 35% to 45% across the industry. The gap between actual and theoretical performance is an opportunity to believe that CoreWeave can narrow it down, primarily through some new software innovations.

One such software innovation is Sonk, a software platform that combines Kubernetes with the Slurm open source software platform. Kubernetes is a platform for containerized workloads in traditional cloud environments, and AI customers are also used to serve the model. Meanwhile, Slurm is a popular open source software that coordinates large parallel computing workloads for AI training.

Traditionally, customers had to choose between the two for each computing cluster. However, using CoreWeave’s Sunk Platform allows SluRM to run within Kubernetes, allowing developers to use both. This increases the efficiency of calculation utilization.

The second software innovation is CoreWeave’s tensorariser, optimization software for inference and training. For inference, the tensors can route models in storage to optimally close GPU nodes for clients. According to data cited in the S-1, this results in faster load times than rival Hug Face and Safetensors. For training, tensors can reduce training time through similar efficiency optimizations.

In addition to software, CoreWeave may have more advantages to the market than other clouds. Because of Nvidia’s investment, CoreWeave could be the front line of the line, or very close, for the latest and largest Nvidia GPUs. On the S-1, CoreWeave has been in the market for the NVIDIA H100 and H200 systems, and in fact, NVIDIA GB200 NVL72-based instances are generally the first cloud available recently.

Image source: Getty Images.

Finance shows super growth

Of course, there’s nothing more to say about the positive side of CoreWeave than its finances. As you can see, the company has seen explosive growth over the past two years.

Data Source: CoreWeave S-1.

The big jump in 2024 was a 737% growth rate, an incredibly high rate for so-called startups. The flip to profitability of operations is certainly promising. Notably, the results over the past year have been based on 32 deployment data centers hosting approximately 250,000 GPUs.

It is unclear what valuations CoreWeave has decided to publish, but certain analysts estimate they are trying to raise between $3.5 billion and $4 billion with a market capitalization of $32 billion. Therefore, the stock trades with approximately 16 times subsequent revenues and 100 times operating profits. But before the IPO, CoreWeave had $7.9 billion in debt and $1.4 billion in cash, making it a bit more expensive on a corporate value basis.

The risk may be high enough to distress some

At first, CoreWeave might look like the next big AI juggernaut. Its valuation is superficially high, but considering the company’s current growth rate and the long-term growth potential of its generation AI, the stock price valuation is not that expensive.

However, the structure of its growth may raise questions. In 2024, 62% of CoreWeave’s revenue came from just one company, Microsoft. Microsoft complements the Azure cloud by renting spare GPU capacity from CoreWeave, but also spends tens of thousands of dollars a year on cloud infrastructure.

You might ask why Microsoft is such a huge customer when other major clouds, Amazon, and Alphabet, are not listed as major CoreWeave customers. This may be because Amazon and Alphabet have a rather mature custom ASIC program itself. Alphabet designed its own tensor processing unit chip in 2015, and Amazon unveiled the Irsentia chip in 2019 and the Trainium AI chip in 2021.

Microsoft was late to custom AI chip games, but about a year ago, in November 2023, they unveiled the Maia AI chip. It is unclear whether the lack of custom ASIC is the complete reason for Microsoft’s use of CoreWeave. After all, Microsoft might appreciate CoreWeave’s ability to run data centers for another reason. But that’s also an important part of it.

So, if Microsoft has upgraded the game and Maia matures to the level of Google TPU or Amazon’s AI chips, Microsoft may use less CoreWeave’s infrastructure.

Without a doubt, Nvidia GPUs are still in demand today and will likely be in the future. However, since Alphabet and Google can supplement certain workloads with their own chips, Microsoft’s own chip scaling could free up a lot of dollars to buy Nvidia chips directly for their own data centers.

Don’t forget that cloud companies can purchase custom ASICs at casting prices, but Nvidia has a total margin in the 70% medium term range. So basically it’s 3-5 times more expensive to buy an Nvidia GPU than to design your own chip and buy it directly from the foundry.

Furthermore, thanks to Nvidia’s investment, part of CoreWeave’s appeal may be early access to the most advanced Nvidia chips. Therefore, the fate of Coreweave appears to be very closely linked to Nvidia’s future progress.

Of course, it’s a great place right now. But if AI buildouts are slow, or something happens in Nvidia’s competitive position, it will have a big impact on CoreWeave as well.

CoreWeave is a fascinating company

In a volatile market, CoreWeave could prove to be an unstable and controversial inventory once it’s published. The filing shows justification for investors to buy into an IPO, but also shows some major risks that will at least put this investor in an announcement of an IPO.