CoreWeave (CRWV -2.30%) is making some big moves. Recently, it signed a deal to rent out $14 billion of its computing capacity to Meta Platforms (META -2.29%). It also expanded its agreement with OpenAI, the makers of ChatGPT, by $6.5 billion. These monster deals are pushing the stock higher, and CoreWeave's stock is up almost 34% since September started.
CoreWeave's business goal is to become the artificial intelligence (AI) cloud king. To do that, it will need to buy a lot more computing capacity, including chipsets from Nvidia (NVDA -0.77%). That leads to the question: Is it better to buy CoreWeave stock, or are you better off owning Nvidia? Let's take a look.
Image source: Getty Images.
CoreWeave's servers are filled with Nvidia GPUs
As mentioned above, CoreWeave's goal is to become the go-to cloud computing platform for artificial intelligence computing. This is an attractive business model for CoreWeave and clients alike. While most clients are building out their own computing infrastructure, having some flexibility to run workloads on CoreWeave's servers when demand is higher or not needing to build out AI computing capacity all at once is a smart move. On CoreWeave's side, the business model is fairly simple: Rent out the computing power for more than it costs to replace and operate the equipment.
For Nvidia, its graphics processing units (GPUs) are the most flexible computing units available and are a top option for AI companies to run workloads on. However, custom AI chips are starting to increase the competition that Nvidia has to deal with, but it's still the most commonly used computing unit due to its flexibility. Nvidia sells its GPUs to cloud providers like CoreWeave or its competitors because they have no idea what type of workload their computing units will see. AI hyperscalers can purchase their own custom AI chips because they know what workload will be run across them. This is a key distinction for Nvidia, so it's crucial for its success that CoreWeave and other cloud competitors continue buying up Nvidia's GPUs to meet capacity.
In a way, both companies are critical for each other's success, but there's one option that stands out as a much better pick.
CoreWeave isn't profitable during a once-in-a-lifetime boom
CoreWeave is growing at a rapid rate as demand for AI computing capacity explodes. During its most recent quarter, CoreWeave's revenue rose 207% year over year to $1.2 billion. Considering CoreWeave signed a $14 billion agreement through the end of 2031 with Meta and the total value of the OpenAI contract (which runs through 2029) is now $22.4 billion, there's a massive backlog for CoreWeave to churn through.
However, what remains unclear is what will happen after those contracts expire. Both Meta and OpenAI are building out internal computing capacity to meet AI demand. If they build out enough to satisfy capacity, using CoreWeave as a third-party cloud provider may be unnecessary. This would cause CoreWeave to lose substantial business.
For Nvidia, a complete AI computing capacity buildout may sound scary, but it really isn't. GPUs utilized for AI computing are run incredibly hard and can have lifespans of one to three years. This means that current computing capacity must be replaced every couple of years, which will allow Nvidia to maintain a strong revenue base. CoreWeave has no similar guarantee, and I think that's a key point investors must understand.
Furthermore, right now is about as good as it's going to get for CoreWeave. Demand for AI computing capacity is high, and although it's still growing, it's hard to imagine this growth lasting over the long term, say five to 10 years out. The current problem is that CoreWeave isn't profitable. It posted a net loss margin of 24% during Q2. CoreWeave losing money during arguably its best time to be making profits is a huge red flag, and it makes sense why companies like Meta and OpenAI are renting computing capacity from them; they can likely rent it from CoreWeave for cheaper than they can build it. If CoreWeave increases prices to become profitable, this may no longer be the case. At that point, customers will likely build their own.
This makes the future precarious for CoreWeave, and I'd much rather own Nvidia stock than CoreWeave as a result.
