CoreWeave IPO: The Next Big AI Player Or A Ticking Time Bomb?

CoreWeave was the first big tech IPO to happen during the current AI craze on Wall Street. An opportunity for private investors?

With CoreWeave ($CRWV), on 28th March 2025 a larger tech company has finally managed to go public again after a long dry spell. The data center operator, which specializes in AI workloads, has seen its revenues explode 120-fold in just three years, from a meager $16 million in 2022 to $1.9 billion. Its last round of funding took place just 4 months ago, in December 2024, at a valuation of $23 billion.

Just a few weeks ago, CoreWeave announced that it had signed OpenAI as a major customer, the largest deal in the company's history. OpenAI will pay CoreWeave up to $11.9 billion over the next 5 years for its AI services. In addition, OpenAI will invest $350 million in CoreWeave.

CoreWeave's revenue is expected to more than double to $4.6 billion in 2025, even though the OpenAI deal is not yet contributing to revenue, and another triple-digit growth to $10.8 billion is expected in 2026 thanks to the OpenAI mega-deal.

And now CoreWeave, a new AI hyperscaler often touted as "one of the hottest players in the AI space," is going public. And it didn’t go too well. The IPO was expected to raise up to $2.7 billion and there were rumors of a $35 billion valuation. But at the end of the day they raised just $1,5 billion at an issue price of only $40 and the valuation came in at around $23 billion (fully diluted).

Sounds like a great AI company with a future, and maybe even a good deal for private investors?

Well, yes.

After all, CoreWeave is the first major AI company to go public after OpenAI's breakthrough with the release of ChatGPT in 2022.

So let's take a look at the company:

The History Of CoreWeave

CoreWeave was founded in 2017 as a crypto mining company and, more or less by accident, owned a large amount of the coveted Nvidia GPUs at the beginning of the AI boom, which it had originally purchased for Ethereum mining. After the launch of ChatGPT at the end of 2022, they were able to lease them at scarcity prices to Meta, IBM, Microsoft and others for much more lucrative prices. Thus, in 2023, the company pivoted from being an Ethereum miner to renting out AI servers.

CoreWeave seized the opportunity and made a massive purchase from Nvidia, ordering billions of GPUs in bulk. At the same time, Nvidia took a 6% stake in CoreWeave for $100 million in April 2023. This intense partnership with Nvidia made CoreWeave a preferred customer, and today it has an impressive fleet of more than 250,000 Nvidia GPUs across 32 data centers in the US and Europe. This infrastructure positions CoreWeave as one of the leading providers of GPU-based cloud services, particularly for artificial intelligence applications.

But 62% of the record revenue in 2024 came from the major customer, Microsoft. This is a company that is also a competitor with its Azure services.

Another customer, which CoreWeave has not officially named, accounts for another 15% of revenue in 2024. This means that CoreWeave recently generated more than three-quarters of its revenue from just two large customers.

The Information reported that this second major customer was Nvidia. The chip giant put $100 million into the startup in early 2023. This stake is now worth more than $700 million at the IPO price. It sent hundreds of thousands of high-end graphics processing units to CoreWeave. And they agreed to rent back their own chips from CoreWeave for $1.3 billion over four years, through August 2027. These kinds of financial deals are a great example of how money for AI often goes back and forth between the same places.

In any case, this high customer concentration carries significant risks. If Microsoft were to terminate or reduce its contracts, the financial consequences could be severe. Microsoft has reportedly already discontinued some of CoreWeave's services due to delivery problems and delays, adding to the uncertainty. However, CoreWeave has denied these reports.

Financing On Shaky Ground

CoreWeave has pursued an aggressive expansion strategy in recent years, incurring significant debt in the process. Total debt amounts to approximately $8 billion. This debt was incurred to expand its GPU provisioning capacity as demand for computing power exploded, particularly from Microsoft and OpenAI.

This investment spree has left CoreWeave with a negative cash flow of $6 billion in 2024, more than three times its revenue. According to another article in The Information, the company is expected to have a cash burn of $15 billion in 2025. This means that the proceeds from the IPO will likely not be enough, and that more debt will be needed to fund the company in the short term.

CoreWeave's business strategy seems very risky, as CoreWeave's future revenues may not be sufficient to service the interest, principal, and rent payments for the large data centers. A significant drop in demand or the termination of large contracts could well lead to insolvency.

Dubious Partner

The partnership between CoreWeave and Core Scientific, a company founded in 2017 that originally worked in bitcoin mining and had to file for bankruptcy in 2022 after a disastrous SPAC IPO in the same year, is also noteworthy.

Meanwhile, Core Scientific is back on the Nasdaq after a financial restructuring. Although Core Scientific has no proven expertise in high performance computing (HPC) to date, CoreWeave plans to deliver over 500 MW of capacity with its partner in the coming years. CoreWeave's planned 500 MW capacity is equivalent to the combined power of 5 to 10 large hyperscale data centers operated by companies such as Google or Amazon.

The bankruptcy of Core Scientific and the primary focus on crypto mining raises the question of whether the company has the expertise, financial resources, and infrastructure to implement CoreWeave's ambitious plans. Building the planned data centers requires significant investment and expertise in handling large amounts of data. If the partnership with Core Scientific were to fail, it could not only impede expansion, but also jeopardize CoreWeave's entire business model.

Do We Need More Red Flags?

The founders of CoreWeave have already sold nearly $500 million worth of stocks prior to the IPO. This is highly unusual and reminiscent of the founder of WeWork, who raised similar sums before going public. What followed was a failed IPO, a backdoor IPO two years later (via SPAC), and another two years after that, WeWork was insolvent. I'm interested to see how CoreWeave's valuation will develop in the coming weeks and months after their underwhelming IPO.

The CoreWeave IPO is an acid test for the AI industry. I think it was a very smart move by Microsoft to rely on CoreWeave, at least for a while, to build AI datacenter capacity. Not only did it allow them to bypass the GPU bottlenecks and deliver faster, but it also gave them a little more flexibility with their already skyrocketing cap-ex.

Now, depending on how AI demand evolves, Microsoft can either continue to grow its own AI capacity, or reduce CoreWeave's capacity relatively quietly. I think I can see the latter happening. Although CoreWeave has denied terminating existing contracts, it is now being reported that Microsoft has let a $12 billion option to lease additional capacity from CoreWeave lapse. To be honest, that doesn't make much difference to me.

Even though the Nvidia CEO keeps repeating like a mantra that the demand for GPU chips will continue to grow dramatically in the coming years: I won't believe it until I see application software vendors doing great business with their new AI capabilities and experiencing a growth spurt. But nine months after Sequoia formulated AI’s $600 billion question, there is still no answer to that.

SaaS companies' expectations for the current fiscal year are muted, and there is no evidence that a broad layer of user companies is now spending large sums of real money on AI software.

Until then, the AI economy will remain a closed loop, with money moving mainly between Nvidia, Microsoft, a few other hyperscalers, OpenAI, SoftBank and the remaining VC-funded AI players. This will continue as long as Nvidia and Microsoft shareholders and venture capitalists continue to fund the whole thing.

The CoreWeave IPO can be seen as a litmus test for the general state of the AI industry. So far, all AI protagonists (especially Nvidia and OpenAI, but also Microsoft) have strongly denied that future spending on AI could be lower than previously assumed. To maintain this narrative, the AI industry needs a sustained successful CoreWeave IPO.

Personally, I think CoreWeave might even be another candidate for a short position in my long/short portfolio. If you want to follow CoreWeave's progress with me, you can subscribe to my free weekly newsletter here:

*Disclaimer: This post is an expression of opinion and not investment advice. The author and/or persons or companies associated with him hold short positions in Nvidia stock (as of March 29th, 2025).