By Adam Levine
CoreWeave, a rapidly growing AI cloud company, is expected to price its initial public offering later tonight. The stock will open for trading on the Nasdaq Friday morning.
Last week the company disclosed a likely price range of $47 to $55 for the stock, which would give CoreWeave a market value of $27 billion to $32 billion.
The IPO price could change, depending on investor demand for the offering.
CoreWeave is a pure play on AI, with all of its revenue coming from cloud rentals of AI servers that use Nvidia chips.
The IPO and early trading of the stock will offer a dual test: It will gauge enthusiasm for the AI trade, which carried the market in 2023 and 2024, but has flagged so far in 2025. Start-ups and investment bankers will also have a keen interest in the listing as a test of the IPO market, which has been weak since 2022.
CoreWeave was founded in 2017 as a crypto miner that pivoted to supplying AI cloud services after the 2018 crypto bust. When ChatGPT hit in late 2022, it was at the right place at the right time. Before the AI investment boom began, this was a company with $16 million in 2022 revenue. It finished 2024 with sales of $1.9 billion, a 737% annual jump. Given the fast growth, the fourth-quarter may be more illustrative, suggesting a $3 billion annual run rate.
The fast-growing company is already profitable on an operating basis, with a 17% operating margin in 2024. Excluding the benefit of substantial customer deposits, it had $700 million in 2024 operational cash flows, a 37% margin.
CoreWeave has benefited from its close relationship with Nvidia, which is an investor, supplier, and customer, and it was able to get access to Nvidia's latest technology in 2023 and 2024 as it rolled out.
But CoreWeave faces plenty of risk. Two big customers, Microsoft and likely Nvidia, made up 77% of its 2024 revenue, with Microsoft providing the lion's share at 62% of sales. Nvidia uses CoreWeave to do some of its own AI work, but for Microsoft, this is overflow capacity it resells to its Azure cloud unit customers.
CoreWeave scored a third big customer in March, OpenAI, for a contract value of "up to" $11.6 billion, though neither would say what the term of the agreement is. Not including that deal, CoreWeave's backlog stood at $15 billion at the end of 2024, with $8 billion of that coming due in 2025 and 2026.
But the precariousness of the business model goes beyond this customer concentration risk.
In CoreWeave's presentation to potential IPO investors, Chief Financial Officer Nitin Agrawal referred to CoreWeave's "relentless focus on financing," because the operation is built on a mountain of debt that will need to keep elevating. At the end of 2024, CoreWeave had borrowing arrangements for over $12 billion, out of which they had $7.9 billion in loans outstanding. Though its more recent loans are at much lower interest rates, it had a weighted average interest rate of about 12%.
That debt at high rates adds up to $361 million in 2024 interest expenses that more than offset its $324 million operating profit.
Because of its capital-intense business model the company's borrowing will likely grow from here. It had $4.4 billion left to borrow on existing loans at the end of last year, and another $1.4 billion in cash.
The IPO has been expected to bring in $2.3 billion to $2.7 billion in new cash, but that number could change if the IPO price is revised. CoreWeave will also receive $350 million from OpenAI in exchange for stock.
CoreWeave could use the liquidity and likely more. Some $5.6 billion of its loans must be repaid in 2025 and 2026. It will use some of the IPO proceeds to begin that process, with $1 billion going to pay off a 12%-interest loan which comes due at the end of the year. A key metric going forward will be its average interest rate, as high payments can destroy earnings like it did in 2024.
In addition to debt repayments, CoreWeave has $1.1 billion in lease payments to make in 2025 to 2026, a number that will likely rise.
But the bigger lift is the future capital expenditures it will need to keep this flywheel spinning. Capex was $8.7 billion in 2024, up 196% from the year before and it will have to keep growing at rapid rates to support the operations that the valuation portends. Because of the high capex, CoreWeave had a free cash flow loss of $8 billion in 2024, excluding customer deposits.
Medium and long term, a bullish outlook is founded on a belief that the demand growth for AI cloud computing will continue to outstrip supply, giving a smaller player like CoreWeave a lot of runway, even if just as overflow capacity. It's essentially the same pitch Nvidia CEO Jensen Huang made at length during last week's GTC conference keynote. In effect, CoreWeave would be a leveraged play on Nvidia.
But should demand growth rates dial down and be exceeded by supply growth, the CoreWeave flywheel could stall. That's a real risk if Big Tech, which is spending hundreds of billions on AI data centers, pulls back on those expenses in the near term.
In that scenario, CoreWeave's high depreciation and interest expenses would become a greater burden on its income statement. For example, should Microsoft no longer need the overflow capacity CoreWeave provides, it could quickly erase CoreWeave's operating profit.
CoreWeave's management has tried to mitigate some of this risk with a unique cloud business model. It signs "take-or-pay" contracts that force customers to pay a penalty if they don't use contracted capacity. Ranging from two to five years in length, the contracts start with a cash deposit of 15% to 25% of the contract value. At the end of 2024, CoreWeave had $4.1 billion in customer deposits on its balance sheet, $769 million due to be recognized as revenue in 2025.
Only after the contract signing does CoreWeave start to build the required capacity for the customer, a process that takes three months. While this reduces its risk of overbuilding in the short term, that three month delay may discourage some customer sign-ups.
Investors should also know some things about CoreWeave's capital structure. Class A shares, which will trade publicly, get one vote per share. A smaller group of non-trading Class B shares will get 10 votes apiece. That means Class B shareholders will control a majority of votes.
The stock could also be subject to early volatility: only 9% of fully diluted shares will be available to trade.
Ultimately, CoreWeave is a high risk-reward proposition for the long term investor. If AI computing demand continues to exceed supply, the stock could have a strong run out of the gate. But if AI spending falters, CoreWeave shares could be an immediate casualty.
Write to Adam Levine at adam.levine@barrons.com
This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
March 27, 2025 10:53 ET (14:53 GMT)
Copyright (c) 2025 Dow Jones & Company, Inc.
Comments