CoreWeave (CRWV) is expected to start trading Friday, March 28. At the midpoint of its IPO price range, $51 a share, CoreWeave earns an unattractive stock rating.
While CoreWeave’s expected $27 billion valuation is lower than the previously reported $35 billion target, it remains much too high.
CoreWeave leases GPUs, CPU’s, and equipment within data centers to customers. Its solutions, connected via “proprietary software and cloud services”, allow customers to deploy and develop artificial intelligence (AI) workloads. With booming revenue growth, CoreWeave and its bankers are looking to capitalize on the AI hype – the same market narrative that’s driven many other AI and AI-related stocks to great heights.
However, in my view, CoreWeave’s IPO looks more like a plan to boost Wall Street insider bonuses and throw a lifeline to a cash-burning business, rather than the launch of the next NVIDIA.
The company faces persistent threats, not just from ever-changing/upgrading technology, but also from its own customers. Plus, the current proposed $27 billion valuation is far too high as it implies that CoreWeave will grow revenue at a 30% compounded annual growth rate for 7 years, while also maintaining tech-like margins, a highly unlikely feat.
Add in that the company hides 200% of its debt off balance sheet, presents misleading profits, has weak internal controls, offers almost no voting rights to IPO investors, and the risks to potential IPO investors become crystal clear.
Below, I’ll detail these risks and use my reverse discounted cash flow (DCF) model to prove the current valuation is too expensive.
My firm’s IPO research provides investors with an alternative to the popular narratives on IPOs that is based on more reliable fundamental research.
Revenue Soaring, But Core Earnings Remain Negative
After moving away from its crypto mining roots, CoreWeave has seen its revenue soar amidst growing demand for AI computing power. In 2024, CoreWeave’s revenue grew 737% year-over-year (YoY).
However, CoreWeave’s Core Earnings remained negative, at $36 million in 2024. See Figure 1.
Figure 1: CoreWeave’s Revenue & Core Earnings: 2023 – 2024
Unlike many of the unprofitable companies that went public during the booming IPO market of 2020-2021, CoreWeave is efficiently scaling its business, at least from an income statement perspective.
In 2022, total expenses, which include cost of revenue, technology and infrastructure, sales and marketing, and general and administrative, were 245% of revenue. In 2023, total expenses fell to 106% of revenue before falling to 83% of revenue in 2024.
The biggest improvements come from:
- Cost of revenue: 77% to 26% of revenue from 2022 to 2024
- Technology and infrastructure: 114% to 50% of revenue from 2022 to 2024
- General and administrative: 38% to 6% of revenue from 2022 to 2024
The ability to grow revenue rapidly while managing costs is no small feat, especially in an era of reckless spending.
Figure 2: CoreWeave’s Falling Expenses: 2022 – 2024
Projecting Strong Market Growth
We’ve seen that any product and service even tangentially related to AI has received significant investor interest in recent years…for good reason. Many analysts and industry experts believe AI has the ability to transform the global economy for years to come. For instance, CoreWeave notes in its S-1 that:
- IDC estimates that AI will add nearly $20 trillion to global GDP by 2030.
- Bloomberg Intelligence estimates the market for AI inference/fine-turning, AI workload monitoring, and training infrastructure will grow at a 38% compound annual growth rate (CAGR) from 2023-2028.
With such high growth rates, it’s not surprising to see billions of capital flow to the AI market. And, Wall Street is, so far, successfully selling that CoreWeave offers the next big opportunity to invest in AI.
However, as I’ve cautioned before, investing in the wrong AI could crush your portfolio. No matter how good an AI company looks now, I think there will be many more DeepSeeks looking to disrupt the market and garner a share of the huge amount of capital flowing into the space. Anytime billions flow into an industry, competition ramps up and stock valuations soar. Investors need to avoid stocks that are too expensive because they already embed huge revenue growth and super high margins far into the future.
There are other risks of investing in CoreWeave, as I’ll show below.
Customers Are Heavily Concentrated
Just two customers accounted for 77% of CoreWeave’s revenue in 2024. Microsoft (MSFT), the company’s biggest customer, accounted for 62% of CoreWeave’s revenue.
Revenue concentration is growing, not falling. In 2023, CoreWeave’s top three customers accounted for 73% of its revenue. In 2022, the company’s top three customers accounted for 41% of its revenue.
From Michael Porter’s five forces, we all know that large customers can exert more influence over their suppliers. It’s fair to say that 2 customers accounting for 77% of revenue can exert a lot of influence.
Should Microsoft decide that CoreWeave costs too much, they have significant leverage to negotiate better terms. In fact, The Financial Times reports that Microsoft is already reducing commitments to CoreWeave due to “delivery issues and missed deadlines.”
While CoreWeave denies the reporting, such a story highlights the risk of such high customer concentration.
Biggest Customers Can Take Business In-House Too
Microsoft forcing lower prices on CoreWeave may be the least of the company’s (and investors’) concerns.
Microsoft could decide that leasing compute power and data center space from CoreWeave is simply too expensive, and that it would be better to bring such operations in house…to its own Azure data centers. In such a scenario, CoreWeave’s revenue and profit growth would take a material dive.
This risk is not a hypothetical one. Microsoft Azure already operates 300 data centers across the globe, and Microsoft recently reiterated its plans to invest billions in AI infrastructure in 2025. When the company doesn’t have enough capacity through its own Azure data centers to meet compute demand, it pays CoreWeave to fill the gap. However, should Microsoft build out enough Azure capacity to meet demand, it would no longer need to use CoreWeave.
Worse yet, in a scenario where AI compute demand falls across the entire industry, neither Microsoft nor other potential customers would need CoreWeave. Such a scenario may be more likely than the market is pricing in. Alibaba’s Chairman Joe Tsai recently noted there may be a bubble forming in data center construction as the pace of investment may outstrip demand for AI services.
Microsoft isn’t the only company that could find itself with no need for CoreWeave’s services either. In the leadup to its IPO, CoreWeave is touting a new agreement to provide cloud computing capacity to OpenAI. The deal is worth up to $11.9 billion through 2030. However, OpenAI is also building its own data centers, which could limit demand for CoreWeave’s services moving forward.
One fact remains clear, CoreWeave faces lots of competition from a variety of firms, including its largest customers.
Competitors’ Margins Remain Much Higher
Not only is CoreWeave’s biggest customer also a competitor, but CoreWeave is much less profitable than all its main competitors. In its S-1, CoreWeave notes its main competitors include the most well-known hyperscalers Amazon (AMZN) AWS, Alphabet (GOOGL), and Microsoft (MSFT), as well as International Business Machines (IBM), and Oracle (ORCL). Because these competitors have numerous business lines outside of AI computing, I think CoreWeave will struggle to gain any competitive advantage over its larger and more profitable competitors.
Per Figure 3, CoreWeave’s invested capital turns (a measure of balance sheet efficiency) and return on invested capital (ROIC) rank last amongst its competitors.
The company’s invested capital turns of 0.1 are particularly low and speak to the capital-intensive nature of the company’s business model, which requires leasing expensive data centers and filling those data centers with expensive equipment. With such low IC turns, the company only earns a 2% ROIC, even though its NOPAT margin is comparable to its competitors.
Outfitting its data centers with the latest technology to handle and process AI workloads will continue to require large amounts of capital, which means capital efficiency is unlikely to improve.
Figure 3: CoreWeave’s Profitability Vs. Competition: TTM
Nearing Zombie Stock Status
Not only is CoreWeave less profitable than its competitors (who are also customers), but it is running out of capital to sustain its business.
In 2024, CoreWeave burned $6.9 billion in free cash flow (FCF). As of December 31, 2024, CoreWeave had just $1.4 billion in cash and cash equivalents on its balance sheet. CoreWeave could sustain its 2024 burn rate for only ~2.5 months from December 31, 2024. In other words, CoreWeave needs a capital raise to remain a going concern, which could explain the timing of its IPO and willingness to continue an IPO at a lower valuation than the company’s previously reported target.
Even if I include the proceeds of the IPO, between $2.3 and $2.7 billion (depending upon the underwriters’ decision to exercise their over-allotment option) in the cash on CoreWeave’s balance sheet, CoreWeave’s cash on hand would be at most $4.1 billion.
With $4.1 billion in cash, the company could only sustain its 2024 cash burn rate for ~9 months from December 2024.
In other words, CoreWeave needs to significantly improve operations and drastically slow cash burn to remain a going concern.
CoreWeave, with its high FCF burn and short cash runway, nearly qualifies as a Zombie Stock before it even goes public. The only criteria keeping it from being a Zombie Stock is its positive interest coverage ratio. Should this ratio fall into negative territory, CoreWeave would meet all the criteria of a Zombie Stock.
200% Hidden Debt Adds More Risk
At first glance at CoreWeave’s balance sheet, investors may presume that CoreWeave has just $8 billion in total debt. In fact, in the company’s S-1, it notes as of December 31, 2024, that its “total indebtedness was $8.0 billion.” That figure understates the company’s total debt by over 3x, as I’ll show.
Only looking at short- and long-term debt risks missing a material liability in the form of operating leases. As part of its ongoing operations, CoreWeave leases data centers, brings in the technology to run its software, and then rents/leases that system back out to its end customers. Accordingly, operating leases are a key part of CoreWeave’s business.
While the company reports $213 million in current operating lease liabilities and another $2.4 billion in non-current operating lease liabilities, these values only scratch the surface of CoreWeave’s actual lease obligations.
As part of Accounting Standards Update (ASU) 2016-02, companies can exclude “Not-Yet-Commenced” leases from the calculation of their operating lease obligation. To ensure I capture all of a company’s obligations, I included not-yet-commenced leases in my present value of future operating leases calculation.
When doing so, I find that, with $15 billion in not-yet-commenced leases on its books, the total value of CoreWeave’s operating lease debt is $16.5 billion. When combined with CoreWeave’s reported debt, I find the company’s Total Debt is actually $24.5 billion, or over 3x higher than the $8 billion figure quoted in the “Risks” section of the company’s S-1.
Red Flag #1: Presenting Misleading Results
Many high growth and/or unprofitable companies present non-GAAP metrics to appear more profitable than they really are, in hopes of justifying a higher valuation. CoreWeave is no different. CoreWeave provides investors with the popular Adjusted EBITDA metric. Not surprisingly, Adjusted EBITDA gives a more positive picture of the firm’s business than GAAP net income and economic earnings.
For instance, CoreWeave reports adjusted EBITDA of $1.2 billion in 2024. Meanwhile, GAAP net income is -$863 million and economic earnings are even lower at -$1.4 billion. See Figure 4.
Why would the company overstate its results by so much? Maybe, the bankers are pushing this narrative to reap a big payday with the IPO?
Figure 4: CoreWeave’s Adjusted EBITDA, GAAP Net Income, and Economic Earnings: 2023 – 2024
Big Red Flag #2: Weak Internal Controls
Even after I adjust the company’s numbers to get the right metrics in place, I don’t know if I can trust the financials because the company identified material weaknesses in its internal control over financial reporting. Specifically, the material weaknesses pertained to:
- the lack of effectively designed, implemented, and maintained IT general controls over applications that support our financial reporting processes,
- insufficient segregation of duties across financially relevant functions, and
- lack of sufficient number of qualified personnel within our accounting, finance, and operations functions who possessed an appropriate level of expertise to provide reasonable assurance that transactions were being appropriately recorded and disclosed.
The company notes that “management has made improvements”, but the identified material weaknesses remain un-remediated. CoreWeave expects remediation efforts to continue to take place through 2025 and 2026.
Weaknesses in internal controls increase the risk that the company’s financials are fraudulent and/or misleading. Making matters worse, the remediation efforts are expected to continue into 2026, thereby increasing the time during which the company’s financials are less trustworthy and increase the likelihood of restatements.
The controls are so bad that they will take years to fix. Not a good sign.
Big Red Flag #3: IPO Investors Get Almost No Voting Power
In its S-1, CoreWeave disclosed it will go public with dual class shares. Class A shares will be offered as part of the IPO. These class A shares will receive one vote per share and class B shares will receive 10 votes per share. This share and voting structure will give the three co-founders significant say over corporate governance matters.
After the IPO, CoreWeave’s three co-founders, Michael Intrator (Chief Executive Officer), Brian Venturo (Chief Strategy Officer), and Brannin McBee (Chief Development Officer) will hold 79% of the voting power in the company. If the three co-founders elect to exercise their stock options to purchase Class A shares and exchange them for an equal number of class B shares (pursuant to their Equity Exchange Rights), they would hold 83% of the voting power in the company.
In other words, new shareholders will not have much say in the corporate governance of this business.
Big Red Flag #4: Emerging Growth Company Status Limits Disclosures
By confidentially filing its S-1 with the SEC when the company was still an emerging growth company, CoreWeave can, and did, take advantage of reduced disclosure requirements in its S-1.
CoreWeave notes in its S-1 that it is using extended transition periods for complying with new or revised accounting pronouncements, which is one of the reduced requirements available to emerging growth companies.
No matter the company, reduced disclosure is bad for investors. More disclosure is beneficial to all involved, unless the company has something to hide.
These three red flags do no present a promising pattern of behavior for any company, especially one about to go public at a valuation that implies nothing but lots of good news for he foreseeable future.
Valuation Implies Rapid Revenue Growth and High Margins
I think the stock holds large downside risk at its IPO valuation, especially considering all the red flags I’ve identified.
When I use my reverse discounted cash flow (DCF) model to analyze the future cash flow expectations baked into CRWV, I find that shares, at $27 billion, embed very optimistic assumptions about margins and growth.
To justify its midpoint IPO valuation of $51/share, my model shows that CoreWeave would have to:
- maintain NOPAT margin at 27% from 2025-2032 and
- grow revenue by 30% compounded annually through 2032.
In this scenario, CoreWeave would generate $15.6 billion in revenue in 2032, which is 8.2x its 2024 revenue. While this revenue may pale in comparison to the revenue levels of CoreWeave’s customers/competition, it dwarfs the revenue of comparable data centers, which is the core of CoreWeave’s operation. $15.6 billion in revenue would be 1.1x greater than the combined revenue of Equinix (EQIX) and Digital Realty (DLR).
This scenario also implies CoreWeave grows NOPAT to $4.2 billion in 2032, compared to $513 million in 2024.
I think that it is overly optimistic to believe CoreWeave will maintain SAAS-level margins over the long-term without risking losing its biggest customers (who increasingly move services in-house) while also growing revenues at a 30% CAGR.
Companies that grow revenue by 20%+ compounded annually for such a long period are “unbelievably rare”, making the expectations in CoreWeave’s midpoint IPO price even more unrealistic.
Up until now, CoreWeave’s moat has been capital based, i.e. the ability to acquire GPUs and sell compute power. Capital as a weapon will last only so long as IPO investors are willing to subsidize it. Without any other meaningful competitive advantage, margins are likely to fall.
There’s 43%+ Downside If Margins Fall in the Slightest
Even if I instead assume CoreWeave can maintain high revenue growth rates, as the AI boom continues, and margins only slightly fall, the stock has large downside risk.
If I assume:
- NOPAT margin falls to 20% (above IBM and traditional data center operators DLR and EQIX), and
- revenue grows 30% compounded annually through 2032, then
the stock would be worth just $29/share today – a 43% downside to the midpoint IPO price. This scenario still implies CoreWeave’s NOPAT improves to $3.1 billion in 2032, or 6x its 2024 NOPAT.
Should AI demand slow, or worse, demand for CoreWeave’s services fall precipitously as customers move their operations in-house, the downside risk could be even greater.
Figure 5 compares CoreWeave’s implied future NOPAT in these scenarios to its historical NOPAT. For reference, I include traditional data center operators Digital Realty and Equinix.
Figure 5: Estimated IPO Price Looks Overvalued
Each of the above scenarios assume CoreWeave grows revenue, NOPAT, and FCF without increasing working capital or fixed assets. This assumption is highly unlikely but allows us to create best-case scenarios that demonstrate the high level of expectations embedded in the midpoint IPO valuation. For reference, CoreWeave’s invested capital increased 36% YoY in 2024.
Disclosure: David Trainer, Kyle Guske II, and Hakan Salt receive no compensation to write about any specific stock, style, or theme.
Read the full article here
