NEXT PICK · Market Insights
Hundreds of billions in orders hitting a 15% bond interest rate—CRWV in jeopardy?
Thursday, October 22, 2026
Revenue doubled, orders surpassed 100 billion, yet the stock price halved in a year—CoreWeave's key to success was not AI demand, but its balance sheet.
A company's quarterly revenue more than doubled year-on-year, with $104 billion in outstanding orders, yet its stock price fell 10.06% in a single day on October 22, closing at $55.6, marking a new yearly closing low, with its market value shrinking to $30.67 billion.
Even more glaring, one of the company's bonds with a coupon rate of 9.625% once yielded as close as 15%—the price offered by the bond market was close to pricing for high-risk borrowers.
The stock market looks at orders, while the bond market looks at repayment ability. When two markets give completely opposite answers to the same company, who should we trust? This is the CoreWeave ($CRWV) we're breaking down today.
|
Latest price
$55.6
▼ -10.06%
|
Order backlog
$104B
▲ Q3 initial new signings exceeded $25B
|
Bond yields
15%
▼ Reaching a high position
|
It doesn't sell clouds, but rather an AI powerhouse
CoreWeave is not just another "small version of AWS." From day one, it does only one thing: connect thousands of GPUs into large-scale AI clusters capable of training and inference, then lease the entire set to companies that need computing power.
In its product line, it packages GPU and CPU computing, networking, hosting services, virtual machines, and bare metal servers, stacking CoreWeave Kubernetes, runtime acceleration, data and storage solutions, and a mission control management system that covers individual nodes, entire cabinets, and the entire server lifecycle.
The acquired Weights & Biases extended its reach into model and agent development tools, extending its business boundaries to visual effects rendering, pixel streaming, and batch processing. It aims to upgrade from being a "landlord selling computing power" to an integrated service provider that "bundles renovation and property management."
The scale is no longer at the pilot level
Revenue for fiscal year 2024 was $1.92 billion, jumping to $5.13 billion in fiscal year 2025, a year-over-year increase of about 167%; Quarterly revenue for Q2 2026 was $2.575 billion, up 112% from $1.212 billion in the same period last year.
But the fuel for this "power plant" is money. Data centers, GPUs, and electricity are all heavy assets. Capital expenditure for the full year of 2025 is $14.9 billion, and this year the company has raised its full-year capital expenditure guidance to $35 billion to $39 billion.
So, just how strong is its irreplaceability? The research report gives it a score of 6 out of 10: its advantage lies in hardware-software integration and experience in scaled operations, rather than patents or exclusive chips. Currently, there is no verifiable data proving it is cheaper, faster, and more comprehensive than Azure, AWS, Google Cloud, or Oracle, yet it is not unbeatable. **
Behind the 4.24 price-to-sales ratio lies a miswritten cash flow figure
By valuation snapshot, CRWV has a price-to-sales ratio (P/S) of 4.24 times, a price-to-book ratio (P/B) of 6.10 times, and a return on equity (ROE) of -35.15%; Due to ongoing losses, neither the price-to-earnings ratio nor the enterprise value multiple can produce meaningful figures.
The profit and exit statement is in the red
Net loss of $1.167 billion in 2025, with a loss per share of $2.81; In Q1 and Q2 2026, losses of $740 million and $626 million respectively, with a loss per share of $1.14 in Q2. Operating profit was $51.9 million in Q3 2025, then fell back to -$144 million and -$49 million in Q1 and Q2 this year.
The most attractive figure in the report is the "free cash flow yield of 9.97%," corresponding to $3.06 billion in 2025 and $2.98 billion in Q1 2026 "free cash flow." But after checking the financial report, these figures actually represent operating cash flow, not free cash flow after deducting capital expenditures.
Doing the math reveals a completely different picture
In 2025, operating cash flow will be about $3.06 billion, but capital expenditure will be $14.9 billion; In the first half of 2026, operating cash flow will be $3.663 billion, capital expenditure $14.117 billion, indicating deep negative free cash flow. Media statistics show its rolling 12-month free cash flow has reached -$13.66 billion.
Therefore, that nearly 10% "yield" is not money the company can freely distribute to shareholders, but rather the gross cash that hasn't even deducted the costs of building the data center and buying graphics cards. The research report itself warned that it cannot be mechanically interpreted as "undervalued," but now, looking at this warning, it's still a bit light.
The valuation model in the report is anchored mainly by P/S: low 2.5x, benchmark 4.24x, high 6.0x, with the conclusion stretched (somewhat expensive)—the growth premium has already been paid, but evidence of profitability and cash flow has not yet arrived.
From another perspective, the company's full-year revenue guidance is $12.4 billion to $13.2 billion. Based on this standard, the forward price-to-sales ratio at current prices is less than 2.5 times. The same company may be expensive in the future but cheap in the future; the difference is whether those tens of billions in debt can smoothly roll down. For the value per share and real-time buy-ask levels corresponding to the three scenarios, see the official account.
Revenue has taken off, but the stock price has fallen back to the starting point
The report positions CRWV in the breakout phase: quarterly revenue will grow from $982 million in Q1 2025, to $1.21 billion, $1.36 billion, then $2.08 billion in Q1 and $2.575 billion in Q2 this year, far beyond the proof of concept and commercial pilot stages.
But breakthroughs do not mean maturity. Net profit remained negative, operating profit turned negative again, and the combined outstanding shares of Class A and Class B rose from 526 million shares at the end of January to 552 million shares at the end of July, about 4.9% in just over half a year, indicating heavy financing dependence.
The market is clearly no longer paying for a "breakout."
The current price is about 35% below the 50-day moving average and nearly 40% below the 200-day moving average, with both moving averages holding above the price. Fundamentals are accelerating, but stock prices are falling rapidly, and the market is reclassifying it from a "growth stock" to a "credit risk stock."
A challenger caught between giants and new elites
The first layer of competitors are Microsoft Azure, AWS, and Google Cloud. They have capital, channels, and a complete ecosystem, but their weakness is a high proportion of general cloud resources, so they may not be specifically optimized for frontier training and inference. CRWV excels in focus but loses in scale, and is more sensitive to financing and supply chains.
The second layer is Oracle Cloud Infrastructure. Oracle has enterprise customers, database ecosystems, and the ability to deliver large contracts, and its balance sheet can better withstand financing cycles. CRWV focuses more on dedicated accelerator clusters, Kubernetes native orchestration, and developer tools, but there is currently no data to determine which unit economics are superior.
The third layer is similar "neocloud" platforms like Crusoe, Lambda, RunPod, and Nebius. They compete for the same batch of GPUs and high-power data centers, and may be more flexible in inference or price-sensitive markets. The scarcer the GPU, the more attractive CRWV becomes; Once chip supply becomes loose, price wars will come knocking.
On October 15, Gartner issued a risk warning for the new cloud industry, and on the 22nd, Nebius and IREN also plunged, indicating the market is repricing the entire sector, not just CRWV.
Its moat is more like a puzzle piece
Hardware-software integration, customer migration costs, deployment scale, plus physical resource barriers like power, grid connection, and site selection that are difficult to replicate in the short term. Lack of evidence from the patent dimension, and network effects are just a weak version where "the more you use, the more experience you have."
The question is, when the moat mainly relies on "electricity and cards secured first," is it a barrier or a deposit that must be continuously renewed? The verification metrics listed in the report—customer retention, gross profit per unit computing power, GPU utilization rate, delivery cycle—have not yet been fully disclosed by the company.
In trillions in AI investment, how much of the share can it get?
The company did not disclose its official Total Market Size (TAM). The report borrows Morgan Stanley's judgment: AI investment cycles could reach $1 trillion per year, lasting 5 to 10 years, assuming 15% to 30% of that flow to computing infrastructure, resulting in an infrastructure pool of $150 to $300 billion annually, with a dedicated AI cloud market of about $60 to $180 billion.
The report starts from a valuation snapshot and estimates annualized revenue of about $7.23 billion, creating a three-year scenario: under the low 25% growth scenario, revenue in 2029 is about $14.1 billion; in the 40% baseline scenario, about $19.8 billion; and in the 60% high scenario, about $29.6 billion.
Reality outpaces research reports
The company indicates an annualized revenue run rate of $18.5 billion to $19.5 billion by year-end, with about 3.7 GW of contracted electricity and 1.5 GW of commissioned electricity. The ceiling is not the problem; the problem is that the ladder to climb requires borrowing money.
The bulls' trump card: orders are real, demand is real
Bulls don't need to convince you that AI will go viral; they just need to prove that CoreWeave can safely convert orders into cash:
| ▲ Bull Case |
| ① | Q2 revenue was $2.575 billion, up 112% year-on-year, marking several consecutive quarters of rapid growth |
| ② | At the end of June, orders backlog was about $104 billion, and at the beginning of Q3, more than $25 billion in new customer commitments were signed |
| ③ | The full-year revenue guidance has been raised to $12.4 billion to $13.2 billion, with a forward-looking price-to-sales ratio of less than 2.5 times based on guidance |
| ④ | Adjusted EBITDA for the second quarter was $1.51 billion, with a margin of 59%, indicating solid operating efficiency before interest |
| ⑤ | Caterpillar, Bentley, Grammarly, and other new clients have joined; Databricks and Cognition have expanded their cooperation, resulting in a fragmented customer structure |
| ⑥ | RSI (14) is 26.6, down 37.5% over 30 days, indicating an extremely pessimistic sentiment; Goldman Sachs is counter-trending to buy |
The bear's knife: the fragile balance sheet in a good industry
Bears do not deny AI demand; they question how much shareholders can actually get behind the $30.67 billion market cap:
| ▼ Bear Case |
| ① | Net interest expense in Q2 was $640 million, more than doubling in one year, with Q3 guidance rising to $860 million to $940 million |
| ② | At the end of Q2, cash was $5.524 billion, interest-bearing debt exceeded $35 billion, and about $4 billion in corporate bonds fell below 80 cents |
| ③ | After deducting capital expenditures, the depth of free cash flow is negative; the research report cites a "yield" of about 10%, but in reality, it is based on operating cash flow |
| ④ | Outstanding shares increased by about 4.9% in half a year, and the co-founder has reduced holdings by approximately $3.3 billion since the IPO |
| ⑤ | Operating profit turned from profit to loss, with a second-quarter operating loss of $49 million, and scale expansion has yet to bring operational leverage |
| ⑥ | HSBC maintains its underweight rating, Bernstein lowers its target price and maintains an underperforming rating |
In the next three months, three keys will determine the direction
The next pace of stock prices will most likely be determined by earnings reports, credit markets, and industry sentiment. Whoever gives the answer first gains pricing power:
| • | Q3 earnings report (third-party calendar expected mid to early November): Whether the revenue guidance of $3.45 billion to $3.6 billion can be met is relatively high |
| • | Bond yields and credit default swap trends: If financing costs fall, stock prices will be the first to recover, with a high impact |
| • | New financing structure implementation: If lower-cost, less diluted financing emerges, the bearish logic will be weakened, resulting in a high impact |
| • | Operating profit and free cash flow disclosure: Whether the situation can continue to improve and break down the cash flow composition will affect medium- to long-term valuation |
| • | Mega clients like Meta are increasing capital expenditure and inference load, with external cloud capacity demand being transmitted and having a moderate impact |
From $141 to $55, what fell was trust
Over the past 12 months, CRWV returned -54.2%, with a 52-week high of $141.79 and less than 40% of the current peak. Ironically, revenue during the same period rose from $1.36 billion quarter to $2.575 billion, and orders increased from $66.8 billion by the end of 2025 to $104 billion.
The real turning point came in mid-October: on the 13th, the yield on its 2032 bonds hit 15%; on the 14th, the stock price fell 15.5% in a single day, with a weekly cumulative drop of nearly 25%; Subsequently, HSBC reiterated its bearish stance, Bernstein lowered its target price, and news of founder selling shares fermented, causing the stock price to fall below a 52-week low.
Looking back, the expectation of "strong demand and faster revenue growth" has basically come true; The expectation that "growth naturally translates into profits and cash" has not been realized; "Oversold will lead to a rebound" has yet to be confirmed by fundamentals. The market is not punishing growth, but the cost of growth.
|
⚠️ Risk Notice
|
|
🟡 Neutral Orders and growth are real, but debt pressure is real—neither side has won yet. |
|
💬 Discussion Bottom line: Do not hold heavy positions before financing costs fall. Live buy and sell positions, post CRWV on the official WeChat account. |
Data source
| • | Source: NextPick real-time snapshot + company earnings and investor relations announcements + SEC filings + mainstream financial media coverage. |