CoreWeave Surged on a 168% Revenue Beat — but It Still Fails the Halal Screen (2026)
CoreWeave reported after the close on Tuesday 10 November 2026, and the print was the clearest demand test the artificial intelligence trade has had in weeks. Third-quarter revenue was $3.66 billion, up 168 percent from about $1.36 billion a year earlier and ahead of the $3.55 billion analysts expected, the company raised its full-year 2026 guidance on every line that matters, and the revenue backlog — contracted future computing work — reached $132.5 billion, up from $104.2 billion at the end of June. The shares, which had fallen more than 40 percent in a month and about 65 percent from their year-to-date high, were quoted about 5 percent higher after hours at $51.38. The other half of the print was the bill for that growth. The GAAP net loss widened to $709 million, or $1.27 a share, from a $110 million loss a year earlier, capital spending was $14.5 billion in the quarter alone, debt stood at almost $47 billion at the end of the quarter, and the chief financial officer said the company does not intend to issue additional parent-level unsecured debt at current yields through 2027. This article lays out what actually printed, why the loss widened while revenue tripled, what the guidance raise really commits the company to spend, what the $132.5 billion backlog is and is not, how much debt the company carries, and the part that matters to a halal investor: why a stock can beat by 168 percent, rise 5 percent, and still fail the Shariah screen by the widest margin of any name in this run of articles. This article was written after the close on Tuesday 10 November 2026. The numbers below are the result, not the forecast.
Updated November 2026 · Reviewed by the Canadian Halal Investor team
The print: revenue up 168 percent to $3.66 billion
CoreWeave is the company at the centre of the argument about whether artificial intelligence infrastructure is being overbuilt, and this quarter was the argument’s first real test since the market started asking it in earnest. The company rents out access to clusters of graphics processors — the chips that train and run AI models — to customers who sign long-term contracts, and it builds the data centres that house them. For much of 2026, investors have wrestled with a simple question: are the tech giants overbuilding AI infrastructure ahead of actual demand? A company that sells the picks and shovels of the AI build-out, and that reports a 168 percent revenue increase, is the closest thing to a direct answer the market gets, and the answer this quarter was that demand is still outrunning supply. Revenue of $3.66 billion, up 168 percent from about $1.36 billion in the third quarter of 2025, beat the $3.55 billion analysts expected by about 3 percent, and the company said it added more than $1 billion of sequential quarterly revenue in the quarter alone. Adjusted operating margins crossed into double digits a full quarter ahead of the company’s own plan, with adjusted operating income of $379 million, and the company added more than 200 megawatts of power capacity during the quarter, taking active power to over 1.7 gigawatts. The print landed on a day when the market was already on edge: Tuesday’s hot CPI reading had pushed the December rate hike to near-certain and sent the 10-year Treasury yield to almost 5 percent, and Nvidia, CoreWeave’s most important hardware partner, had fallen 5.5 percent on the day. Against that backdrop, a revenue beat of this size was always going to move the stock, and it did, by about 5 percent after hours to $51.38, against a regular-session close near $48.73.
The context of that 5 percent matters as much as the 5 percent itself, because the stock had been falling for a month. CoreWeave had slid more than 40 percent over the past month and about 65 percent from its year-to-date high, as investors soured on heavily indebted cloud infrastructure startups and as the broader market grappled with the board probe into Nvidia chief executive Jensen Huang, the hot CPI print, and the question of whether AI spending can ever pay for itself. A 5 percent after-hours rise from $48.73 takes the stock to $51.38, which is still roughly a third below where it traded a month ago, and an overnight marker near $54.50 the following morning would still leave it well below its highs. The move is a relief rally inside a sell-off, not a reversal of it, and the options market had priced a move of about 21.5 percent in either direction for the print, the largest of the week, which is the market saying it did not know which way this one would go. The revenue beat resolved one half of the uncertainty, the demand half. The other half, the financing half, is what the rest of this article is about, and it is the half that matters to a halal investor.
There is also a detail in the print that deserves attention because it explains how a company can grow revenue 168 percent and still lose money faster than it did a year ago. The GAAP net loss widened to $709 million, or $1.27 a share, from a $110 million loss in the same quarter of 2025, a more than sixfold increase in the reported loss even as revenue nearly tripled. Part of the gap between the adjusted figures and the GAAP figures is stock-based compensation, a recurring scrutiny point for CoreWeave, and part of it is the cost of the build-out itself: the company is buying graphics processors and building data centres at a pace that revenue, however fast it grows, cannot yet cover. Adjusted operating income of $379 million against $3.66 billion of revenue is a margin of about 10 percent, and the company said it would be profitable on an adjusted basis in the first half of 2028. That is the target the market is being asked to believe in. It is a 2028 target, not a 2026 result, and the distance between the two is measured in debt.
The loss widened anyway: $709 million, and interest expense is the tell
A revenue beat that comes with a widening loss is not a contradiction. It is the normal shape of a company that is building capacity faster than it is selling it, and CoreWeave is building at a pace that has few precedents. Capital spending in the third quarter was $14.5 billion, about four times the quarter’s revenue, and the company’s own guidance for third-quarter interest expense was $860 million to $940 million, up from $640 million in the second quarter and well above the $200 million to $260 million of adjusted operating income it had targeted for the period. Read those two numbers together and the shape of the business becomes clear: the company is paying more in interest in a single quarter than it earns in adjusted operating income, and the interest bill is rising faster than the operating income. That is not a criticism of the business model. It is the business model. CoreWeave borrows money, buys graphics processors, builds data centres, and rents the capacity out on long-term contracts. The revenue arrives over years. The interest arrives every quarter. The question is whether the contracted revenue is large enough, and durable enough, to cover the interest and the principal, and that question is the one the backlog is supposed to answer.
The interest expense is also the number that connects this earnings report to the rest of the week’s news, and to the halal screen, in a way that the revenue number does not. Interest expense is not interest income, and the screen’s third gate measures interest income, not interest expense, so the $860 million to $940 million of interest expense does not itself move the screen. But it is the cost of the debt that the screen’s second gate measures, and it is the reason the debt gate matters. A company that pays close to a billion dollars of interest in a quarter is a company whose lenders have a large claim on its future cash flow, and the screen’s debt ceiling exists precisely to keep that claim from dominating the company. The rate environment makes the claim heavier: Tuesday’s hot CPI print put a December rate hike at near-certain, with a 15.3 percent chance of a half-point move, and the two-year Treasury yield rose 17 basis points to 4.71 percent. Higher rates raise the cost of the debt CoreWeave must service and the cost of the debt it must still raise, and the chief financial officer, Nitin Agrawal, said that at current yields the company does not intend to issue additional parent-level unsecured debt through 2027. That is a company telling the market that its own borrowing costs have become too expensive to bear at the parent level, and that it will fund the build-out another way instead.
The other way, according to the chief executive, Mike Intrator, is the customers themselves. He told analysts that most of the company’s capital requirements in fiscal 2027 will be sourced from clients via prepayments on infrastructure services and from hardware partners such as Nvidia, and that about $9 billion of customer prepayments and deposits were expected over the next three months. That is a real funding source, and it is also a concentration risk: the prepayments come from a small number of very large customers, and the hardware comes from a single supplier whose chief executive is under a board investigation. The company also raised about $4.2 billion through convertible senior notes and established an equity issuance programme covering up to 35 million shares, which is a source of dilution for existing shareholders. The picture is a company that is funding a $37 billion to $39 billion annual build-out with customer prepayments, convertible notes, and new shares, because the unsecured debt market has priced it out. That is a workable funding plan, and it is also a fragile one, and the fragility is the point: a company that cannot borrow at the parent level at acceptable rates is a company whose debt is already large enough that the market is charging it for it.
Guidance went up everywhere — and so did the spending
The guidance raise was the part of the print that moved the stock, and it was broad. Full-year 2026 revenue is now guided at $13.0 billion to $13.3 billion, up from $12.4 billion to $13.2 billion, with the floor rising $0.6 billion and the ceiling just $0.1 billion. Adjusted operating income is guided at $1.275 billion to $1.375 billion, up from $960 million to $1.15 billion, a midpoint increase of about 26 percent and above the $1.07 billion analysts had pencilled in. The exit annualised revenue run-rate was raised to $20 billion to $21 billion, from $18.5 billion to $19.5 billion, a midpoint increase of about 8 percent. The most significant upgrade was the adjusted operating income line, which moved by a quarter, and it is the line the company is using to tell the market that the business is scaling: adjusted operating margins crossed into double digits a full quarter ahead of the internal plan, and the company said adjusted profitability should arrive in the first half of 2028. A guidance raise that lifts the operating income forecast by 26 percent while lifting the revenue forecast by less than 3 percent is a company saying it is getting more profitable per dollar of revenue, and that is the number the market rewarded.
The spending guidance went up too, and by a similar logic it is the number that limits how much the raise can be celebrated. Capital spending is forecast at $37 billion to $39 billion for the full year, up from $35 billion to $39 billion, with the floor rising $2 billion and the ceiling unchanged. At the midpoint, that is about $38 billion of capital spending against about $13.15 billion of guided revenue, or about 2.9 times revenue. The third quarter alone spent $14.5 billion, about four times the quarter’s revenue, and the company is on track to spend roughly three dollars on capacity for every dollar of revenue it takes in this year. That ratio is the whole argument about CoreWeave in a single number. If the capacity is filled by the $132.5 billion backlog, the spending is an investment. If it is not, the spending is a liability, and the liability is funded with debt. The guidance raise does not resolve that argument. It raises both sides of it at once: more revenue, and more spending to get it. The company’s own numbers say the same thing. The revenue floor rose $0.6 billion. The capital spending floor rose $2 billion. For every dollar of extra revenue the company now expects, it expects to spend more than three dollars of extra capital to earn it.
There is a second way to read the guidance, and it is about what the company did not raise. The revenue ceiling rose by only $0.1 billion, from $13.2 billion to $13.3 billion, while the floor rose by $0.6 billion. The range narrowed from the top and widened from the bottom, which is the shape of a company that is more confident about the minimum it will earn and no more confident about the maximum. The adjusted operating income range moved by a similar shape, from $960 million to $1.15 billion up to $1.275 billion to $1.375 billion, with the whole range shifting up rather than widening. The capital spending range narrowed from below, from $35 billion to $39 billion up to $37 billion to $39 billion, which is the company committing to spend at least $37 billion and no more than $39 billion. The picture is a company that knows what it will spend, is increasingly confident about the floor of what it will earn, and is not yet willing to promise more at the top. That is a reasonable posture for a company spending three dollars to earn one. It is also a posture that leaves the financing question open, because $37 billion to $39 billion of spending against $13 billion to $13.3 billion of revenue is a gap that has to be funded from somewhere, and the somewhere, this year, is debt, prepayments, and new shares.
The $132.5 billion backlog: contracted, not collected
The backlog is the number that does the most work in CoreWeave’s story, and it is the number most often misunderstood, so it is worth stating plainly what it is. The revenue backlog was $132.5 billion at 30 September 2026, up from $104.2 billion at the end of June, a rise of about $28 billion in a single quarter and about 138 percent year on year. It is contracted future revenue, mostly from committed customer contracts, and the company says 98 percent of its revenue comes from committed contracts, with named customers including Microsoft and OpenAI. It is not cash in hand. It is not revenue earned. It is the total value of the contracts the company has signed, to be delivered as it builds the capacity and hands it over, over years. The backlog grew by more than the entire quarter’s revenue, which is why the market read it as proof that demand is still outrunning supply, and why the company could raise its guidance with confidence. A backlog that grows $28 billion in a quarter is a company whose customers are signing faster than it can build, and that is a real signal about demand for AI computing.
The signal is real, and it is not the whole story, because a backlog is also a promise the company has to keep, and keeping it costs money. Every dollar of backlog revenue requires capacity that has to be built, and the capacity costs about three dollars of capital spending for every dollar of revenue, at the ratio this year’s guidance implies. The backlog of $132.5 billion, at that ratio, implies something on the order of hundreds of billions of dollars of future capital spending to deliver it in full, and the company’s own guidance says it will spend $37 billion to $39 billion this year alone. The backlog is also concentrated: it comes from a small number of very large customers, and the chief executive said about $9 billion of customer prepayments and deposits were expected over the next three months, which is the mechanism by which the customers fund the build-out in advance. That mechanism works as long as the customers keep paying in advance, and it stops working if they do not. The backlog, in other words, is a measure of demand and a measure of obligation at the same time, and the two sides of it move together. The bigger the backlog, the more demand there is, and the more the company has to spend to meet it.
There is a useful comparison inside the company’s own history. At the end of June, the backlog was $104.2 billion. At the end of September, it was $132.5 billion. In the same three months, the company added more than 200 megawatts of power capacity, taking active power to over 1.7 gigawatts, and spent $14.5 billion on capital. The backlog grew faster than the capacity, which is the direction a bull wants to see, and the capacity grew faster than the revenue, which is the direction a bear points to. Both are true at the same time, and they will stay true until the capacity catches up with the backlog or the backlog stops growing. The next data point on that question is not CoreWeave’s next earnings report. It is Nebius, the other neocloud, which reports on Thursday, and then Nvidia’s own report on 17 November, which is the last of the Magnificent Seven to report and the one that will say whether the customers who are signing CoreWeave’s contracts are still buying Nvidia’s chips at the same pace. A backlog is a promise. The next two reports are the first check on whether the promise is being kept.
The debt question: almost $47 billion, and a market capitalisation that shrank
The debt is the part of the CoreWeave story that does not fit in a headline about a revenue beat, and it is the part that decides the screen. CoreWeave owed almost $47 billion in debt at the end of the quarter, according to reporting on the results, up from $35.1 billion of interest-bearing debt on the filed balance sheet at 30 June 2026, the figure this site’s screen used. The increase is the cost of the build-out: the company borrowed to buy graphics processors and build data centres, and the borrowing shows up as debt before the revenue shows up as backlog conversion. The third-quarter interest expense guidance of $860 million to $940 million, up from $640 million in the second quarter, is the quarterly cost of that debt, and it is rising faster than the adjusted operating income it has to cover. The company raised about $4.2 billion through convertible senior notes and set up an equity issuance programme covering up to 35 million shares, and the chief financial officer said the company does not intend to issue additional parent-level unsecured debt at current yields through 2027, projecting under $5 billion of parent-level debt financing over that span. The message is that the unsecured market has become too expensive, and the company will fund itself with customer prepayments, convertible notes, and new shares instead. That is a funding plan. It is also a confession that the debt is already large enough that the market is charging a premium for more of it.
The halal screen measures debt against market capitalisation, and both sides of that ratio have moved the wrong way since this site screened the company on 2 October 2026. The debt figure is higher, at almost $47 billion against $35.1 billion on the filed June balance sheet. The market capitalisation is lower, because the share price has fallen more than 40 percent in a month and about 65 percent from its year-to-date high, and a lower share price means a smaller denominator for the same debt. The October screen measured debt at 74.7 percent of market capitalisation, more than double the roughly 33 percent ceiling, and that was before the debt rose and the share price fell. This article does not compute a new ratio from the quarterly figures, because the screen is measured on audited annual figures and the share count and the market capitalisation at the quarter end are not the same as the filed balance-sheet date, and a quarterly ratio would not be the screen result. But the direction is not ambiguous, and it is worth stating plainly: the debt is bigger, the market capitalisation is smaller, and the ratio that was already 74.7 percent has moved further from the ceiling, not closer to it. A revenue beat does not shrink a debt ratio. It grows the denominator of a different ratio, the revenue ratio, and the screen does not measure that one.
The rate environment is the second pressure on the same ratio, and it is the one the rest of the week’s news supplies. Tuesday’s hot CPI print — headline prices up 0.4 percent against 0.3 percent expected, core up 0.3 percent against 0.2 percent expected, for a third straight month — put a December rate hike at near-certain, with a 15.3 percent chance of a half-point move, and the 10-year Treasury yield rose to almost 5 percent. Higher rates raise the cost of the debt CoreWeave must service and the cost of the debt it must still raise, and they raise the yield on the cash its customers hold, which is the other side of the same trade. The chief financial officer’s statement that the company will not issue parent-level unsecured debt at current yields is a direct response to that environment: the rates are too high, so the company will not borrow there. The bond market is closed on Wednesday 11 November for Veterans Day, while the stock market opens as usual, so the rate reaction to the print will pause for a day, and the stock reaction will not. The debt question, in the end, is not whether CoreWeave can borrow. It is whether it can borrow at a price that leaves anything for the shareholders after the interest is paid, and the answer this quarter is that it is choosing not to try at the parent level at all.
The screen result: CoreWeave still fails — and the fail got wider
This site screened CoreWeave on 2 October 2026, and the result was a fail, and nothing in this quarter’s print changes it. The screen has three gates, and they are applied in order. Gate one is business activity: the company must not earn its living from alcohol, tobacco, gambling, weapons, adult entertainment, pork, or conventional finance. CoreWeave passes gate one. It operates a single segment, GPU cloud and AI computing services, with no interest-based lending and no disclosed revenue from any prohibited activity, and the filing mentions past crypto-mining revenue that was discontinued before 2022 and contributes zero current revenue. Gate two is debt: total debt must be below about 33 percent of market capitalisation, with cash and cash-like securities below about 33 percent as well. CoreWeave fails gate two, decisively. Interest-bearing debt of $35.1 billion against a market capitalisation of about $46.92 billion is 74.7 percent, more than double the ceiling, and the full method is set out in the screening methodology. Gate three is non-compliant income: interest income and other non-compliant income must be below 5 percent of total revenue. CoreWeave passes gate three, at about 0.43 percent, far under the ceiling, though the figure rests on a note disclosure rather than a standalone income-statement line. The screen result is a fail, because a fail at any gate is a fail, and the debt gate is the one that decides it. The full screen is on the CoreWeave screener page, and the company sits on the screener hub with the other 938 screened names.
The third-party screeners agree, which is worth reporting because it is not always the case. Zoya publishes a CoreWeave page and currently rates it not Shariah-compliant under AAOIFI guidelines, based on the latest financial reports, consistent with this site’s debt-gate finding. No Musaffa rating page for CoreWeave could be verified from public sources, and no ShariaPortfolio rating was found. Third-party screeners apply different assumptions and update on different schedules, and their results can change when new financial statements are published, so the agreement is a cross-check, not a verdict. There is no BDS listing for CoreWeave and no moral-screen finding, so the boycott and military-ties sections that appear on other screener pages do not apply here. The screen result is a fact about a methodology, and the methodology is a choice, and the choice belongs to the investor, with a qualified scholar for the religious questions. Providers disagree on the debt ceiling itself, 30 percent versus about 33 percent, and on whether the income ceiling should be measured on total revenue or on a different base, and the disagreement is reported, not resolved, because it decides how close to the line a name really is. CoreWeave is not close to the line. At 74.7 percent, it fails under either ceiling, by more than double, and the debt has grown and the share price has fallen since the measurement, so the margin has widened, not narrowed.
The contrast with the other names in this run of articles is what makes the CoreWeave fail instructive rather than just large. Cloudflare fails at the income gate, at 6.05 percent non-compliant income, on $131.2 million of 2025 interest income, even after a 39 percent revenue beat and a 20 percent share surge, and Zoya independently rates it not compliant. AMD and Arista pass, with AMD at 0.32 percent debt and Arista at zero debt and 4.26 percent non-compliant income, the closest to the income ceiling among the recent United States passes. Nvidia passes with debt of 0.15 percent and non-compliant income of 1.07 percent, even after falling 5.5 percent on Tuesday to $168.89. The pattern is that the screen does not care about the share price, the revenue growth, or the backlog. It cares about the business, the debt, and the interest income, and CoreWeave is the name where the business is clean, the interest income is tiny, and the debt is so large that neither of the other two gates can save it. A company can beat revenue by 168 percent, raise its guidance, grow its backlog by $28 billion in a quarter, and rise 5 percent after hours, and still fail the screen by the widest margin in this run of articles, because the screen is not a measure of how fast a company is growing. It is a measure of how it is funded, and CoreWeave is funded with debt.
The AI-trade context: Nvidia fell, AMD and Broadcom rose, and the bubble question
CoreWeave did not report in a vacuum. It reported into a week in which the AI trade split down the middle, and the split is the context the print has to be read against. On Tuesday, the Philadelphia Semiconductor Index slipped just 0.36 percent, but that small move hid a sharp division inside it. Nvidia fell 5.5 percent to $168.89, Micron fell 1.10 percent to $725.11, and SanDisk fell 2.99 percent to $859.99. On the other side, Broadcom rose 3.91 percent to $331.98, AMD rose 1.30 percent to $515.21 after touching an intraday high of $539.25, a one-month high, and TSMC gained 2.02 percent. The rotation out of Nvidia into AMD and Broadcom extended a second session: on Monday, AMD had risen 13.26 percent and Broadcom 8.78 percent while Nvidia fell 4.27 percent, and on Tuesday the same pattern repeated, with money moving from the stock at the centre of the week’s bad news toward its rivals. The split is not about the chips themselves. It is about which company investors want to own while Nvidia’s board probe, the export-control debate, and the inflation report all hang over the market at once. The week before had wiped more than $2 trillion off the market on Thursday 5 November, with the Nasdaq down 5.52 percent and the semiconductor index down 12.25 percent, and Friday’s bounce took back about a fifth of it. The tape is still trading the same argument it has been trading for a week: whether artificial intelligence spending can ever pay for itself, and what the cost of money will be while the market waits for the answer.
CoreWeave’s print is the most direct evidence yet on the first half of that argument, and it cuts both ways. The revenue beat and the backlog growth say demand for AI computing is still outrunning supply, and the raised guidance says the company expects that to continue. The widening loss, the $14.5 billion of quarterly capital spending, and the almost $47 billion of debt say the supply is being built with borrowed money at rising rates, and the chief financial officer’s refusal to issue more parent-level unsecured debt at current yields says the borrowing has become expensive enough to avoid. The bubble question is not whether demand exists. The backlog proves it does. The bubble question is whether the demand is worth the price being paid to serve it, and the price is $37 billion to $39 billion of capital spending this year against $13 billion to $13.3 billion of revenue, funded with debt that costs close to a billion dollars a quarter in interest. The market’s answer on Tuesday night was to buy the stock 5 percent higher. The market’s answer over the past month, in which the stock fell more than 40 percent, was to sell it. Both answers are in the price, and the next two reports will say which one holds: Nebius on Thursday, and Nvidia on 17 November, the last of the Magnificent Seven to report and the one whose chips fill CoreWeave’s data centres.
There is also a development outside the earnings calendar that frames the whole AI trade this week, and it is worth naming because it will be the next test of whether the demand CoreWeave is seeing is durable. Anthropic, the AI company whose models run on the infrastructure CoreWeave and its peers build, is pressing ahead with an initial public offering at a valuation of approximately $1.3 trillion, with shares expected to begin trading next Thursday following investor meetings this week. An offering of that size is a litmus test for the AI trade broadly: it asks the public market to put a price on a company that has never been public, at a valuation larger than most of the companies in the S and P 500, and the answer will say something about how much appetite remains for AI exposure at any price. CoreWeave’s own IPO, in March 2025 at $40 a share, is the smaller version of the same test, and the stock has traded the whole arc of the AI argument since then, from the highs of the year to a fall of about 65 percent. The Anthropic offering will not change CoreWeave’s screen result, and it will not change its debt, but it will change the price the market is willing to pay for AI growth, and the price is the denominator of the debt ratio. A market that pays more for AI growth makes the debt ratio smaller. A market that pays less makes it bigger. The screen is measured on audited annual figures, not on prices, so the IPO will not change the result today. It will change the conditions under which the next refresh is measured.
How a halal investor should approach CoreWeave
A 168 percent revenue beat triggers a specific set of instincts, and they are worth naming before acting on any of them. The first is to buy the beat, on the theory that a company growing revenue 168 percent must be a good investment. The second is to buy the dip, on the theory that a stock down more than 40 percent in a month and 65 percent from its high must be cheap. The third is to buy the backlog, on the theory that $132.5 billion of contracted revenue must be worth something. The fourth is to buy the story, on the theory that artificial intelligence is the future and the companies that build it must win. None of the four is a framework. A framework is a set of questions, answered in order, before any order is placed, and for CoreWeave the questions are these.
First, screen before you react. The screen result for CoreWeave is a fail, as of 2 October 2026, at the debt gate, at 74.7 percent of market capitalisation against a roughly 33 percent ceiling, and a print at the close on Tuesday does not change it, because the screen is measured on audited annual figures, not on prices and not on headlines. A halal investor who buys CoreWeave because the revenue beat was 168 percent has let the headline do the screening, and the headline is not an input to the screen. Second, separate the growth from the funding. The revenue grew 168 percent. The debt grew to almost $47 billion. The capital spending was $14.5 billion in a quarter. The interest expense was guided at $860 million to $940 million. The growth and the funding are the same story told from opposite sides, and the screen measures the funding side. Third, do not trade a headline. The backlog will be revised, the next quarter will arrive in three months, and the Federal Reserve will meet on 8 and 9 December with more data than it has today. Fourth, remember that you do not use margin. A halal portfolio carries no margin debt, which means no lender can liquidate it on a bad morning, and a 5 percent after-hours rise is a price move, not a forced decision. Fifth, watch the right dates: the Federal Reserve’s December meeting, the Bank of Canada’s decision on 9 December, Nebius on Thursday, Nvidia on 17 November, and the 2026 10-K, which is where the debt figure will actually appear. Sixth, recheck at the refresh. If the ratios move, the next screen refresh will show it in the numbers. The framework is the position: screen first, separate the growth from the funding, do not trade a headline, no margin, watch the right dates, and wait for the refresh.
There is one more question the framework asks, and it is the one that matters most for a company like CoreWeave, because it is the question the screen cannot answer. The screen measures the debt, the business, and the interest income. It does not measure whether the business model works. A company that borrows almost $47 billion to build capacity, rents the capacity on long-term contracts, and targets adjusted profitability in the first half of 2028 is a company whose model depends on the contracts lasting longer than the debt, and on the customers paying in advance, and on the hardware arriving on time, and on the rates not rising faster than the revenue. Those are business questions, not screen questions, and they belong to the investor, with the screen as the first filter and the business analysis as the second. The screen says fail. The business analysis says the model is unproven at this scale, funded with debt, and dependent on a small number of very large customers. Neither answer is a buy or sell call. Both are facts, and the framework is to hold both of them at once, and to wait for the refresh.
What to watch next
The next number that matters for CoreWeave is not CoreWeave’s. It is Nebius, the other neocloud, which reports on Thursday, and which will say whether the demand CoreWeave is seeing is an industry-wide phenomenon or a CoreWeave-specific one. If Nebius reports the same pattern — revenue up sharply, backlog up sharply, loss widening, spending up sharply — then the demand story is an industry story, and the financing story is an industry story too. If Nebius reports something different, then the CoreWeave print is a company story, and the two stories have different implications. The earnings calendar does not let up after that. Nvidia reports on 17 November, the last of the Magnificent Seven to report, a week after Thursday’s rout and Friday’s bounce, and after the board probe that knocked the stock down on Monday and again on Tuesday, and it is the biggest one on the calendar because its chips fill CoreWeave’s data centres and its customers sign CoreWeave’s contracts.
Closer to the screen, watch the dates that decide the cost of money. The Federal Reserve meets on 8 and 9 December, and the Bank of Canada decides on 9 December as well, the same week, and the two decisions will be read together by Canadian investors, because Canadian rates set the cost of money at home, and American rates set the cost of money for the companies in a Canadian portfolio. The December hike is now near-certain after Tuesday’s hot CPI print, with a 15.3 percent chance of a half-point move, and the January pricing has a cumulative 50-basis-point hike at 48.8 percent. A steeper rate path raises the cost of the debt CoreWeave must service and the cost of the debt it must still raise, and the chief financial officer has already said the company will not issue parent-level unsecured debt at current yields. The 2026 10-K, due in early 2027, is the document that will decide whether the debt ratio improves or worsens, and the next screen refresh is where the ratio will show up in the numbers. And watch the Anthropic initial public offering, expected to begin trading next Thursday at a valuation of approximately $1.3 trillion, because it will say how much appetite the public market still has for AI exposure at any price, and the price is the denominator of the debt ratio. For readers who hold the names in this article, the practical next step is the same as it was after Friday’s jobs report: note the screen result as it stands, note the margin, and recheck both at the refresh. Run it through the screening methodology, check the company on the screener hub, and build the rest of the decision from there. A print is an evening. The screen is the house. Read the print, and keep the house in order.
FAQ
What did CoreWeave report in the third quarter of 2026?
Revenue of $3.66 billion, up 168 percent from about $1.36 billion a year earlier, ahead of the $3.55 billion analysts expected. The GAAP net loss widened to $709 million, or $1.27 a share, from a $110 million loss a year earlier, while adjusted operating income was $379 million. Management raised full-year 2026 guidance to revenue of $13.0 billion to $13.3 billion and adjusted operating income of $1.275 billion to $1.375 billion, and reported a revenue backlog of $132.5 billion at 30 September, up from $104.2 billion at the end of June.
Why did the shares rise if the loss widened?
Because the market was testing demand, not profitability. The revenue beat and the raised guidance, plus a backlog that grew by about $28 billion in a quarter, told investors that customers are still signing long-term contracts for AI computing faster than CoreWeave can build it. The shares were quoted about 5 percent higher after hours at $51.38, after falling more than 40 percent in a month, so the rise was also a bounce from a beaten-down level. Management said adjusted profitability should arrive in the first half of 2028.
How much debt does CoreWeave have?
Almost $47 billion at the end of the quarter, according to reporting on the results, up from $35.1 billion of interest-bearing debt on the filed balance sheet at 30 June 2026. Its own guidance put third-quarter interest expense at $860 million to $940 million, up from $640 million in the second quarter. The chief financial officer said the company does not intend to issue additional parent-level unsecured debt at current yields through 2027, and projected under $5 billion of parent-level debt financing over that span.
Does CoreWeave pass the halal screen?
No. This site’s screen of 2 October 2026 rates CoreWeave a fail at the debt gate: interest-bearing debt of $35.1 billion against a market capitalisation of about $46.92 billion is 74.7 percent, more than double the roughly 33 percent ceiling. The business gate passes, because GPU cloud and AI computing services carry no disclosed alcohol, gambling, pork, tobacco, weapons or adult-entertainment revenue, and non-compliant income passes at about 0.43 percent, far under the 5 percent ceiling. Zoya independently rates CoreWeave not Shariah-compliant. There is no BDS listing and no moral-screen finding for the company.
What is the $132.5 billion backlog — and why does it not settle the debt question?
The backlog is contracted future revenue, mostly from committed customer contracts, and it grew from $104.2 billion at the end of June. It is a demand signal, not cash in hand: the revenue arrives only as CoreWeave delivers the computing capacity, and delivering it costs money — $14.5 billion of capital spending in the third quarter alone, and $37 billion to $39 billion planned for the full year. The backlog explains why lenders and customers keep funding the build-out. It does not change the screen, which measures debt against market capitalisation and interest income against revenue, not orders against anything.
What should a halal investor watch next on CoreWeave?
Three things. First, the next screen refresh, where the filed annual figures will show whether the debt ratio improved or worsened — the debt figure is already higher than the $35.1 billion measured on 2 October, and the share price has fallen more than 40 percent in a month, so the direction is the wrong way. Second, the December Federal Reserve meeting on 8 and 9 December, because a hike raises the cost of the debt CoreWeave must service. Third, Nebius, the other neocloud, which reports on Thursday, and Nvidia’s own report on 17 November, which is the next real test of whether AI demand is still outrunning AI spending.