CoreWeave Revenue Jumps 168% to $3.66 Billion as Backlog Hits $132.5 Billion — But the Loss Widens to $709 Million

CoreWeave, the company that rents out the computing power behind the artificial intelligence boom, has reported a quarter that captures the whole argument about the AI industry in a single set of numbers. Revenue more than doubled and a half again, the order book swelled to a record, and the stock jumped by almost a quarter in a day. At the same time, the loss grew sharply, the debt pile kept climbing, and the bill for building new data centres reached a scale that would have been hard to imagine even a year ago. The results, released on November 10, 2026, landed in the middle of a fierce market debate about whether the trillions of dollars flowing into AI infrastructure are building the next great computing platform or inflating the next great bubble, and the numbers gave both sides something to point to. This is the story of a company growing faster than almost any other in corporate America, spending even faster, and asking investors to believe that the gap between the two will close — a story best summed up by the CoreWeave Q3 revenue $3.66 billion November 2026 report.

1. The Headline Numbers: Revenue Soars, Losses Widen

Few companies of CoreWeave’s size have ever grown as fast as it did in the third quarter of 2026, and few have lost as much money while doing it. The two facts belong together, because they describe the same business model from opposite ends: a company that buys enormous quantities of advanced computing hardware, installs it in data centres, and rents it to customers building artificial intelligence systems, and that pays for the hardware largely with borrowed money. When demand is strong, revenue can multiply in a year. When interest rates and construction costs are high, the financing bill can swallow the operating profit whole. The third quarter, which ended on September 30, 2026, delivered both sides of that bargain at full strength, and the market’s first reaction was to celebrate the growth and worry about the cost later.

1.1 Revenue up 168% to $3.66 billion

CoreWeave reported third-quarter revenue of $3.66 billion, an increase of 168 percent on the same quarter a year earlier, when revenue was about $1.37 billion. The figure beat the analyst consensus of roughly $3.54 billion, and it was more than $1 billion higher than the revenue of the preceding quarter, a sequential jump that underlines how quickly the company’s capacity is expanding. In plain terms, the business is now generating in three months roughly two and a half times what it generated in the same three months last year, a rate of growth that places it among the fastest-expanding large companies in the world. The growth was driven, the company said, by strong demand for its specialised AI cloud platform, the service through which customers rent access to clusters of advanced graphics processors rather than buying the hardware themselves. That demand has been the engine of CoreWeave’s rise since it pivoted from cryptocurrency mining to AI infrastructure, and the third quarter showed no sign of it slowing. For context, the revenue of a single quarter now exceeds what many established technology companies earn in a full year, and it arrives from a company that was a niche player only a few years ago. The scale of the increase also matters because it answers, at least for now, the question of whether customers are still willing to pay for AI computing at the pace they were a year ago. They are, and they are paying more of it, faster.

1.2 A $709 million net loss despite operating gains

The other side of the quarter was less flattering. CoreWeave’s net loss widened to $709 million, or $1.27 per share, compared with a loss of $110 million, or $0.22 per share, in the same quarter of 2025. The loss grew even as the underlying operating business improved markedly: operating income on a generally accepted accounting principles basis rose to $207 million from $52 million a year earlier, lifting the operating margin from 4 percent to 6 percent, and adjusted operating income reached $379 million, about 10 percent of revenue, beating the top end of management’s own guidance of $200 million to $260 million. Adjusted earnings before interest, tax, depreciation and amortisation, a measure the company uses to show the cash-generating power of its operations before financing costs, came in at $2.26 billion, a margin of 62 percent. The gap between a healthy operating performance and a large net loss is explained almost entirely by what happens below the operating line, and in CoreWeave’s case that means interest and depreciation on the vast sums it has borrowed and spent. On an adjusted basis, stripping out some items, the net loss was $529 million, against $41 million a year earlier, and the adjusted net loss margin widened from 3 percent to 14 percent. The pattern is familiar from the company’s recent history: the machines earn good money, but the cost of owning the machines is heavier still. Whether that pattern reverses depends on whether revenue keeps growing faster than the financing bill, and the third quarter left that question open.

2. The Backlog: $132.5 Billion of Future Demand

If revenue tells the story of the past three months, the backlog tells the story of the years ahead, and in the third quarter the backlog was the number that most excited the company’s supporters. A backlog in this business is the total value of signed customer contracts for computing capacity that has not yet been delivered or billed, and for CoreWeave it has become the single most watched figure in its reports, because it is the closest thing the market has to a forward view of demand. The backlog does not guarantee that every dollar will arrive on schedule, and contracts can be renegotiated, delayed or, in extreme cases, cancelled. But a backlog that keeps setting records is hard to dismiss as a mirage, and the third quarter set another one.

2.1 What the backlog actually means

CoreWeave reported a revenue backlog of $132.5 billion as of September 30, 2026, a record high for the company. To put that figure in perspective, it is more than ten times the revenue the company expects to generate in the whole of 2026, and it represents years of future work already under contract. The backlog is built from multi-year agreements with customers who need guaranteed access to AI computing capacity and are willing to commit to paying for it in advance, sometimes years ahead of delivery. For the customers, the commitment secures scarce capacity in a market where the most advanced chips are in short supply. For CoreWeave, it provides visibility over future revenue and, crucially, a basis on which lenders are willing to finance new data centres, since the debt is backed by contracted future income rather than by hope. That is the logic of the whole business: sign the customer first, borrow against the contract, build the capacity, and collect the revenue over time. The backlog is the starting point of that chain, and its size is the reason the company can keep borrowing at the pace it does. It is also, however, a figure that invites scrutiny, because not all backlog is equal. Some of it depends on data centres that have not yet been built, power supplies that have not yet been connected, and chips that have not yet been delivered, and the value of the backlog will only turn into revenue if all of those pieces arrive on time.

2.2 How fast the backlog is growing

The backlog grew 138 percent from a year earlier and about 27 percent from the end of the second quarter, when it stood at $104 billion, an increase of roughly $28.5 billion in three months. That quarterly increase alone is larger than the company’s entire annual revenue only a short time ago, and it suggests that customers are not merely renewing existing arrangements but signing substantially larger and longer commitments. Management pointed to strong conversion of backlog into active capacity as one reason it raised its full-year guidance, a sign that the pipeline is not just growing on paper but turning into working data centres. The growth also reflects the structure of the AI market, in which a small number of very large customers account for an outsized share of demand, and in which those customers are competing with each other to secure computing power for the next generation of models. When one large customer signs a multi-billion-dollar, multi-year deal, the backlog can jump in a single step, and the third quarter appears to have contained several such steps. The question for the quarters ahead is whether the pace can continue, because the backlog can only keep growing at this rate if customers keep signing deals of ever-larger size, and at some point the pool of customers able to make such commitments is finite. For now, though, the direction is unmistakable, and it is the strongest evidence the company can offer that demand for its services is not a passing fashion.

3. The Debt Question: $47 Billion and Rising Interest Costs

Every discussion of CoreWeave eventually arrives at the same place: the debt. The company has borrowed on a scale that would be extraordinary for a business of any age, let alone one that has only recently become a household name on Wall Street, and the third quarter showed both why it borrows and what the borrowing costs. The debt is not an accident or a sign of distress in the ordinary sense. It is the deliberate funding mechanism of the business model, the way the company pays for data centres and chips before the revenue from them arrives. But the size of the bill is now so large that it shapes everything else in the results, and it is the first thing sceptics point to when they question whether the AI boom can pay for itself.

3.1 Interest expense now exceeds operating income

CoreWeave’s net interest expense reached $846 million in the third quarter, up from $310 million in the same quarter a year earlier, a near-tripling that reflects both the growth of the debt pile and the cost of servicing it. The company’s total debt stood at about $47 billion at the end of the quarter. The interest bill alone was more than four times the company’s GAAP operating income of $207 million, which is why a business whose operations are improving can still report a widening net loss. Interest is, in effect, the price of growing this fast with borrowed money, and that price is rising: management forecast fourth-quarter interest expense of $1.05 billion to $1.15 billion, higher again than the third quarter. On a free cash flow basis, the company lost $7.7 billion in the quarter, a figure that captures the cash going out the door for new infrastructure after accounting for the cash coming in from operations. None of this is hidden from investors; it is disclosed plainly in the earnings release and discussed on the earnings call. The company’s argument is that the spending buys assets that will generate revenue for years, and that the interest burden will shrink as a share of revenue as the business scales. The counter-argument is that $47 billion of debt, carrying more than $3 billion a year in interest at the current run rate, leaves little room for error if demand slows, if a large customer falters, or if the cost of borrowing rises further. The third quarter did not settle that argument, but it made the stakes clearer: the interest bill is now one of the largest single items in the company’s accounts, and it will get bigger before it gets smaller.

3.2 The $9 billion prepayment plan

One of the most closely watched disclosures in the third quarter was the company’s expectation of receiving approximately $9 billion in customer prepayments and deposits over the next three months. Prepayments matter because they change the funding equation. When a customer pays in advance for computing capacity, CoreWeave receives cash before it has to spend it on the hardware, reducing the amount it needs to borrow and easing the pressure on its balance sheet. The $9 billion figure, if it arrives as expected, would be a substantial inflow, equivalent to more than two quarters of current revenue, and it would arrive at a moment when the company is spending heavily on new capacity. Management presented the prepayments as evidence of customer commitment, and in a sense they are: a customer does not hand over billions of dollars in advance unless it is confident it will need the capacity and confident the supplier will deliver it. But prepayments also create obligations. The money has to be earned by delivering the computing power, on schedule, in data centres that must be built, powered and cooled, and any delay in delivery turns an advance payment into a liability rather than a windfall. The chief financial officer, Nitin Agrawal, also indicated on the earnings call that the company does not intend to issue additional unsecured debt at the parent-company level through 2027 at current yields, a signal that management is aware of the cost of borrowing and is looking for other ways to fund growth. Between the prepayments, the backlog and the operating cash the business generates, the company has several sources of funding beyond new debt, and how those sources balance out over the next year will be one of the clearest tests of the business model.

4. Spending to Grow: $14.5 Billion of Capital Expenditure

If the debt is the cost of CoreWeave’s growth, the capital expenditure is where the money actually goes, and in the third quarter it went out at a pace that startled even seasoned observers of the AI infrastructure boom. Capital expenditure is the money a company spends on long-lived assets, and for CoreWeave that means almost entirely data centres, the power systems that feed them, the cooling systems that keep them running, and the advanced processors inside them. The third-quarter figure was so large that it exceeded the company’s entire quarterly revenue by a factor of four, a ratio that illustrates just how capital-hungry the AI cloud business has become.

4.1 Where the money is going

CoreWeave reported company-defined capital expenditure of $14.511 billion in the third quarter, compared with $1.85 billion in the same quarter a year earlier, an almost eightfold increase. The spending pays for the physical build-out of AI capacity: land and buildings for new data centres, the electrical infrastructure to deliver tens or hundreds of megawatts to each site, the cooling systems needed to keep thousands of high-powered processors from overheating, and the processors themselves, which are among the most expensive components in modern computing. The company is expanding capacity to meet the demand reflected in its backlog, and the lead times are long: a data centre can take a year or more from breaking ground to switching on, and the chips inside it must be ordered months in advance. That is why the spending runs so far ahead of the revenue it will eventually produce, and why the company’s cash flow is deeply negative even while its operating margins improve. The spending also reflects a competitive race. CoreWeave is not the only company building AI data centres, and capacity that arrives late can mean a lost customer, so there is a strong incentive to build ahead of demand rather than behind it. The risk is that capacity built ahead of demand becomes a burden if demand disappoints, and with $14.5 billion spent in a single quarter, the cost of being wrong is now very large indeed.

4.2 The full-year spending plan

For the full year, CoreWeave raised its capital expenditure guidance to $37 billion to $39 billion, a figure that dwarfs the revenue it expects to generate in the same period and that underlines how much of the company’s future depends on the bets it is placing now. The guidance implies that the fourth quarter will see spending on a similar scale to the third, as the company races to bring new capacity online. Management’s case is that the spending is matched by contracted demand, with the $132.5 billion backlog providing the cover for the outlay, and that the assets being built will generate returns for years. The company’s adjusted operating margin, which rose from about 1 percent in the first quarter to 5 percent in the second and above 10 percent in the third, is presented as evidence that the model improves with scale, and that as more capacity comes online the revenue it generates will begin to cover the costs more comfortably. Sceptics note that the same argument has been made about many capital-intensive booms, and that the test comes when the building stops and the bills still have to be paid. The guidance also raises a practical question about execution: spending $37 billion to $39 billion in a year requires finding sites, securing power, hiring engineers and managing construction on a scale that few companies have ever attempted, and any bottleneck, from grid connections to chip supply, can delay the revenue that the spending is meant to produce. The fourth quarter will show whether the pace can be sustained without stumbling.

5. The Market Reaction: Shares Jump After the Report

The stock market’s verdict on CoreWeave’s third quarter was swift and emphatic, and it told its own story about how investors are weighing the company’s growth against its costs. Shares that had been under pressure in the weeks before the report, as doubts about the sustainability of AI spending spread through the market, surged once the numbers were out, rewarding those who had bet that the demand story was still intact and punishing those who had bet against it. The move was large even by the standards of a stock known for violent swings, and it turned the earnings report into one of the most talked-about market events of the week.

5.1 The stock’s wild ride since the IPO

CoreWeave’s shares rose in after-hours trading on Tuesday, November 10, immediately after the results were released, and then climbed more than 15 percent in premarket trading on Wednesday before closing the day up 24.71 percent. The scale of the move reflects both the strength of the numbers and the state of the stock beforehand, since a heavily debated and heavily traded name can move sharply when an earnings report resolves a question that had been hanging over it. CoreWeave listed on the stock market earlier in 2025, and its shares have been among the most volatile of any large company since, swinging on every new piece of news about AI demand, chip supply, interest rates and the health of its biggest customers. The company sits at the centre of the market’s argument about artificial intelligence: if the AI buildout continues, the providers of computing capacity are among the first to benefit; if it stalls, they are among the first to suffer, because their costs are fixed and their debts are large. That is why the stock can rise by a quarter in a day on a strong backlog number and fall just as fast on a hint of slowing demand. The third-quarter jump does not settle the argument, but it shows that, for now, the market is willing to pay up for growth, and that the demand side of the story remains persuasive to a large number of investors. It also shows how much of the company’s value depends on sentiment, since a single quarter’s numbers, however strong, cannot by themselves justify a move of this size unless they change what investors believe about the future.

5.2 What the chief executive said about Wall Street’s doubts

The earnings call also gave CoreWeave’s chief executive, Michael Intrator, a chance to answer the sceptics directly, and he did not hold back. According to reports of the call, Intrator said that Wall Street had not understood CoreWeave, a remark that captured the frustration of a management team that believes the market has focused on the debt and the losses while missing the strength of the demand and the quality of the backlog. The point he was making, in substance, is that the company’s critics are applying the logic of a conventional business to an unconventional one: a company that rents out scarce computing capacity on multi-year contracts is not the same as a company selling a product into an uncertain market, and the backlog, the prepayments and the rising operating margins are, in his telling, the evidence that the model works. Whether that argument convinces the market over time will depend on the numbers in the coming quarters, and in particular on whether the operating margin keeps climbing towards the levels that would make the interest bill manageable. Intrator’s confidence is understandable: the quarter delivered record revenue, a record backlog, better-than-guided operating income and raised guidance, a combination that few chief executives get to report. But confidence is not the same as proof, and the history of fast-growing, debt-fuelled companies is full of confident management teams whose models worked until they suddenly did not. The market’s reaction suggests that, for one day at least, the chief executive’s version of events carried the day.

6. The Guidance: A Bigger 2026 Than Planned

Beyond the numbers for the quarter just ended, CoreWeave’s management also told investors to expect more from the rest of the year than they had previously forecast, raising its guidance for revenue, operating income and the annualised revenue run rate at which it expects to exit the year. Guidance matters because it is management’s own forecast, and raising it is a statement of confidence that is hard to take back. It also matters because it sets the bar against which the next quarter will be judged, and a raised bar that is then missed can be more damaging than a cautious forecast that is beaten.

6.1 Revenue raised to $13 billion-plus

CoreWeave raised its full-year 2026 revenue forecast to between $13.0 billion and $13.3 billion, up from the previous range of $12.4 billion to $13.2 billion, an increase that reflects both the strength of the third quarter and the pace at which backlog is converting into active, revenue-generating capacity. The company also raised its guidance for adjusted operating income to between $1.275 billion and $1.375 billion, about 26 percent higher at the midpoint than the previous guidance, a sign that management expects the operating leverage seen in the third quarter to continue. For the fourth quarter specifically, the company guided to revenue of $4.7 billion to $5.0 billion, which would be another large sequential step up from the third quarter’s $3.66 billion and would mean that the final three months of the year generate more revenue than the whole of 2025’s first half. The guidance implies that the company’s growth is not a one-quarter event but a trajectory, driven by data centres coming online and contracts beginning to bill. It also implies, however, that the spending will continue at a similar pace, since the revenue growth depends on capacity that has to be built and paid for first. Investors who focus on the revenue line will find much to like in the new guidance; those who focus on the cash flow will note that the higher revenue forecast does not by itself solve the funding question, since the money to build the capacity that generates the revenue still has to come from somewhere.

6.2 The $20 billion run-rate target

Perhaps the most striking figure in the guidance was the company’s projection for its exit annualised revenue run rate, the revenue it would generate in a year if it continued at the pace expected at the end of December. CoreWeave raised that target to between $20 billion and $21 billion, a level that would make it, by revenue, one of the larger infrastructure companies in the world, and that would represent a more than fivefold increase on the revenue of two years earlier. The run-rate figure is based on contracted revenue and the company’s capacity deployment plans, and it is meant to show investors the scale the business is heading towards rather than the scale it has already reached. It is also a figure that depends heavily on execution: the capacity has to be built, the power has to be connected, the chips have to arrive, and the customers have to take delivery on schedule. Any slippage in that chain pushes the run rate out, and the history of large infrastructure projects is full of delays. But if the target is met, it would mark a transformation in the company’s standing, from a fast-growing challenger to a major infrastructure provider in its own right, with a revenue base large enough to begin absorbing the interest bill that has so far outweighed its operating income. The run rate is, in that sense, the number on which the whole investment case rests: the company is asking investors to value it not on what it earns today, but on what it will earn when the capacity now being built is fully working, and the $20 billion to $21 billion target is the clearest statement yet of what that future is supposed to look like.

7. The Bigger Question: Is the AI Buildout Sustainable?

CoreWeave’s third quarter arrived at a moment when the question hanging over the entire artificial intelligence industry is no longer whether the technology works, but whether the money being spent on it will ever come back. Trillions of dollars are being committed to data centres, chips and power generation around the world, and the companies doing the spending are, for the most part, borrowing heavily to do it. CoreWeave is the purest expression of that bet: a company whose entire business is building the infrastructure on which AI depends, and whose accounts show, in unusual clarity, both the scale of the opportunity and the scale of the risk. The debate about its future is, in miniature, the debate about the whole AI economy.

7.1 The case that demand is real

The case for CoreWeave, and for the AI buildout more broadly, starts with the backlog. Customers do not sign $132.5 billion of multi-year contracts for computing capacity they do not expect to use, and they do not pay billions of dollars in advance for capacity they do not need. The demand is coming from the developers of the largest AI models, from cloud providers racing to offer AI services to their own customers, and from enterprises beginning to deploy AI in their operations, and each of those groups has reason to secure capacity early, because the most advanced chips are scarce and the data centres that house them take time to build. The revenue growth supports the demand story too: a 168 percent increase, a beat against analyst expectations, and a sequential jump of more than $1 billion are not the marks of a market that is cooling. The improving operating margin, from about 1 percent in the first quarter to above 10 percent in the third, suggests that the business gets more profitable as it gets bigger, which is what should happen if the underlying economics are sound. And the raised guidance, including the $20 billion to $21 billion exit run rate, indicates that management, which has the best view of the order book, expects the demand to continue. For supporters, the losses and the debt are the temporary costs of building a position in a market that will be enormous, and the right comparison is not with a mature company but with the builders of the railways, the electricity grids or the early internet, all of which consumed vast amounts of capital before they produced reliable profits.

7.2 The case that the spending cannot last

The case against is equally straightforward, and it starts with the same numbers. A company that loses $709 million in a quarter, carries $47 billion of debt, spends $14.5 billion in three months and burns $7.7 billion of free cash flow is a company that depends entirely on the continued willingness of lenders and customers to keep funding it, and that willingness is not guaranteed. The interest bill alone, at more than $846 million a quarter and rising, will consume an ever-larger share of revenue if growth slows, and the history of leveraged infrastructure booms is that they end not when demand disappears overnight but when the cost of funding rises, a customer delays a project, or a lender asks for its money back. The backlog, impressive as it is, is concentrated among a small number of very large customers, and the fortunes of those customers are themselves tied to the same AI bet, which means the risk is correlated rather than spread. If one large customer were to slow its spending, the effect on CoreWeave would be immediate and large. There is also the question of what happens to the price of computing capacity as more of it is built. Today’s high prices reflect scarcity; as the industry adds capacity at the current pace, scarcity will ease, and with it the prices and margins that make the current model work. The sceptics do not need to argue that AI is a fad to be worried. They need only argue that the rate of spending cannot be sustained, that the returns will arrive later and smaller than expected, and that the companies that borrowed the most to build the fastest will be the most exposed when the music slows. The third quarter gave both sides new evidence, and neither side has yet been proved wrong.

8. What to Watch Next

The third quarter answered some questions and raised others, and the answers will come in the months ahead, in the fourth-quarter results, in the delivery of the prepayments, and in the pace at which new capacity comes online. CoreWeave is now a company whose every quarter is a referendum on the AI boom, and the next few will be watched more closely than most.

8.1 The fourth quarter and the prepayments

The most immediate test is the fourth quarter, for which the company has guided to revenue of $4.7 billion to $5.0 billion and interest expense of $1.05 billion to $1.15 billion. Meeting the revenue guidance would confirm that the backlog is converting into billed capacity at the pace management claims, while the interest figure will show whether the cost of funding is stabilising or still climbing. The $9 billion of customer prepayments expected over the next three months is the other number to watch, because it will show whether customers are as committed in cash as they are on paper, and because the inflow would materially change the company’s funding position. Any shortfall in the prepayments, or any delay in their arrival, would raise questions about the strength of the demand story, while their arrival on schedule would be a powerful confirmation of it. Beyond the numbers, investors will watch for signs of how the capacity build-out is progressing: whether data centres are opening on time, whether power connections are keeping pace, and whether the chip supply is arriving as planned. The company’s full-year capital expenditure guidance of $37 billion to $39 billion implies a fourth quarter of spending on a similar scale to the third, and the execution of that spending, without delays or cost overruns, is itself a test of management’s ability to operate at this scale. The fourth quarter will also bring the first full reading of how the company is managing its debt, including whether it can fund its growth without returning to the unsecured debt markets it has said it will avoid through 2027.

8.2 The Bottom Line

CoreWeave’s third quarter was a quarter of superlatives: record revenue, a record backlog, a record loss, record spending and a record interest bill, all at once. The company grew 168 percent, beat expectations, raised its guidance and saw its shares jump by almost a quarter in a day, and it did all of that while losing $709 million, carrying $47 billion of debt and spending $14.5 billion on new capacity in three months. That combination is the story of the AI infrastructure boom in a single earnings report, and it is why the company has become a symbol of both the promise and the peril of the moment. The bulls see a business with demand stretching years into the future, margins improving with scale, and a revenue run rate heading towards $20 billion. The bears see a business that burns cash faster than it earns it, owes more than it is worth on some measures, and depends on a boom that cannot last forever at this pace. The truth, as usual, will probably lie somewhere between the two, and it will be decided not by this quarter’s numbers but by the ones that follow. What is certain is that CoreWeave is no longer a footnote in the AI story. It is one of its central characters, and its fortunes will say a great deal about whether the trillions being spent on artificial intelligence are building the future or financing a bubble. The coming days will provide the answers.

Comments

comments