Nebius Beat by 8% and Still Fell 18% on a $27 Billion Spending Bill: the Halal Investor's Read (2026)
This is the third beat-and-fall of the earnings season, and the clearest one yet about what the market is actually pricing. Nebius Group reported third-quarter revenue of $980.8 million before the open on Thursday 12 November 2026, up 571 percent from $146.1 million a year earlier and above the $907.5 million consensus tracked by LSEG and cited by CNBC, with adjusted EBITDA of $465.8 million against a loss of $5.2 million a year earlier and a net loss that narrowed to $37.5 million from $119.6 million. The shares opened down about 10 percent, fell about 16 percent by mid-morning according to CNBC, and were down about 18 percent by late morning, to about $127 from Wednesday's close of $154.69. Nothing in the quarter itself explains the fall. Everything in the guidance does. Nebius raised its 2026 capital spending plan to about $27 billion from a $20 to $25 billion range, guided 2027 spending at $40 to $45 billion against about $28.8 billion expected, pushed 2026 revenue to the low end of its $3.0 to $3.4 billion range, and pushed some contracted capacity, including a New Jersey site for Microsoft, into 2027. This article separates the quarter from the guidance, lays out the spending bill in full, and reads the story the way a halal investor should: what the screen says about Nebius, why the debt gate is the one to watch when a company plans to spend more than its own market value in a single year, and the framework for holding a passing company whose pass is getting thinner.
Updated November 2026 · Reviewed by the Canadian Halal Investor team
What happened: a beat, a bill, and an 18% fall
Nebius reported before the open on Thursday, and the shape of the morning was set within minutes. The revenue beat was real and large: $980.8 million against about $907.5 million expected, an 8 percent beat on a base that grew 571 percent year on year and 68 percent from the second quarter's $582.3 million. Adjusted EBITDA was $465.8 million, against a loss of $5.2 million in the same quarter a year earlier, and the company kept its full-year adjusted EBITDA margin forecast at about 40 percent. The net loss from continuing operations narrowed to $37.5 million, or 13 cents a share, from $119.6 million, or 50 cents a share, a year earlier, and adjusted net profit was $90.9 million, against a market expectation of a $271.4 million loss, according to Zhitong Finance. On the numbers alone, this was a beat across the board, and the shares fell anyway. They were down about 10 percent in premarket trading, down about 16 percent by mid-morning according to CNBC, and down about 18 percent by late morning, trading at about $126.29 at 10:53 a.m. Eastern according to Webull data, and about $127.49 at 11:50 a.m. Eastern. The market did not dispute the quarter. It repriced the plan.
The plan is what changed, and it changed in three places at once. First, the spending bill got bigger: 2026 capital expenditure is now about $27 billion, up from a $20 to $25 billion range, above the $23.85 billion FactSet expected and the roughly $22 billion analysts expected on LSEG data cited by Reuters, and 2027 capital expenditure is guided at $40 to $45 billion, against a FactSet expectation of $28.77 billion. Second, the revenue got smaller in the near term: 2026 revenue is now expected at the low end of the $3.0 to $3.4 billion range, against a FactSet consensus of $3.35 billion, near the top of the range the company now says it will only reach at the bottom. Third, the timetable slipped: some third-party data-centre capacity expected this year is now scheduled for the first half of 2027, and the company said it will reach its target of more than $7 billion in annual run-rate revenue about two months later than planned, in February rather than December. Founder and chief executive Arkady Volozh said in his shareholder letter that the slippage pushes revenue into the following year, and that the contracts behind the delayed capacity remain in place. The market heard the second and third changes as a warning and the first as a bill, and it priced all three at once.
There is a pattern to name here, because this is the third time this earnings season that a beat has been followed by a fall, and the pattern is now the season's defining feature. Apple and Palantir both beat and fell on 2 November, Nvidia's board story moved its shares without a number attached, and now Nebius has beaten by 8 percent and fallen 18 percent. The pattern is not that beats do not matter. It is that the market is no longer grading the quarter. It is grading the funding: whether the spending that produces the next quarter's growth can be paid for, at what price, and on whose balance sheet. Nebius is the purest test of that question so far, because it is the smallest of the artificial intelligence infrastructure names, the fastest growing, and the one whose spending plan is largest relative to its own size. The quarter was the best it has ever reported. The bill was the largest it has ever presented. The market chose the bill.
The context around the print matters too, because Nebius did not report in isolation. It reported two days after CoreWeave surged about 25 percent on its own 168 percent revenue beat, and after IREN also posted better-than-expected results, so the artificial intelligence infrastructure trade had just been given a clean bill of health by its two closest peers. It reported the morning after a hot October consumer price print, headline up 0.4 percent against 0.3 percent expected and core up 0.3 percent for a third straight month, which made a December rate hike near-certain and put the cost of borrowing at the centre of every spending plan in the market. And it reported into a week in which Nvidia's board met to discuss ousting its chief executive and Anthropic's initial public offering was reported oversubscribed several times over at about a $1.4 trillion valuation. The Nebius fall is therefore not a verdict on Nebius alone. It is the market's answer to the question the whole week has been asking: how much debt-funded growth is too much, and who pays for it.
The quarter itself: $980.8 million, up 571%
The quarter deserves a fair reading before the guidance takes it apart, because it was, on its own terms, a very strong three months. Revenue of $980.8 million was up 571 percent from $146.1 million in the third quarter of 2025 and up 68 percent from $582.3 million in the second quarter of 2026. The artificial intelligence cloud business brought in $973.8 million, about 99 percent of group revenue, according to the company. Adjusted EBITDA was $465.8 million, against a loss of $5.2 million a year earlier, a swing of more than $470 million in twelve months. The net loss from continuing operations narrowed to $37.5 million from $119.6 million, and the loss per share narrowed to 13 cents from 50 cents. The company also said its expanded customer base now includes Databricks, Datadog, Palantir and Snowflake, and that Nvidia held about $329 million in Nebius shares at the end of June, a holding that ties the smallest infrastructure name to the largest chipmaker in the trade. On growth, profitability on an adjusted basis, and customer breadth, the quarter was a step change, not an incremental improvement.
There is one number inside the quarter that complicates the beat, and it is the one the market spent the morning on. About $325 million of the quarter's revenue was non-recurring, tied to the company's new asset-light model, in which partners fund the capacity and Nebius books revenue without owning the underlying asset. That is about a third of the quarter's $980.8 million, and it is revenue that does not repeat in the same form. The distinction matters because the beat was about $73 million above consensus, and the non-recurring portion was more than four times that. Strip it out and the recurring revenue was about $656 million, still up sharply from $146.1 million a year earlier and from $582.3 million in the second quarter, but a smaller beat on a smaller base. The asset-light model is not a problem in itself. It is a way of growing without borrowing, which is exactly the question this article is about, and the partners who fund the capacity take the balance-sheet risk that Nebius does not. But a beat that leans on non-recurring revenue is a thinner beat than the headline suggests, and the market's job on Thursday was to look past the headline and price the recurring business. It did, and the recurring business, while still growing fast, did not justify the spending plan that came with it.
The other number inside the quarter that deserves attention is the cost of the debt that is already on the books. Under generally accepted accounting principles, the net loss from continuing operations was $37.5 million, and the main loss driver was interest expense of $112.5 million, according to Zhitong Finance. That is the cost of the borrowing that funded the capacity that produced the revenue, and it is rising: interest expense for the first half of 2026 was $182.8 million, so the third quarter alone added more than 60 percent of the first half's total in three months. The company also reported cash and cash equivalents of about $7.43 billion at September 30, 2026, down from $8,042.1 million at June 30, 2026, as capital spending consumed cash faster than operations produced it. The quarter's capital expenditure was about $7.3 billion, which is more than seven times the quarter's revenue. A company that spends $7.3 billion in a quarter to make $980.8 million is not a company with a profitability problem. It is a company with a funding problem, and the funding problem is the subject of the next section.
One more change in the quarter is easy to miss and worth naming, because it changes how the company will be judged from now on. Nebius said it no longer sees annualised run-rate revenue as the right measure of its business, and from 2027 it will give guidance for revenue, pre-tax earnings and capacity instead. The run-rate target of $7 to $9 billion by December, which the company had forecast, will now arrive in February, about two months late, and then the measure itself will be retired. The change is honest in one direction and convenient in the other. It is honest because run-rate revenue, which annualises a single month, flatters a company whose capacity comes online in lumps, and a pre-tax earnings guide is a harder test than a revenue run-rate. It is convenient because the measure being retired is the one the company just missed. A halal investor should note the change and judge the company on the new measures when they arrive: revenue, pre-tax earnings and capacity, reported as actuals rather than run-rates, are harder to flatter and easier to screen.
The bill for growth: $27 billion this year, $40–45 billion next
The spending plan is the story, and its scale is difficult to overstate. Nebius now expects about $27 billion of capital expenditure in 2026, up from its earlier $20 to $25 billion range, and it has given an initial 2027 outlook of $40 to $45 billion. Put those two numbers against the company's own revenue and the scale becomes clear: 2026 revenue is now expected at the low end of $3.0 to $3.4 billion, and 2027 revenue is guided at $12 to $13 billion. The 2026 spending plan is roughly nine times the 2026 revenue it will produce, and the 2027 plan is more than three times the 2027 revenue it is expected to produce. The company expects to reach pre-tax breakeven in the fourth quarter of 2027, which means it will spend about $67 to $72 billion across two years before the business, on its own guidance, stops losing money before tax. That is not a normal ratio for any company, and it is not a ratio the market has seen before from a company of this size. It is the reason the shares fell 18 percent on a beat.
The funding question is the one the chief executive addressed directly, and his answer is the most important sentence in the shareholder letter. According to Reuters, Nebius said 75 to 85 percent of its 2027 spending will be covered by customer prepayments and asset-backed financing of existing contracts, and the company expects $17 to $20 billion of customer prepayments next year, according to the shareholder letter. The logic is that the spending is not speculative: it is tied to contracts that customers have already signed, and the customers pay in advance, or the financing is secured against the contracts themselves. That is a real answer, and it is better than borrowing unsecured against a hope. But it is not a complete answer, and the part that is missing is the part the market priced. Asset-backed financing is still debt, secured against the asset rather than the company, and customer prepayments are still a liability, money received for capacity not yet delivered. If the capacity slips, as the New Jersey capacity for Microsoft just did, the prepayment sits on the balance sheet while the revenue it was meant to fund moves into the next year. The funding model works if the capacity arrives on time. The quarter's news was that some of it did not.
The capacity delay is the concrete fact behind the revenue guide, and it deserves to be stated plainly. Some third-party capacity Nebius expected to switch on this year is now scheduled for the first half of 2027, and the contracts behind that capacity remain in place, according to Reuters. Capacity in New Jersey that is contractually committed to Microsoft, for instance, is now not expected before the second quarter of 2027. The delay is not a lost customer. It is a slipped timetable, and the revenue it was meant to produce moves with it into 2027. But a slipped timetable is expensive for a company that has already spent the money: the capital expenditure is incurred when the capacity is built, and the revenue arrives when the capacity is switched on, so a delay widens the gap between spending and earning in exactly the year the spending is largest. That is the mechanics of the 2026 guide: revenue at the low end of the range, spending at the top of it, and the gap between the two funded by borrowing, prepayments and asset-backed financing. The company said the contracts remain in place, and the market believed it. It also believed that the gap had widened, and it priced the widening.
There is a second funding fact that belongs in this section, because it sets the limit on how much of the bill can be paid without new borrowing. The company's balance sheet at September 30, 2026 showed about $7.43 billion in cash and cash equivalents, down from $8,042.1 million at June 30, 2026, and the third quarter's capital expenditure alone was about $7.3 billion, roughly equal to the entire cash balance. A company cannot fund a $27 billion year from a $7.4 billion cash balance, and it cannot fund a $40 to $45 billion year from customer prepayments of $17 to $20 billion alone. The arithmetic leaves a gap of tens of billions of dollars, and that gap is filled by borrowing, by asset-backed financing, or by new equity. The chief executive's statement that 75 to 85 percent of 2027 spending will be covered by prepayments and asset-backed financing leaves 15 to 25 percent, or $6 to $11 billion, to be covered some other way, and the asset-backed portion itself is debt that will appear on the balance sheet. The question for a halal investor is not whether the spending is justified by the contracts. It is what the borrowing does to the debt ratio, which is the subject of the debt-watch section below. The market's answer on Thursday was that the borrowing will be large, and the shares fell accordingly.
The market reaction: down about 18%, and 57% below the June high
The fall on Thursday was large, fast and broad-based within the artificial intelligence infrastructure trade, and it is worth recording precisely because the numbers will be quoted for weeks. Nebius closed Wednesday at $154.69, fell about 10 percent in premarket trading, was down about 16 percent by mid-morning according to CNBC, and was down about 18 percent by late morning, trading at about $126.29 at 10:53 a.m. Eastern according to Webull data, about $127.21 at 10:42 a.m. Eastern, and about $127.49 at 11:50 a.m. Eastern, a fall of 17.58 percent on the day. The market capitalisation, about $36.43 billion at that price, is less than the company's 2027 capital expenditure plan alone. The stock is down 9.91 percent over five sessions and 38.37 percent over the past month, and at about $127 it stands about 57 percent below its 52-week high of $299.86, reached on June 22, 2026, on the arithmetic of the two prices. The fall is not a single bad morning. It is the continuation of a slide that began well before the earnings report, and the report accelerated it rather than causing it.
The slide before the report is part of the story, because it shows that the market had already begun to doubt the funding model before the company confirmed the numbers. Over the past month the shares fell 38.37 percent, and over five sessions they fell 9.91 percent, as the wider artificial intelligence trade sold off on the same questions: whether data-centre spending can continue at its current pace, whether the hyperscalers will keep funding it, and whether the infrastructure names that borrow to build it can service the debt. Last Thursday's rout, when the Nasdaq fell 5.52 percent and the semiconductor index fell 12.25 percent, and more than $2 trillion was wiped off the market, was the first test of those questions, and Friday's bounce, when the chip names recovered about a fifth of the fall, was the second. Nebius fell through both, and the earnings report gave the market a concrete reason to keep selling: a spending plan that grows faster than the revenue it is meant to produce, funded by borrowing that grows faster than the cash it is meant to replace. The 18 percent fall is the market's judgment on the plan, not on the quarter, and the month-long slide is the market's judgment on the model.
The spillover on Thursday was mixed rather than uniform, and the mix is informative. The wider chip trade was under pressure from the same forces that have moved it all week, with Nvidia down about 7 percent in premarket trading on its board story, Micron down 3.4 percent and SanDisk down 3.7 percent, while AMD rose 3.7 percent, defying the trend a week after its record quarter. The infrastructure names were the exception to the morning's weakness in one place: CoreWeave, which surged about 25 percent on Wednesday after its own 168 percent revenue beat, held up better than Nebius, and the difference between the two is the difference between a beat that came with a raised revenue guide and a beat that came with a raised spending bill. The market is not selling artificial intelligence infrastructure. It is selling the infrastructure names whose spending plans outrun their revenue, and it is buying the ones whose revenue outruns their spending. That is a funding judgment, not a technology judgment, and it is the judgment a halal investor should be making too, because funding is what the screen measures.
There is one more market fact that belongs in this section, because it frames the size of the fall in the right units. At about $127, Nebius is worth about $36.43 billion, and its 2027 capital expenditure plan of $40 to $45 billion is larger than the entire company. A company that plans to spend more than its own market value in a single year is a company whose future depends entirely on the contracts it has signed and the financing it can raise against them, and the market's 18 percent fall is a vote on whether that financing will arrive on the terms the company needs. The vote is not final. The company has $17 to $20 billion of customer prepayments expected next year, contracts with Microsoft and others that remain in place, and a 2027 revenue guide of $12 to $13 billion that, if achieved, would be roughly four times the 2026 figure. But a vote is a vote, and the market's vote on Thursday was that the plan is bigger than the balance sheet that has to fund it. The next section is about whether the screen agrees.
The screen results: Nebius still passes
None of Thursday's moves changed a single screen result, and that is the central fact of this article. The screen is measured on audited annual figures and the most recent filings, not on prices and not on headlines, and no price, however large, and no guidance change, however sharp, changes a debt ratio or an income ratio until the next filing reports it. Nebius passes this site's October 2026 screen, screened on October 2, 2026, with interest-bearing debt of 13.4 percent of market capitalisation and non-compliant income of 3.9 percent of revenue, both below their ceilings. The business-activity gate comes first: Nebius is an Amsterdam-based artificial intelligence cloud company, spun out of Yandex in 2024, building full-stack infrastructure for the global artificial intelligence industry, large graphics-processor clusters rented out as compute capacity for training and inference, with consolidated businesses that also include Avride, an autonomous-vehicle platform, and TripleTen, an edtech service. None of the disclosed business lines, artificial intelligence cloud infrastructure, autonomous vehicles or edtech, are prohibited lines, so the business gate passes. The full method, gate by gate, is set out in the screening methodology.
The debt gate is the second gate, and it is the one this article is about. At June 30, 2026 Nebius reported debt, current, of $46.7 million and debt, non-current, of $8,499.0 million, total interest-bearing debt of $8,545.7 million, per the second-quarter 2026 financial results. Against a market capitalisation of about $63.7 billion, the shares closed at $232.28 on October 1, 2026 with 271,855,218 shares outstanding, debt divided by market capitalisation is about 13.4 percent, well below the roughly 33 percent ceiling. Operating lease liabilities of $1,510.4 million, non-current, sit outside the debt lines, and including them would bring the ratio to about 15.8 percent, still under the ceiling. Cash and cash equivalents of $8,042.1 million are about 12.6 percent of market capitalisation, within the roughly 33 percent guideline. The debt-and-cash gate passes, and it passes comfortably on the October figures. The question is not whether it passed in October. The question is whether it will pass at the next refresh, when the debt that funded the third quarter's $7.3 billion of capital expenditure is on the balance sheet, and the answer to that question is not in this article, because the filing that contains it has not been published yet.
The income gate is the third gate, and it is the one that has moved in the company's favour since the screen was run. Interest income was $38.3 million for the six months ended June 30, 2026, $24.1 million in the second quarter, against revenue of $981.3 million, $582.3 million in the second quarter, about 3.90 percent of revenue, under the 5 percent non-compliant income ceiling. The interest is earned on cash balances, not from lending. On a second-quarter-only basis the ratio is about 4.14 percent, also under 5 percent. The third quarter's revenue of $980.8 million, alone larger than the entire first half's $981.3 million, pushes the ratio down further on the arithmetic, because the denominator is growing far faster than the interest on a shrinking cash balance: cash fell from $8,042.1 million at June 30 to about $7.43 billion at September 30, so the interest the cash earns is falling while the revenue it is measured against is rising. The income gate is therefore the comfortable one, and it is getting more comfortable as the company grows. The debt gate is the tight one, and it is getting tighter as the company borrows. The two gates are moving in opposite directions, and the screen result at the next refresh will depend on which one moves further.
Two flags belong in the article, reported as context and not as a screen change, and the first is a disagreement that the site has reported honestly since October. Zoya covers Nebius and, as of September 2026, flags it as not Shariah-compliant, based on its fiscal 2025 figures, where it cites $31.8 million of interest income on $529.8 million of revenue, about 6 percent, over its 5 percent threshold. This site's screen uses the more recent first-half 2026 figures, where the ratio is 3.9 percent, and the two assessments disagree because they use different periods, not because one of them is wrong about the numbers. No rating or coverage of Nebius was found on Musaffa or ShariaPortfolio in a search conducted in October 2026, honestly reported as absent, not invented. The disagreement is reported here, not resolved, because it is a disagreement about which period to use, and the next filing will give both sides a new period to use. A reader who follows Zoya will already have excluded the name. A reader who follows this site's screen will have it as a pass, and will need to watch the next refresh more closely than any other name in the portfolio, because this is the pass with the thinnest margin for error on the debt side.
The second flag is the absence of two others. Nebius does not appear on the BDS movement's published boycott, divestment or pressure lists, as checked in October 2026. That flag is informational only. It does not change the Shariah screening result, and it is not a religious ruling. Readers who follow the BDS movement's lists will weigh it for themselves; readers who do not will note it and move on. The site's moral screen, which checks documented business ties to the Israeli and United States militaries and to weapons manufacturing, publishes no findings for Nebius. A row with no finding means none surfaced in the review, not proof of none, and the screen results stand on the financial gates alone: Nebius passes, and nothing else in this article changes that. The spending plan is a debt question, not a moral-screen finding, and it is reported here as one.
The debt watch: the June debt figure against today's market value
The debt gate is where the Nebius story becomes a halal story, and the arithmetic is worth doing in public, because it is the arithmetic the next filing will either confirm or overturn. The screen's debt ratio is measured against market capitalisation, so when the price falls the denominator shrinks and the ratio rises. At the October screen, the ratio was 13.4 percent: $8,545.7 million of debt against about $63.7 billion of market capitalisation. At Thursday's price of about $127, the market capitalisation is about $36.43 billion. The June 30 debt figure alone, $8,545.7 million, divided by today's market capitalisation, is about 23.5 percent, on the arithmetic of the two figures. That is still under the roughly 33 percent ceiling, and it is still under the 30 percent ceiling that some providers use, which is the disagreement reported in the methodology, but the margin has narrowed from about 20 percentage points to about 10, and from about 17 percentage points to about 7 on the stricter reading. The pass has not flipped. The room for error has halved. That is the labelled arithmetic, and it is labelled because it uses the June 30 debt figure, not a new one: the debt has grown since June 30, and the next filing will say by how much.
The growth since June 30 is the part of the arithmetic that cannot be done yet, and it is the part that matters most. The third quarter's capital expenditure was about $7.3 billion, and the company's cash fell from $8,042.1 million at June 30 to about $7.43 billion at September 30, a fall of about $600 million in three months while the company spent $7.3 billion. The difference between the spending and the cash consumed was funded by borrowing, by prepayments and by asset-backed financing, and the balance sheet at September 30 will show how much of it was borrowing. The company's own 6-K for the results was not yet on the SEC's EDGAR system when the first reports were written, so the figures in this article come from the company's own release and shareholder letter, cross-checked against the Reuters report, and the debt figure at September 30 is not yet public. What is public is the direction: the debt is higher than $8,545.7 million, the market capitalisation is lower than $63.7 billion, and the ratio is therefore higher than 13.4 percent and higher than the 23.5 percent that the June figure alone would produce. How much higher is the question the next filing answers, and it is the only question that decides whether the pass holds.
The scale of the plan is the reason the question is urgent, and it is worth stating the scale in the units the screen uses. The 2027 capital expenditure plan of $40 to $45 billion is larger than the company's entire market capitalisation of about $36.43 billion. The 2026 plan of about $27 billion is about three quarters of it. A company that spends more than its own value in a year, funded in large part by borrowing secured against the contracts it is building for, is a company whose debt ratio will move by tens of percentage points in a single filing, not by tenths. The chief executive's statement that 75 to 85 percent of 2027 spending will be covered by customer prepayments and asset-backed financing is the mitigation, and it is a real one: prepayments are not debt in the same way, and asset-backed financing is secured against the asset rather than the company. But asset-backed financing is still interest-bearing debt, and it still appears on the balance sheet, and the 15 to 25 percent of the plan that is not covered, $6 to $11 billion, has to be funded some other way. The screen does not distinguish between kinds of debt. It measures the total against the market value, and the total is about to grow very fast.
There is a comparison that makes the scale concrete, and it is the comparison the market made on Wednesday. CoreWeave's debt is about $47 billion, and its market capitalisation is far larger than Nebius's, so its ratio is 74.7 percent and it fails. Nebius's debt at June 30 was $8,545.7 million, and its market capitalisation was $63.7 billion, so its ratio was 13.4 percent and it passed. The difference between the two is not the model. It is the stage: CoreWeave has already borrowed the money, and Nebius is about to. The 2027 spending plan, if funded even partly by borrowing, closes that gap in a single year, and the market's 18 percent fall is a vote that the gap will close faster than the revenue will grow. The vote is not a screen result. The screen result arrives with the filing. But the vote is the reason to watch the filing, and the filing is the reason to hold the judgment until it arrives.
One more mechanical point belongs in this section, because it is the point most often misunderstood about the debt gate. A falling share price does not change the company's debt, and it does not change the company's revenue. It changes the denominator of the debt ratio, and for most companies the change is small. For Nebius it is not small, because the fall has been large: from $232.28 on October 1 to about $127 on Thursday, a fall of about 45 percent, which shrinks the denominator by almost half and nearly doubles the ratio on the same debt. That is the arithmetic of the 23.5 percent figure, and it is the reason a price fall matters to a screen even though the screen is not measured on prices. The fall does not change the business. It changes the room the business has to borrow, and the business is about to borrow more than it ever has. The disciplined position is to note the labelled arithmetic, note that it is not a new screen, and wait for the filing that replaces the June figure with the September figure, and then the December figure. The screen is the house. The filing is the inspection. Read the filing when it arrives.
The income gate: why the third gate is moving the other way
The third gate is the good news in this article, and it deserves the same careful reading as the debt gate, because the two gates are moving in opposite directions and the screen result will depend on which one moves further. The non-compliant income gate measures interest income as a percentage of revenue, with a ceiling of 5 percent, and Nebius sits at 3.9 percent on the first-half 2026 figures, with interest income of $38.3 million against revenue of $981.3 million. The third quarter's revenue of $980.8 million, alone larger than the entire first half's revenue, pushes the ratio down on the arithmetic, because the denominator is growing far faster than the numerator: the interest is earned on cash balances, and the cash balance fell from $8,042.1 million at June 30 to about $7.43 billion at September 30, so the interest the cash earns is falling while the revenue it is measured against is more than doubling. The income gate is therefore getting more comfortable as the company grows, and it will get more comfortable still if 2027 revenue reaches the guided $12 to $13 billion, because the interest on any plausible cash balance is a smaller fraction of $12 billion than it is of $1 billion. The third gate is not the problem. It is the cushion.
The cushion has a limit, and the limit is the rate environment, which is the thread that runs through every article this week. October's consumer price print came in hot on Tuesday, headline up 0.4 percent against 0.3 percent expected and core up 0.3 percent for a third straight month, and the December hike went from likely to near-certain, with FedWatch putting a 25-basis-point hike at 84.7 percent and a 50-basis-point hike at 15.3 percent after the release. Higher rates raise the interest income of cash-rich companies, and interest income is exactly what the third gate measures, so a higher rate path pushes the numerator up while the revenue growth pushes the denominator up faster. For Nebius the revenue growth is so fast that the denominator wins, and the gate stays comfortable. For the names near the ceiling, the arithmetic is tighter: Regeneron sits at 4.998 percent, a rounding error below the 5 percent ceiling, and Arista sits at 4.26 percent, and on the failing side Cloudflare fails at 6.05 percent and Palantir fails at 5.12 percent. Those are the ratios the December hike actually threatens, and Nebius, at 3.9 percent and falling, is not among them. The rate thread matters to this article because it is the reason the income gate is safe, and the reason the debt gate is the one to watch instead.
There is a subtlety in the income gate that is worth naming, because it is the subtlety that separates Nebius from the names that fail it. The interest income is earned on cash balances, not from lending, and the cash balance is falling because the company is spending it on capacity. A company that spends its cash on data centres earns less interest on it, and the income ratio falls, even as the debt it takes on to fund the same capacity pushes the debt ratio up. The two gates are therefore linked by the same decision: spend the cash, and the income gate improves while the debt gate worsens. That is the trade-off the company has chosen, and it is the trade-off the screen will measure at the next refresh. A halal investor should understand the trade-off before judging it, because it is not a trade-off between compliance and growth. It is a trade-off between two gates, and the company has chosen the one that is currently comfortable over the one that is currently tight. The screen will say whether that choice was right, and it will say it in the numbers, not in the press release.
The Zoya disagreement is the last piece of the income gate, and it is worth restating here because it is the piece a reader who follows Zoya will already know. Zoya flags Nebius as not Shariah-compliant on its fiscal 2025 figures, where it cites $31.8 million of interest income on $529.8 million of revenue, about 6 percent, over its 5 percent threshold. This site's screen uses the more recent first-half 2026 figures, where the ratio is 3.9 percent, and the two assessments disagree because they use different periods. The disagreement is not a dispute about the numbers. It is a dispute about which period to use, and the next filing will give both sides a new period to use. A reader who follows Zoya will have excluded the name on the 2025 figures, and a reader who follows this site's screen will have it as a pass on the 2026 figures, and both readers should watch the same filing, because the filing will show whether the 2026 improvement was a trend or a single half-year. The honest position is to report the disagreement, not to resolve it, and to let the next filing do the resolving. That is what this article does, and it is what the framework in the next sections does too.
The CoreWeave contrast: same model, opposite screen result
The most useful comparison in this article is the one the market made on Wednesday, and it is the comparison between Nebius and CoreWeave, the two artificial intelligence infrastructure names that reported two days apart with opposite market reactions and opposite screen results. CoreWeave reported third-quarter revenue of $3.66 billion, up 168 percent, beat the $3.55 billion expected, raised its 2026 guidance, and surged about 25 percent on Wednesday, with the backlog reaching $132.5 billion. Nebius reported third-quarter revenue of $980.8 million, up 571 percent, beat the $907.5 million expected, raised its spending plan, and fell about 18 percent on Thursday. The difference between the two is not the growth. Nebius grew faster. The difference is the funding: CoreWeave's revenue guide went up, and Nebius's spending guide went up, and the market rewarded the first and punished the second. The screen, which measures funding rather than growth, agrees with the market's direction, but not with its size: CoreWeave fails, and Nebius passes, and the distance between them is the distance the next filing will close.
CoreWeave's screen result is the one to state plainly, because it is the benchmark against which Nebius will be measured. CoreWeave fails this site's October 2026 screen at the debt gate, with debt of 74.7 percent of market capitalisation against a roughly 33 percent ceiling, and non-compliant income of 0.43 percent, which passes. Its debt is now about $47 billion, against a market capitalisation that has fallen with the shares, and its third-quarter capital expenditure was $14.5 billion, about four times its revenue, with 2026 capital expenditure guided at $37 to $39 billion and a 2026 revenue guide of $13.0 to $13.3 billion. The model is the same as Nebius's: borrow to build data centres, rent the capacity to customers, and grow revenue faster than the debt that funds it. The difference is the stage: CoreWeave has already borrowed the money, and its ratio is 74.7 percent, more than twice the ceiling. Nebius is about to borrow the money, and its ratio is 13.4 percent on the October figures, less than half the ceiling. The question is not whether the model works. It is whether the model can be funded without the debt ratio crossing the ceiling, and CoreWeave is the answer to that question for a company that has already tried.
The comparison is useful because it shows what the screen is actually measuring, and what it is not. The screen does not measure growth, and it does not measure whether the spending is justified by the contracts. CoreWeave's revenue grew 168 percent and its backlog is $132.5 billion, and it still fails, because the debt that funded the growth is 74.7 percent of the market value. Nebius's revenue grew 571 percent and its customer base now includes Databricks, Datadog, Palantir and Snowflake, and it still passes, because the debt that funded the growth is 13.4 percent of the market value on the October figures. The screen is a funding test, not a growth test, and the funding test is the one that matters to a halal investor, because the funding test is the one that measures whether the company's growth is built on interest-bearing debt. A company can grow 571 percent and still pass, if it funds the growth without borrowing too much. A company can grow 168 percent and still fail, if it funds the growth by borrowing too much. The growth is not the question. The borrowing is.
There is a second comparison inside the same trade that is worth naming, because it shows the range of outcomes the model can produce. IREN, the third infrastructure name, also posted better-than-expected results over the past week, according to CNBC, and the three names together, CoreWeave, IREN and Nebius, are the market's test of whether the artificial intelligence buildout can be funded at its current pace. The test has produced three different answers in three days: CoreWeave surged on a raised revenue guide, IREN beat, and Nebius fell on a raised spending bill. The market's answer is that the buildout can be funded, but not by every company, and not on any terms. The screen's answer is the same, in different units: the buildout is halal if the funding is halal, and the funding is halal if the debt ratio stays under the ceiling. CoreWeave is past it. Nebius is under it, for now. The next filing will say whether it stays under it, and the filing is the only answer that counts.
How a halal investor should approach a beat that fell
A beat that falls triggers a specific set of instincts, and they are worth naming before acting on any of them. The first is to buy the dip, on the theory that an 18 percent fall in a company that just grew 571 percent is a bargain, or that a $127 price is a floor. The second is to sell everything, on the theory that a spending plan larger than the company's own market value is the start of something worse. The third is to do nothing and feel guilty about it, on the theory that doing nothing is not a decision. None of the three is a framework. A framework is a set of questions, answered in order, before any order is placed, and for a beat that fell the questions are these.
First, screen before you react. Nebius passes the screen, and that fact did not change on Thursday morning. A halal investor who sells a passing name because it fell 18 percent has let the price do the screening, and a halal investor who buys a failing name because it fell has done the same thing in the other direction. The screen is the first question, and the price is not an input to it. Second, separate the quarter from the guidance. The quarter was a beat, and the guidance was a warning, and the market priced the guidance. A halal investor should do the same: note the revenue, note the non-recurring $325 million inside it, note the adjusted EBITDA, and then note the spending plan, because the spending plan is what the debt gate will measure. Third, separate the business from the funding. The business is artificial intelligence cloud infrastructure, a permissible business, and the funding is interest-bearing debt and asset-backed financing, which the screen measures. The business did not change on Thursday. The funding plan did, and the funding plan is what the next filing will show.
Fourth, do not trade a headline. There is no new debt figure in this story, no new filing, and no new screen result, and the debt figure that decides the pass has not been published yet. The disciplined position is to note the screen result as it stands, note the labelled arithmetic that the June debt figure against today's market value would be about 23.5 percent, and wait for the filing. Fifth, 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 an 18 percent fall is a price, not a forced sale. That is the practical difference the screen's debt gate buys you: the companies you hold do not borrow heavily, and neither do you. Sixth, watch the filings, not the price. The price will move every day, and the filings arrive every quarter, and only the filings change the screen. Seventh, recheck at the refresh. If the borrowing grows faster than the market value, the next screen refresh will show it in the numbers, and the pass will flip to a fail on the evidence, not on a headline. The framework is the position: screen first, separate the quarter from the guidance, separate the business from the funding, do not trade a headline, no margin, watch the filings, and wait for the refresh.
There is one more question the framework has to answer, and it is the one the spending plan raises directly. A company that plans to spend $40 to $45 billion in a year, more than its own market value, is a company whose future depends entirely on the contracts it has signed and the financing it can raise against them, and a halal investor has to decide whether that dependence is a risk worth taking. The screen does not answer that question. It answers whether the debt is under the ceiling, and on the October figures it is. The judgment about whether the plan is prudent, whether the contracts will hold, and whether the financing will arrive on the terms the company needs, is a business judgment, not a screen result, and it belongs with the investor, not with the screen. The screen is the gate. The judgment is the walk through it. Keep the two apart, and let neither do the other's work.
What to watch next
The next number that matters for Nebius is not a price. It is a filing. Watch first for the company's 6-K for the third-quarter results, which was not yet on the SEC's EDGAR system when the first reports were written, because it will contain the September 30 balance sheet, and the September 30 debt figure is the one that replaces the June 30 figure in the screen. Watch for the debt, current and non-current, and for the cash, and for the interest expense, which was $112.5 million in the third quarter and is rising with the borrowing. Watch for the fourth-quarter and full-year 2026 results, expected in February 2027, because they will contain the December 31 balance sheet, and the December debt figure is the one that will decide whether the pass holds after a year in which the company spent about $27 billion. And watch for the 2027 capacity announcements, because the New Jersey capacity for Microsoft, now expected in the second quarter of 2027, is the test of whether the slipped timetable slips again.
For the wider trade, the next tests arrive this week and next. After the close on Thursday, Cisco Systems and Applied Materials both report, with calls at 4:30 p.m. Eastern, and options markets are pricing post-earnings moves of roughly 8 percent for each, in either direction. Cisco supplies the networking that connects data centres, and Applied Materials supplies the chipmaking tools used to build the processors and memory inside them, so the two prints test the same buildout from opposite ends, independent of the Nebius story. On Tuesday 17 November, Nvidia reports its fiscal third-quarter 2027 results, with consensus revenue of about $108.8 to $109.2 billion, and the report is now a leadership test as well as a demand test, after the board met to discuss ousting its chief executive. On Wednesday 18 November, Anthropic is expected to price its initial public offering, and on Thursday 19 November the shares are expected to begin trading on the Nasdaq as ANT. The Federal Reserve decides on 8 to 9 December, and the Bank of Canada decides on 9 December, after a hot October consumer price print made a December hike near-certain. For readers who hold Nebius, the practical next step is the same as it was after the CoreWeave print and after last Thursday's rout: note the screen result as it stands, note that a guidance change changes none of the ratios until the filing reports it, and recheck them 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 spending plan is a funding question. The screen is the house. Read the plan, and keep the house in order.
FAQ
What did Nebius report on Thursday 12 November 2026?
Nebius reported third-quarter revenue of $980.8 million, up 571 percent from $146.1 million a year earlier and above the $907.5 million LSEG consensus cited by CNBC, with adjusted EBITDA of $465.8 million against a loss of $5.2 million a year earlier. The net loss from continuing operations narrowed to $37.5 million, or 13 cents a share, from $119.6 million, or 50 cents a share, a year earlier. About $325 million of the quarter's revenue was non-recurring, tied to the company's new asset-light model, and the artificial intelligence cloud business brought in $973.8 million, about 99 percent of group revenue.
Why did the shares fall about 18 percent on a beat?
The fall was about the guidance, not the quarter. Nebius raised its 2026 capital spending plan to about $27 billion from a $20 to $25 billion range, guided 2027 spending at $40 to $45 billion against about $28.8 billion expected, and pushed 2026 revenue to the low end of its $3.0 to $3.4 billion range after third-party data-centre capacity slipped into the first half of 2027. The shares fell about 10 percent in premarket trading, about 16 percent by mid-morning according to CNBC, and about 18 percent by late morning, to about $127 from Wednesday's close of $154.69.
How is Nebius funding $27 billion of spending this year and $40 to $45 billion next year?
According to Reuters, Nebius said 75 to 85 percent of its 2027 spending will be covered by customer prepayments and asset-backed financing of existing contracts, and the company expects $17 to $20 billion of customer prepayments next year, according to the shareholder letter. The logic is that the spending is tied to contracts customers have already signed, so the customers pay in advance or the financing is secured against the contracts themselves. Asset-backed financing is still interest-bearing debt, and customer prepayments are still a liability for capacity not yet delivered, and the 15 to 25 percent of the plan that is not covered, $6 to $11 billion, has to be funded some other way.
Does Nebius still pass the halal screen?
Yes. Nebius passes this site's October 2026 screen, with interest-bearing debt of 13.4 percent of market capitalisation and non-compliant income of 3.9 percent of revenue, both below their ceilings of roughly 33 percent and 5 percent. The screen is measured on the filings, not on prices or headlines, so a guidance change changes no ratio until the next filing reports it. The labelled arithmetic is that the June 30 debt figure of $8,545.7 million, measured against today's roughly $36.43 billion market value, would be about 23.5 percent, still under the ceiling but with the margin halved, and the debt has grown since June 30, so the next filing decides whether the pass holds. Zoya flags Nebius as not compliant on its older fiscal 2025 figures, where it cites about 6 percent interest income, and that disagreement is reported, not resolved.
How does Nebius compare with CoreWeave on the screen?
CoreWeave fails this site's October 2026 screen at the debt gate, with debt of 74.7 percent of market capitalisation against a roughly 33 percent ceiling, and debt of about $47 billion, while Nebius passes at 13.4 percent on the October figures, with debt of $8,545.7 million at June 30, 2026. The model is the same, borrow to build data centres and rent the capacity to customers, and the difference is the stage: CoreWeave has already borrowed the money, and Nebius is about to, with a 2027 spending plan of $40 to $45 billion against a market capitalisation of about $36.43 billion. The screen is a funding test, not a growth test, and the next filing will show whether Nebius closes the gap.
Is anything in this article a fatwa or a buy or sell call?
No. Everything here is facts and screening methodology only, with no fatwas and no personalised buy or sell calls. Rulings on your own situation belong with a qualified scholar, and investment decisions belong with advice suited to your circumstances. Not financial advice. Facts and screening methodology only, no fatwas.