COREWEAVE, INC. (CRWV)
Outperform

The Business Beat the Plan While the Bonds Priced Distress: Maintaining Outperform on CoreWeave, Target Cut to $93 on Cost of Capital Alone

Published: By A.N. Burrows CRWV | Q3 2026 Earnings Analysis

Key Takeaways

  • The quarter we called make-or-break cleared its ceiling by 46%. Adjusted operating income of $379M against a $200–260M guide took the adjusted operating margin from 5.0% to 10.4%, one quarter ahead of the low-double-digit target management set at the start of the year. Revenue of $3,661M rose $1,086M sequentially and beat consensus by 3.1%, and the full-year profit guide rose 26% at the midpoint.
  • The stock fell 54.8% between our August upgrade and this print, and the cause was the bond market, not the business. Yields on the 2031 and 2032 unsecured notes passed 15% in October and default-swap spreads crossed 1,100 basis points, while two other listed neoclouds fell by similar amounts. Management answered with a funding plan: no new unsecured debt at current yields through 2027, roughly $9B of customer prepayments in the next three months, an investment-grade term loan indicated at 7–7.5%, and less than $5B of parent-level capital needed through 2027.
  • The residual-value disclosures we asked for in August arrived on this call. More than 130,000 GPUs come off contract in 2027 and about 170,000 in 2028; short-term capacity sells for about $40M per megawatt a year, a figure management gave both as a floor and as a cap, against $20–25M on four-to-five-year contracts; and Ampere and Hopper pricing rose 30–50% sequentially, above original contract levels. Residual value is now a number rather than an adjective.
  • The balance sheet is still the bear case, and it grew faster than the margin. Net debt rose 43.5% in one quarter to $40.4B, stockholders' equity fell $1.0B to $4.0B, and days sales outstanding rose to about 108 from 90. Interest expense of $846M is 2.2 times adjusted operating income, against 5.0 times in Q2, and the fourth-quarter guide implies about 1.4 times.
  • Rating: Maintaining Outperform; target to $93 from $124; conviction to 7 from 6. Two of the four higher-conviction conditions we set in August fired, where one was enough, and none of the five downgrade conditions did, so we honour them. The whole target cut comes from a lower multiple that prices a higher cost of capital, and at $60.77 the stock trades at 3.6 times the exit-2026 run-rate guide.

CRWV financial model

Income Statement · Dollars in millions, except per share

Income statement preview for CRWV. Dollars in millions, except per share. Actual fiscal years followed by our estimates.
Income Statement
ActualEstimate
202420252026E2027E2028E
Revenue$1,915.4$5,131.0$12,808.4$25,408.2$40,026.8
Less: Cost of revenue(493.4)(1,453.0)(4,321.5)(8,300.0)(12,758.0)
Less: Technology and infrastructure(960.7)(2,929.0)(7,091.7)(12,952.4)(19,345.8)
Less: Sales and marketing(18.4)(144.0)(336.4)(620.8)(929.1)
Less: General and administrative(118.6)(651.0)(809.5)(1,236.5)(1,642.1)
Total operating expenses($1,591.1)($5,177.0)($12,559.2)($23,109.7)($34,675.0)
Operating income (loss)$324.4($46.0)$249.2$2,298.5$5,351.7
D&A863.42,454.06,467.111,876.817,810.1
EBITDA (Operating income + D&A)$1,187.8$2,408.0$6,716.2$14,175.2$23,161.8
Gain (loss) on fair value adjustments($755.9)$27.0$0.0$0.0$0.0
Interest expense, net(360.8)(1,229.0)(3,138.8)(5,302.7)(6,963.0)
Other income (expense), net48.233.0189.8127.0200.1
Loss before income taxes($744.2)($1,215.0)($2,699.9)($2,877.2)($1,411.1)
Less: Provision for (benefit from) income taxes($119.2)$48.0($323.6)($287.7)$81.6
Net loss($863.4)($1,167.0)($3,023.5)($3,164.9)($1,329.5)
NOPAT (Operating income x (1 - effective tax rate))376.3(44.2)279.02,528.35,042.2
Net loss attributable to common stockholders, basic($937.8)($1,196.0)($3,023.5)($3,164.9)($1,329.5)
Net loss attributable to common stockholders, diluted(937.8)(1,223.0)(3,023.5)(3,164.9)(1,329.5)
Weighted-average shares, basic (M)217.854435.000549.146574.836594.330
Weighted-average shares, diluted (M)217.854436.000549.146574.836594.330
Net loss per share, basic($4.30)($2.75)($5.51)($5.51)($2.24)
Net loss per share, diluted($4.30)($2.81)($5.51)($5.51)($2.24)
Ratios & Assumptions
YoY Revenue Growth167.9%149.6%98.4%57.5%
Operating Margin16.9%(0.9%)1.9%9.0%13.4%
Adjusted Operating Margin18.6%13.0%7.6%12.7%16.3%
Net Margin(45.1%)(22.7%)(23.6%)(12.5%)(3.3%)

The full workbook adds 10 historical and 10 projected quarters, plus KPI Drivers · Balance Sheet · Cash Flow Statement — every subtotal a live formula, every projection traced to a driver.

Download the full CRWV model (xlsx)3 free models a month with an account, then $9 a monthAbout the CRWV model →

Results vs. Consensus

Q3 2026 Scorecard

MetricQ3 2026 ActualConsensus / guideBeat/MissMagnitude
Revenue$3,661M$3.55B consensusBeat+3.1%
Revenue vs. company guide$3,661M$3.45–3.6BAbove the high end+$61M vs. ceiling; +$136M vs. midpoint
Adjusted loss per share$(0.95)$(1.16)Beat$0.21 narrower
GAAP diluted loss per share$(1.27)$(1.46) to $(1.55)Beat$0.19–0.28 narrower
Adjusted operating income$379M$200–260M guide; ~$234–246M StreetAbove the high end+$119M vs. ceiling; +54% to +62% vs. Street
Adjusted operating margin10.4%5.8–7.2% implied by guideAbove+540bp sequentially
Adjusted EBITDA$2,256M~$2.04–2.05B StreetBeat+10% to +11%; 61.6% margin
Interest expense, net$(846)M$860–940M guideBelow the low end−$14M vs. floor
Capex (management basis)$14.5B$11.5–13.5B guideAbove the high end+$1.0B; Q4 deliveries pulled forward
Operating cash flow$1,975Mn/aRecoveredFrom $679M in Q2
Revenue backlog$132.5Bn/a+$28.3B QoQ+138% YoY
FY2026 revenue guide$13.0–13.3B$12.86B consensus; prior guide $12.4–13.2BRaised+2.3% vs. Street at the midpoint
FY2026 adjusted operating income guide$1.275–1.375BPrior guide $960M–1.15BRaised+26% at the midpoint
FY2026 capex guide$37–39BPrior guide $35–39BLow end raisedEarlier deliveries, not cost
Q4 2026 interest expense guide$1.05–1.15Bn/aNew+24% to +36% vs. Q3
Exit-2026 annualized run rate$20–21BPrior guide $18.5–19.5BRaised $1.5BBoth ends up
Exit-2026 active powerMore than 1.85 GWPrior guide more than 1.85 GWMaintainedRoughly 150 MW still to energise in Q4
Every guided operating line beat its own range. Revenue landed above the top of the range, adjusted operating income landed $119M above the top, and interest expense landed below the bottom. Capex was the only line outside its range on the unfavourable side, and management attributed it to pulling fourth-quarter deliveries forward rather than to cost. Operating expenses grew 31.6% against revenue growth of 42.2%, a gap of 10.5 percentage points against 5.8 in the second quarter. The incremental adjusted operating margin on the quarter's $1,086M of new revenue was 23.1%, after 21.5% in the second quarter. Two consecutive quarters above 20% is the contribution-margin claim showing up in the reported numbers rather than in management's description of them.

Year-Over-Year Comparison

MetricQ3 2026Q3 2025Change
Revenue$3,661M$1,365M+168%
Cost of revenue$1,183M$369M+221%
Technology and infrastructure$2,025M$747M+171%
Sales and marketing$58M$45M+29%
General and administrative$188M$152M+24%
Total operating expenses$3,454M$1,313M+163%
Operating income$207M$52M+$155M
Interest expense, net$(846)M$(310)M+173%
Provision for (benefit from) income taxes$83M$(127)MSwung $210M
Net loss$(709)M$(110)MWidened $599M
Net loss per share, diluted$(1.27)$(0.22)Widened $1.05
Adjusted EBITDA$2,256M$838M+169%
Adjusted EBITDA margin61.6%61.4%+20bp
Adjusted operating income$379M$217M+75%
Adjusted operating margin10.4%15.9%−550bp
Adjusted net loss$(529)M$(41)MWidened $488M
D&A$1,889M$630M+200%
Weighted-average diluted shares559M498M+12%

The year-over-year frame still shows the cost of the build. Adjusted operating margin is 550 basis points below the third quarter of 2025, and depreciation grew 200% against revenue growth of 168%. What has changed is the direction of the gap: a year ago the margin was being harvested from a smaller, older fleet, and the comparison has been moving against the company for five quarters. At 10.4% it is now closing, and the fourth-quarter guide implies roughly 16.5%, well above the 5.6% of the fourth quarter of 2025.

The tax line deserves a sentence because it flatters nothing and hurts the bottom line. A provision of $83M on a pre-tax loss of $626M reflects a valuation allowance on deferred tax assets, so the company is paying tax accounting on losses it cannot yet use. Management said the rate should stay broadly consistent through 2026. The year-ago quarter carried a $127M tax benefit that included a catch-up for the One Big Beautiful Bill Act, which makes the year-over-year net loss comparison look worse than the operating comparison is.

Quarter-Over-Quarter Comparison

MetricQ3 2026Q2 2026QoQ Change
Revenue$3,661M$2,575M+42.2%
Total operating expenses$3,454M$2,624M+31.6%
Operating income (loss)$207M$(49)MSwung $256M
Adjusted operating income$379M$128M+$251M
Adjusted operating margin10.4%5.0%+540bp
Adjusted EBITDA$2,256M$1,510M+49.4%
Adjusted EBITDA margin61.6%58.6%+300bp
Interest expense, net$(846)M$(640)M+32.2%
Interest expense / adjusted operating income2.2x5.0xSecond consecutive improvement
Net loss$(709)M$(626)MWidened $83M
Adjusted net loss$(529)M$(567)MNarrowed $38M
Adjusted loss per share$(0.95)$(1.03)$0.08 narrower
D&A$1,889M$1,393M+35.6%
Operating cash flow$1,975M$679M+$1,296M
Purchases of property and equipment$(9,441)M$(6,422)M+47.0%
Total debt$46,873M$35,068M+33.7%
Net debt$40,407M$28,149M+43.5%
Debt due within one year$10,572M$7,513M+$3,059M
Accounts receivable, net$4,280M$2,541M+$1,739M
Days sales outstanding~108 days~90 days+18 days
Deferred revenue (total)$10,002M$9,692M+$310M
Total stockholders' equity$4,025M$5,024M−$999M
Revenue backlog$132.5B$104.2B+$28.3B
Active powerMore than 1.7 GW1.5 GW+more than 200 MW
Contracted powerMore than 4.6 GW3.7 GW+more than 900 MW

The table splits cleanly in two. Every operating row moved in the shareholder's favour, several of them sharply, and every balance sheet row except backlog moved against. Adjusted operating income rose $251M while interest expense rose $206M, so the operating gain outran the incremental financing cost by $45M, up from a $3M margin in the second quarter, when the two increases were $107M and $104M. That is the shift management described as "the path to earnings", and on this evidence it is a fair description of the income statement.

The balance sheet tells the other half. Total debt rose $11.8B in three months and net debt rose $12.3B, because cash fell while debt grew. Stockholders' equity fell by $999M, driven by the $709M net loss and the $566M paid for capped calls alongside the September convertible. The company now carries $87.5B of liabilities against $4.0B of equity. A business can grow out of that ratio only if the income statement keeps improving at this pace, and it is the reason the credit market treated the equity as it did in October.

Quality of the Quarter

Revenue: A 3.1% beat on consensus is five times the 0.6% of the second quarter, and it came with better composition. Management raised list prices about 25% in July and another 10% later in the quarter, and attributed part of the margin beat to allocating capacity into short-term contracts priced at about $40M per megawatt a year. That mix is good for this quarter's margin and less certain for next year's, because short-term revenue is by definition the revenue that does not sit in the backlog. We count it as genuine revenue with a shorter half-life than the take-or-pay base.

Margins: The line that holds the rating. In August we wrote that a third quarter at or above the $260M high end would make the fourth-quarter step arithmetic rather than aspirational. It came in at $379M. The implied fourth-quarter figure is now $747–847M against $551–801M implied in August, so the required step from the third quarter fell from roughly 2.9 times to roughly 2.1 times while the absolute target rose. Management credited three drivers: execution on bringing capacity live, pricing, and premium products. It declined to say how much each contributed, which matters because only the first is fully repeatable.

EPS: Adjusted loss per share of $(0.95) against $(1.16) expected. The GAAP loss widened sequentially to $(1.27) from $(1.14) even though GAAP operating income turned positive at $207M, because interest rose $206M, other income fell $112M from a second quarter that carried investment gains, and the tax provision reached $83M. The adjusted line is the better read of operations. The GAAP line is the better read of what the capital structure costs the shareholder.

The bridge, six quarters running. Adjusted EBITDA less D&A less net interest: Q2 2025 negative $73.3M, Q3 2025 negative $102.9M, Q4 2025 negative $311M, Q1 2026 negative $526M, Q2 2026 negative $523M, and Q3 2026 negative $479M. In August the series stopped widening. This quarter it narrowed for the first time, by $44M, in a quarter in which depreciation rose $496M sequentially. Adjusted EBITDA less interest, the measure management now asks investors to watch, reached $1,410M against $870M in the second quarter, and management expects it to more than double in 2027. The equity case rests on that line outrunning depreciation on assets that are already built and already contracted.

Business Drivers

CoreWeave reports a single segment. The drivers below are the operating disclosures that carry the investment case, each graded against what we said we would watch in August.

Backlog: $132.5B, and the Quarter's Bookings Were Mostly Added in Early Q3

MetricQ3 2026Q2 2026Q1 2026Q4 2025Q3 2025
Revenue backlog$132.5B$104.2B$99.4B$66.8B$55.6B
Sequential change+$28.3B+$4.8B+$32.6B+$11.2Bn/a
Year-over-year change+138%+246%n/an/an/a
Signed after prior quarter end and disclosed in advanceMore than $25B (early Q3)n/an/an/an/a
Delivery metric as statedMore than 50% where delivery is expected to commence by year endMore than 50% where delivery has commencedNot disclosedNot disclosedNot disclosed

The $28.3B increase is the right size and the wrong shape for a reader looking for acceleration. In August management told us that more than $25B of new commitments had been signed in the first weeks of the third quarter. Gross additions in the quarter were roughly $32.0B once the $3.66B of recognised revenue is added back, which leaves something near $7B signed in the remaining ten weeks. That is consistent with what management said on the call: it is choosing not to sell 2028 capacity yet and is holding back some available capacity in the expectation of better pricing later.

"Of this backlog, more than 50% is attached to a contract where we expect customer delivery to commence by the end of this year." — Nitin Agrawal, Chief Financial Officer

The delivery metric also changed its wording. In August more than half of the backlog was attached to contracts where delivery had already commenced, with a target above two-thirds by year end. This quarter more than half is attached to contracts where delivery is expected to commence by year end. The two statements are not the same metric, and the August two-thirds target was not mentioned.

Assessment: Bookings are not the constraint, and management said so: 10 times as much qualified demand as available capacity through 2027. The quarter's backlog growth is real, but most of it was disclosed in August, and the withheld 2028 capacity is a pricing decision we cannot yet grade. We note the change in the delivery metric's wording and will ask for the delivery-commenced figure again.

Capacity: More Than 1.7 Gigawatts Active, and a Slower Quarter of Energisation

MetricQ3 2026Q2 2026Q1 2026Q4 2025Q3 2025
Active powerMore than 1.7 GW1.5 GWMore than 1 GWMore than 850 MW~590 MW
Added in quarterMore than 200 MWNearly 500 MW~150 MW260 MWn/a
Contracted powerMore than 4.6 GW3.7 GWMore than 3.5 GW~3.1 GW2.9 GW
Deployment ratio (active / contracted)37.0%40.5%~29%~27%~20%
Contracted outside the United StatesMore than 1.3 GWMore than 1 GWNot disclosedNot disclosedNot disclosed
Exit-2026 active power targetMore than 1.85 GWMore than 1.85 GWMore than 1.7 GWMore than 1.7 GWn/a
Construction in progress$9.3B$11.9B$9.6Bn/an/a

Active power reached more than 1.7 GW, double the year-end 2025 footprint, but the quarter's additions of more than 200 MW were well below the nearly 500 MW of the second quarter. That is not the delay it looks like. The second quarter's capacity landed late, roughly 300 MW of it in June, and the third quarter was spent filling it. Construction in progress fell to $9.3B from $11.9B as that infrastructure was placed into service, and capex of $14.5B ran above the guide because fourth-quarter deliveries were pulled forward.

Contracted power grew by more than 900 MW to more than 4.6 GW, which pushed the deployment ratio down to 37.0% for the first time in this coverage. The ratio falls when contracted power grows faster than active power, so a lower ratio here is a sign of contracting momentum rather than of build slippage. International contracted power passed 1.3 GW, with new markets in Indonesia and India and expansion in Denmark and Spain.

Assessment: The exit-2026 target of more than 1.85 GW was held, not raised, which implies roughly 150 MW in the fourth quarter. That is a modest requirement against a quarter that has already received some pulled-forward deliveries. The more important number is outside the United States: 1.3 GW of contracted power abroad is the clearest hedge this company has against the Texas policy risk discussed below.

Pricing and Unit Economics: The First Absolute Numbers

DisclosureQ3 2026 callPrior disclosure
List price changes~25% across SKUs in July, another ~10% later in Q3; Q4 pricing higher again~25% in July (Q2 call)
Short-term capacity, annual revenue per MWAbout $40M, given both as a floor and as a capNever disclosed
New infrastructure on 4–5 year contracts, annual revenue per MW$20–25MNever disclosed
Ampere and Hopper pricing+30% to +50% QoQ, above original contract pricesAt or above year-ago levels
Hardware capex payback, short-term contracts6–18 monthsNever disclosed
Hardware capex payback, long-term contracts1–2 years at current pricingNever disclosed
Weighted average cost of debt7.9%, down 250bp YoYDown almost 300bp over the year (Q2)

For five quarters every pricing statement on these calls was relative: at, above or well above prior levels. This call put dollar figures on it. A megawatt of capacity sold on a short-term basis earns about $40M a year; a megawatt of new infrastructure on a four-to-five-year contract now earns $20–25M. The short-term figure came with two framings: the CEO called it "more than 40 million" and the CFO "up to as high as up to $40 million", so we treat $40M as a point estimate rather than as a floor. As a rough cross-check, the exit-2026 run-rate guide of $20–21B spread across the more than 1.85 GW expected at year end is about $11M of annualised revenue per megawatt, so the blended book still sits well below today's new-contract pricing.

"Selling capacity at more than 40 million per megawatt annually while pricing for new infrastructure on 4- and 5-year contracts is moving to 20 million to 25 million per megawatt." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: This is the pricing disclosure we have asked for since the third quarter of 2025, and it is more favourable than we modelled. The gap between today's contract pricing and the blended book is the mechanism behind the 5–10 point contribution-margin step-up claimed in August: older, cheaper contracts roll off and new ones arrive at roughly twice the blended revenue per megawatt. The risk is equally plain. Prices this far above the installed base are a market-clearing price in a shortage, and a shortage is the condition the oversupply debate is about.

Capital Structure and Funding: The Parent-Level Market Closed, the Asset-Level Market Did Not

ItemSeptember 30, 2026June 30, 2026Change
Recourse debt$41,600M$31,405M+$10,195M
Non-recourse debt$5,273M$3,663M+$1,610M
Total debt$46,873M$35,068M+$11,805M
Cash, restricted cash and marketable securities$6,466M$6,919M−$453M
Net debt$40,407M$28,149M+$12,258M
Operating lease liabilities$18,697M$16,319M+$2,378M
Undrawn committed DDTL capacityNearly $5Bn/an/a
Prepayments and deposits expected in next three months~$9B, of which more than $3B on recontracted infrastructuren/an/a
Parent-level capital needed through 2027Less than $5Bn/an/a
Next DDTL (indicated)Investment grade, 7–7.5% yield, more than 90% advance rateDDTL 4.0: A-, implied under 6%Higher all-in yield

Two capital-markets transactions were arranged inside the quarter, and one market closed after it. On September 22 the company completed an upsized $4.2B offering of 2.875% convertible notes due 2033 with capped calls carrying a $199.70 cap price, and on September 17 it established an at-the-market equity programme for up to 35 million shares. Both were arranged at a share price near $80. Three weeks later, a social-media post about the 9.625% 2032 notes set off a sell-off in which those bonds yielded 15.77% within two days, the 2031 notes came close to 16%, and five-year default swaps traded above 1,100 basis points. By November 6 the 2032 notes yielded above 16%. The stock fell 15.5% on October 14 alone.

"Looking ahead, we will not use the unsecured debt market to fund our business at current yields. Nor do we need to." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

The response separates the company's debt into two kinds. Asset-level secured debt, the delayed-draw term loans matched to specific contracts, is said to trade around par and remains open; the next facility is guided to investment grade at a 7–7.5% yield, and more than $4B has already been repaid through contract cash flows. Parent-level unsecured debt is closed at current yields, and the company says it does not need it: prepayments, the term loans, OEM financing and operating cash flow are guided to cover the significant majority of needs through 2027, with less than $5B raised at the parent.

Assessment: The breadth of funding access was the point of Bull #4, and it narrowed. A company that funded growth in every market available to it, secured and unsecured, debt and equity, now relies on secured loans, prepayments, OEM financing and operating cash flow, and two of those (prepayments and the term loans) depend on the same customers and the same contracts. The plan is coherent and the numbers support it, but it now has less slack. The indicated 7–7.5% yield on the next investment-grade facility sits above the sub-6% DDTL 4.0 achieved in May. The disclosure does not split that into base rate and spread. We address that against the downgrade trigger we set in August in the scorecard below.

Products and Customers: Managed Inference Hits Its Target a Quarter Early

MetricQ3 2026Q2 2026Target / note
Managed inference booked ARRMore than $250MMore than $100MTarget was at least $250M exiting 2026
New cloud customers already using Forge70%n/aForge launched in September
Fortune 500 companies as customers10%n/aFirst disclosure
Backlog from financial services10%n/aFirst disclosure
Customers committing more than $100M2.5x year over yearn/aNo base given
CoreWeave OmniFirst customer signedFirst deal signedStill unsized
Revenue-basis customer concentrationNot in the release; 10-Q not yet filedTop three at 72%n/a

Managed inference passed the $250M booked-ARR target one quarter early, from more than $100M in August, and management said its growth is limited by the capacity allocated to it rather than by demand. The customer disclosures are new and directionally helpful: 10% of the Fortune 500, 10% of the backlog from financial services, and more than half of this year's new customers from outside AI labs and technology.

None of that answers the concentration question in the form we track. The revenue-basis table that returned in August is not in the release, and the quarterly report had not been filed as of this note. That leaves a gap at an awkward moment: on November 6 a report said Microsoft is evaluating which neocloud partners to keep as it builds more of its own capacity, and CoreWeave was named among its relationships without being described as kept or cut. Microsoft was not mentioned on the call.

Assessment: The product line is ahead of plan and the diversification statistics are real progress, but they are not a substitute for a revenue-basis concentration figure. We move Bear #2 to EMERGING until that figure is published, because the largest single counterparty in the company's history is reported to be reviewing its supplier list.

Key Topics & Management Commentary

Overall Management Tone: Management was assertive and unusually quantitative, and the call was organised around financing rather than demand for the first time in this coverage. The prepared remarks answered the October credit sell-off before they discussed products, and they answered it with numbers (pricing per megawatt, renewal volumes, prepayment totals) rather than with framework language. Tone shifted from the retrospective confidence of August to a defensive-but-evidenced posture; the one area where management stayed general was the split of the margin beat between execution and short-term pricing.

1. Double-Digit Margins a Quarter Early

The margin path set at the start of the year was a first-quarter trough, sequential expansion through the second and third quarters, and low double digits by the fourth. It delivered 1%, then 5%, then 10.4%, so the fourth-quarter destination arrived in the third.

"Adjusted operating margin was approximately 1% in Q1, 5% in Q2, and exceeded 10% in Q3." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

The CFO attributed the outperformance to execution, pricing and premium products, and added that allocating capacity to higher-margin short-term contracts "contributed meaningfully". The full-year guide now implies a fourth-quarter adjusted operating margin of roughly 16.5% at the midpoints.

Assessment: Two consecutive quarters above the top of the guide converts the margin framework from a promise into a record. The part we discount is the short-term contribution, which helps margin in a tight market and reverses in a loose one.

2. A Funding Plan Without the Unsecured Market

The single most consequential sentence of the call was the pledge not to issue unsecured debt at current yields. Management then gave the funding stack: delayed-draw term loans matched to contracts for the significant majority, customer prepayments, OEM financing for short-term capacity, growing operating cash flow, and less than $5B of parent-level capital through 2027.

"Accounting for these sources, we expect to raise less than 5 billion in incremental parent level capital through 2027. In short, we feel confident in our ability to fund our business for growth even in the current macroeconomic conditions." — Nitin Agrawal, Chief Financial Officer

Management also restated, "for the avoidance of doubt", that contracted cash flow from existing commitments is expected to service and repay all debt outstanding, including the parent-level unsecured notes, before counting any renewals.

Assessment: The pledge is credible because the alternatives are cheaper: secured loans at 7–7.5% and prepayments with no coupon at all, against unsecured paper yielding above 15%. The parent-level figure of less than $5B is the commitment we will grade hardest, because the at-the-market programme is the obvious place to raise it and equity issued near $60 is expensive capital.

3. Prepayments Become the Cheapest Capital in the Structure

Customer prepayments have featured in every quarter of this coverage as a balance-sheet footnote. This quarter they moved to the centre of the funding plan.

"This year, we have collected prepayments on 70% of our deals with customers funding up to 100% of the contract value upfront. In the next 3 months alone, we will collect approximately 9 billion in prepayments and deposits from signed customer contracts. This includes more than 3 billion that relate to recontracting existing cloud infrastructure. Which we sold at meaningful premiums to their original price." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Management framed the choice economically: a prepayment is weighed against the price concession it costs, and in the current market prepayments beat unsecured debt. More than $3B of the $9B relates to infrastructure whose original term loans are already repaid, so it requires no debt service at all.

Assessment: Nine billion dollars over three months that largely overlap the fourth quarter is more than the $6.3–8.3B of fourth-quarter capex implied by the guide. Prepayments have one cost the call did not quantify: they are paid for with price, and they tie the funding plan more tightly to the same customers whose contracts already secure the term loans.

4. The Next Term Loan: Investment Grade, at 7–7.5%

"This will be an investment grade facility with multiple rating agencies indicating a high investment grade rating. Based on our early discussions, we expect the facility to price at a spread that implies a 7% to 7.5% yield based on current SOFR while financing more than 90% of the associated CapEx." — Nitin Agrawal, Chief Financial Officer

Management said the asset-level loan market "has remained stable", with existing loans trading around par, and that weighted average cost of debt across the company was 7.9% at quarter end, down 250 basis points year over year. More than $4B of term-loan principal has been repaid from contract cash flows, including nearly half of the first two facilities, which were the most expensive.

Assessment: A high investment-grade rating with an advance rate above 90% is a strong outcome for a company whose unsecured bonds trade at distressed yields. It confirms that lenders separate the contracts from the corporate credit. The facility has not launched, and its final pricing against DDTL 4.0 is the most important financing datapoint of the next quarter.

5. Pricing Rose Twice in the Quarter, and Got a Dollar Figure

"In July, we increased pricing across SKUs by 25%, followed by another 10% increase later in the quarter. Pricing has continued to move higher in Q4, more than offsetting cost increases." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Two increases inside one quarter is an aggressive pricing posture for a company whose customers have long-term contracts, and it applies to new and renewing business only. The dollar figures in the business drivers section put it in context: new four-to-five-year infrastructure at $20–25M per megawatt per year and short-term capacity at about $40M.

Assessment: The phrase "more than offsetting cost increases" answers the component-inflation concern raised in May. Pricing power of this kind exists only while demand exceeds supply, which makes pricing and the oversupply debate the same question asked twice.

6. The Portfolio Approach: Long Contracts for Visibility, Short Ones for Margin

Management described capacity allocation across three uses: long-term take-or-pay contracts, short-term contracts at premium prices, and managed inference. It weighs them contract by contract on revenue, operating cost and financing cost.

"The scale and flexibility of our platform allow us to allocate available strategically to higher margin, short term contracts when those opportunities arise. This portfolio approach contributed meaningfully to our strong Q3 results." — Nitin Agrawal, Chief Financial Officer

The new financing structure from August is what makes this possible: DDTL 5.5 was the first facility to finance contracts shorter than the debt, and the company is now selling more capacity on shorter terms.

Assessment: This is a deliberate trade of backlog duration for margin, and it is the right trade in a shortage. Neither the share of revenue from short-term contracts nor the share of capacity management is comfortable placing there was given, despite a direct question, so we cannot separate the durable margin from the cyclical one.

7. The Renewal Wave Finally Has a Denominator

In August we listed the size of the fleet approaching recontracting as the disclosure we most needed. It arrived in the prepared remarks.

"More than 130 thousand GPUs come off contract in 2027 as do another approximately 170 thousand GPUs in 2028. We already have a robust pipeline of customers competing for this capacity. At current pricing levels, 2027 renewals would generate approaching 4 billion of incremental annualized revenue with no additional CapEx required. This grows to nearly 10 billion of annualized incremental revenue when including 2028 renewals." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Those figures imply roughly $3.50 per GPU-hour for the 2027 cohort at current pricing and about $4 for the 2028 cohort. In Q&A management clarified that the amounts sit outside the backlog, because the backlog covers only contracted capacity, and that they represent the ARR available from recontracting rather than a figure already signed.

Assessment: This is the strongest single input the residual-value pillar has received, but the word "incremental" needs care. These GPUs are generating revenue today under their original contracts. The $4B is the revenue they would earn on renewal, not the uplift over what they earn now, and management did not size the current revenue on that fleet. The net uplift is positive (prices are above original contracts), but it is smaller than $4B.

8. Prior-Generation GPUs Repriced Upward 30–50% in One Quarter

"In Q3, pricing for Amperes and Hoppers increased 30% to 50% quarter over quarter, and at prices above their original contracts." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

This is the sixth consecutive quarter of favourable residual-value evidence and the first with a percentage attached to a sequential move. The CEO added that the company is signing multi-year commitments that extend beyond the GPUs' depreciable lives, which means later years earn contract revenue with little or no depreciation against it.

"We are signing multiyear commitments extending beyond the GPUs depreciable lives. Driving margins even higher in their later years as price and operating costs remain fixed but depreciation dissipates." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: The residual-value question that capped our rating at initiation is now answered with prices, volumes and dates. An A100 contracted into 2029 in August and Ampere pricing up 30–50% in November is not a recovery story; it is evidence that the useful life of this equipment is longer than the accounting life.

9. Capex Pulled Forward, and a Flat 2027 on Committed Work Alone

"CapEx totaled $14.5 billion in Q3. Above the high end of our guided range, reflecting strong execution that allowed us to pull forward certain deliveries from Q4." — Nitin Agrawal, Chief Financial Officer

The full-year range of $37–39B against nine-month capex of roughly $30.7B implies $6.3–8.3B in the fourth quarter, about half the third-quarter level. The CFO added that, if the company built only for its committed backlog, 2027 capex would be in line with 2026 while revenue and EBITDA grow significantly. More than $9B of infrastructure already sits in construction in progress waiting to be placed into service.

Assessment: This is the most important strategic statement for the bear case on leverage. A company growing revenue at a run rate above $20B exiting 2026 while holding capex flat would see capital intensity fall sharply in 2027. Management presented it as an option, not a plan, and the 2027 capex guide is still to come.

10. A Date for Adjusted Profitability

"Looking ahead, we expect adjusted net loss to reduce meaningfully in 2027 and to become adjusted net income positive in the first half of 28 and for the full year 2028." — Nitin Agrawal, Chief Financial Officer

This is the first time management has dated adjusted net profitability. It sits alongside the expectation that adjusted EBITDA less interest more than doubles in 2027 from a 2026 base that reached $2,901M in the first nine months. The CFO summarised the third quarter in one line: adjusted operating income nearly tripled from the second quarter while interest expense rose about a third.

Assessment: A dated target is falsifiable, which is exactly what this coverage has asked of management since initiation. The fourth-quarter guide supports the direction: interest of $1.05–1.15B against implied adjusted operating income of $747–847M puts interest at about 1.4 times adjusted operating income, down from 2.2 times this quarter.

11. The Election and the Texas Moratorium

The November 3 elections produced a Texas governor-elect who campaigned on a moratorium on new data-center approvals until the legislature passes new protections. CoreWeave had already welcomed her clarification that permitted projects may proceed, and on the call management said its 2027 deliveries depend on fully permitted sites under construction.

"We also do not see material risks to our 2027 deliveries based on ongoing construction at fully permitted sites." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

The CEO described a letter to the governor-elect presenting CoreWeave as nonpartisan and a good operator, and pointed to the 1.3 GW contracted outside the United States as the strategic response to regulatory concentration.

Assessment: The answer covers 2027 and is silent on 2028, which is where siting risk shows up. Texas exposure was described as an overwhelming majority online or permitted, not as a share of contracted power. Bear #6 stays EMERGING with more weight behind it: the policy risk materialised in one state this month, and the effect on CoreWeave's numbers has not yet been shown.

12. Demand, Diversification and the Next Architecture

"We have 10x as much qualified demand as available capacity through the end of 27. And demand for 2028 is already several times larger than our available capacity even before we have brought it formally to market." — Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

CoreWeave was first to bring NVIDIA Vera Rubin NVL72 into multi-rack production, with the first production workloads for Cognition achieving 4.8 times the token throughput of GB200 NVL72. Forge, launched in September, has become the entry point for enterprise customers: 70% of the new cloud customers added in the quarter were already Forge users. Omni signed its first customer and remains unsized.

Assessment: Demand claims of this size are not new on these calls and cannot be tested from outside. The new evidence is that the company is acting on them: holding 2028 capacity back from the market and raising price twice in a quarter. Both are consistent with a supplier in a shortage, and both lose value quickly if the shortage ends.

Guidance & Outlook

MetricPrior guide (August)New guide (November)Change
FY2026 revenue$12.4–13.2B$13.0–13.3BRaised; range narrowed
FY2026 adjusted operating income$960M–1.15B$1.275–1.375BRaised 26% at the midpoint
FY2026 capex$35–39B$37–39BLow end raised
Q4 2026 interest expenseNot given$1.05–1.15BNew
Exit-2026 annualized run rate$18.5–19.5B$20–21BRaised $1.5B at both ends
Exit-2026 active powerMore than 1.85 GWMore than 1.85 GWMaintained
2027 ARR target (more than $30B)Not mentioned"Meaningfully exceed"Reaffirmed qualitatively
Adjusted net incomeNot givenPositive in H1 2028 and FY2028New
Adjusted EBITDA less interest, 2027Not givenMore than doubleNew
Parent-level capital through 2027Not givenLess than $5B; no unsecured debt at current yieldsNew

Every guided line moved in the shareholder's favour, and the revenue range narrowed from $800M wide to $300M wide with one quarter left. The capex range narrowed from the bottom, which is the conservative direction given that the third quarter pulled deliveries forward. The interest guide is the one figure that grew against the shareholder, at the pace the debt balance implies.

Implied fourth quarter. Subtracting nine-month results ($8,314M of revenue, $528M of adjusted operating income) from the full-year range gives fourth-quarter revenue of $4,686–4,986M (midpoint $4,836M, up 32.1% sequentially) and adjusted operating income of $747–847M (midpoint $797M). That is an adjusted operating margin of roughly 16.5% at the midpoints, above the low-teens level stated in August. It requires adjusted operating income to roughly double from the third quarter, against the near-tripling that the August guide required.

Interest to adjusted operating income, Q4. Interest of $1.05–1.15B against implied adjusted operating income of $747–847M is a ratio of 1.24 to 1.54 times, roughly 1.4 times at the midpoints. The sequence across 2026 would then read 25.5 times, 5.0 times, 2.2 times and about 1.4 times. For the full year, interest of roughly $3.1B still runs about 2.4 times adjusted operating income.

Implied fourth-quarter capex. $6.3–8.3B, against $14.5B in the third quarter. With roughly $9B of prepayments and deposits due over the next three months, the fourth quarter is the first in this coverage in which customer cash could plausibly cover most of the quarter's capital spending.

Street at: consensus stood at $12.86B for full-year revenue, so the new $13.15B midpoint is 2.3% above it, the second consecutive quarter in which guidance landed above the Street.

Guidance style: across the five quarters in which management has set a quarterly range, revenue has landed at or above it every time, and adjusted operating income has now done so in four of five. The exception remains the fourth quarter of 2025. Raising the annual profit guide by 26% after one quarter is the most aggressive annual raise in this coverage, and it rests partly on short-term pricing, so we model slightly inside the range rather than at it.

What the guide does not contain. CoreWeave Omni and the 2027–2028 renewal revenue are outside guidance and outside our model. The 2027 capex figure is still to come, and so is a numeric restatement of the 2027 ARR target.

Analyst Q&A Highlights

Six participants asked eleven questions in a short session. The questions were constructive and several opened by thanking management for the extra disclosure. The recurring lines of questioning were the oversupply scenario, the regulatory backdrop after the election, and how much of the margin depended on short-term pricing. No one asked about receivables, debt maturities, the at-the-market programme or Microsoft.

What Happens to the Model When Supply Catches Up

The opening question went to the centre of the bear case: an industry adding a wall of new capacity in the next one to two years, with imperfect visibility into demand. The questioner asked explicitly for the implications of a normalised or oversupplied market. The answer described current conditions (demand outpacing supply, margins expanding on each contract, capacity sold out through 2027) and did not describe what the model looks like in the scenario asked about.

Q: "So we could have a debate around when supply demand normalizes, whether it is next year or 5 years from now, My question for you is walk us through the implications for the CoreWeave model if and when the industry gets into a period of normalized or oversupply."
— Gabriela Borges, Goldman Sachs

A: "And so yes, you are correct. There is a significant amount of infrastructure that is coming to market But the proliferation of this technology, the diffusion of this technology across the economy is outpacing the ability of that infrastructure to be able to deliver the tokens, to be able to deliver the infrastructure and the clusters to be able to train and deliver infrastructure. We are largely sold out through 2027."
— Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: The question was not answered as asked. The prepared remarks had offered a version of the answer (contracted demand and success-based capex as the base, a paid-off fleet as the hedge if supply is delayed), but the scenario where supply arrives faster than demand went unaddressed. For a model now selling short-term capacity at nearly twice the long-term price, that is the scenario that matters.

How Costs Are Locked When Revenue Is Contracted

A follow-up asked how the company manages the gap between locking in revenue per megawatt and incurring costs that can move before the capacity is live. The answer was specific: purchase orders are placed when the customer contract is signed, and the data center is identified and costed at signing.

Q: "how do you manage the risk between when you lock in that number on the revenue monetization side versus the cost side of the equation, which can sometimes be ephemeral and or move up in the period of time between when you contract the megawatts with the customer versus when you actually bring it online?"
— Gabriela Borges, Goldman Sachs

A: "The way we manage that is our CapEx is fundamentally success based. As we sign customer contracts, we also place the corresponding POs and lock in the pricing for those as we move ahead."
— Nitin Agrawal, Chief Financial Officer

Assessment: A clear answer to a good question. Matching purchase orders to signed contracts removes most component-price risk from the contracted book, which supports the claim that pricing increases are "more than offsetting cost increases". It does not cover power or construction cost overruns at third-party sites, which is where the 2025 delay came from.

Whether the Credit Sell-Off Changes How Capacity Beyond 2027 Is Contracted

The only question that addressed the October credit market directly asked whether it changes the approach to contracting new capacity beyond 2027. The answer reframed the withheld 2028 capacity as a choice to let demand "season" and described a policy of not locking in power too far ahead of client roadmaps.

Q: "I wanted to ask, how, if at all, does some of the recent concerns in credit markets across the industry impact your approach to contracting new capacity beyond 2027, perhaps as we think around time to power, customer types, duration, and when you line up the funding."
— Brad Zelnick, Deutsche Bank

A: "And it is a great opportunity for us to clarify that we have not brought 2028 to market because we are waiting for that to season further."
— Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: The answer is consistent with the capex flexibility the CFO described, but it does not say whether funding conditions are part of the reason 2028 is being held back. If the credit market is shaping 2028 contracting, that is information shareholders would want, and the answer neither confirmed nor denied it.

How Much of the Margin Came From Short-Term Contracts

The second consecutive quarter in which a direct request to decompose the margin drivers was declined. The question separated execution (bringing capacity live earlier in the quarter) from short-term contract pricing. The answer named three drivers and said all three contributed, without proportions.

Q: "You historically talked about also bringing capacity earlier in the quarter. That tends to drive margin upside. But how should we think about maybe execution versus the contribution from shorter term contracts? In terms of the upside on margin, both in the third quarter and going into the fourth quarter."
— Samik Chatterjee, JPMorgan

A: "All 3 of those vectors contributed to our overperformance this quarter. And we continue to expect, you know, those vectors to kind of continue in our favor through the course of the near term."
— Nitin Agrawal, Chief Financial Officer

Assessment: In August the request was to decompose the contribution-margin step-up; this time it was the operating margin beat. Both times the answer was that the drivers cannot be separated. They can be, internally, and the reluctance suggests the short-term share is large enough to matter.

Exposure to the Post-Election Regulatory Shift, and to Texas in Particular

A recurring line of questioning: first on the national regulatory environment after the elections, then specifically on Texas and the governor-elect. The answers were that no impact is anticipated because the backlog's data centers are permitted and under construction, and that the overwhelming majority of Texas power is online or fully permitted.

Q: "I know you have mentioned that you do not see any delays. I was wondering what gave you that conviction, maybe after that reach out just a little more specific and what you are exposure is to that state?"
— Multiple analysts incl. Benjamin Reitzes, Melius Research; Samik Chatterjee, JPMorgan

A: "So within Texas specifically, the overwhelming majority of our power is either online or fully permitted and under construction. And that is where our statement around that we do not believe or anticipate an impact to our data centers that will ultimately host the backlog that we have brought to market."
— Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: The question asked for the size of the Texas exposure and the answer gave its permitting status instead. "The backlog that we have brought to market" is a precise qualifier: it covers contracted capacity, not the 2028 capacity the company is holding back, which is the capacity most exposed to new permitting rules.

Whether Frontier Labs Building Their Own Compute Changes the Customer Mix

The question raised the risk that the largest AI labs will increasingly lease and build their own compute, and asked how fast the company would shift incremental capacity toward enterprise customers and what the composition of the quarter's backlog growth was. The answer gave diversification statistics and did not describe the composition of the backlog growth.

Q: "I think as everybody on the call hears the Frontier Labs signal a desire to lease and stand up their own compute going forward. I think we would all love a little bit of color on how CoreWeave is going to react to that and the pace at which you would like CoreWeave to allocate incremental capacity to non lab, specifically enterprise customers."
— Karl Keirstead, UBS

A: "10% of our backlog is from financial services. Like, there is real diversification there. More than half of the new customers that we added this year are not within the AI space."
— Michael Intrator, Co-Founder, Chairman and Chief Executive Officer

Assessment: The diversification figures are useful and new. The specific request, who signed the $28.3B of backlog growth, went unanswered, and it is the figure that would show whether concentration is falling in the new business. Paired with the report that Microsoft is reviewing its neocloud roster, this is the open question we carry into the next quarter.

Returns After Financing on Short- Versus Long-Term Contracts

A two-part question: how returns differ after financing costs between the two contract types, and what share of capacity the company is comfortable placing on short-term terms. The answer gave hardware payback periods and did not address the share.

Q: "On a comparable hardware utilization basis, how do the returns differ after financing costs? And then, Mike, maybe you can just talk about what portion of capacity are you comfortable having on shorter term contract?"
— Amit Daryanani, Evercore ISI

A: "When we think about you know, the returns, when you are looking for a shorter duration contract, they kind of pay back CapEx in a much faster duration. Somewhere in the tune of 12. You know, 6 to 18 months. And when you look at the longer dated customer contracts, they pay back CapEx somewhere in the domain of hardware CapEx somewhere in the domain of 1 to 2 years at the current pricing levels that we are seeing."
— Nitin Agrawal, Chief Financial Officer

Assessment: A one-to-two-year hardware payback on long-term contracts at current pricing is a striking figure, and if accurate it answers much of the leverage bear case: the asset repays itself well inside a four-to-five-year term. The figure covers hardware only, not data center leases or power, and the second half of the question, how much capacity can go short-term, was not answered.

Whether the Renewal Figures Are Gross Revenue or Uplift

The sharpest follow-up of the call asked whether the $4B and $10B renewal figures were total revenue or uplift, and whether they were already signed. The answer placed them outside the backlog and described them as the ARR available on recontracted capacity at current pricing, which implies gross revenue on that fleet rather than uplift over its current revenue.

Q: "I think you said, initially 4 billion of incremental revenues from renewed contracts in 2027 and $10 billion in 2028. Is that just the total revenue contribution? Or is that the uplift you expect? And is this thing you have already renewed or you expect to do?"
— Amit Daryanani, Evercore ISI

A: "Those are outside our backlog that you see because that is for contracted capacity today. As we recontract this capacity, that is the available opportunity at the current pricing levels for the ARR to be generated against that infrastructure?"
— Nitin Agrawal, Chief Financial Officer

Assessment: The clarification matters more than the headline. The figures are an opportunity at today's prices, not contracted revenue and not net uplift. A reader who adds $4B to the 2027 run rate is double-counting the revenue those GPUs already earn. We treat the renewal disclosure as strong evidence of residual value and keep it out of the valuation.

What They're NOT Saying

  1. The net uplift on renewals. The $4B and $10B are the ARR available on recontracted GPUs, and management did not state the current revenue on that fleet. Without it, the genuinely incremental part of the renewal opportunity cannot be calculated.
  2. Microsoft. A report four days before the call said Microsoft is evaluating which neocloud partners to keep. The name did not appear on the call, and no questioner raised it.
  3. Revenue-basis customer concentration. The table that returned in August is not in the release, and the call offered diversification statistics instead. The top-three figure of 72% is the number we track, and it has no update.
  4. Why receivables rose $1.7B. Days sales outstanding rose to roughly 108 from roughly 90. Receivables grew faster than revenue in a quarter when the company was asking investors to focus on cash generation, and the subject did not come up.
  5. The at-the-market programme. Established in September for up to 35 million shares, mentioned in the release, and not discussed on the call. Whether any shares have been sold, and at what price, was not stated.
  6. Debt due within a year. $10.6B is now classified as current, up from $7.5B. The call discussed new financing at length and did not mention near-term maturities or how they will be met.
  7. The deferred revenue mismatch. The cash flow statement credits $2,127M of deferred revenue inflow in the quarter while the deferred revenue balance rose only $310M and other current liabilities rose $1,945M. The release does not reconcile the two, and the quarterly report had not been filed.
  8. The share of capacity on short-term contracts. Asked directly, not answered. It is the number that separates the durable margin from the cyclical one.
  9. Texas as a share of contracted power. Described only as an overwhelming majority online or permitted. The 2028 capacity being held back from market was not discussed in the context of new permitting rules.
  10. The first self-build site. Expected online later this year, according to the last two calls. Not mentioned on this one.
  11. Sizing for CoreWeave Omni. First customer signed and presented as a lower-capital growth lever. Third consecutive quarter without a revenue figure, pricing model or pipeline.

Market Reaction

  • Pre-print setup: close of $48.73 on November 10, down 32.0% year to date from $71.61, down 40.6% over the trailing 30 days from $82.08, and down 53.9% over the trailing twelve months from $105.61. The 52-week closing range was $43.66 to $137.98, with the low set on November 5. The stock entered the print 54.8% below the $107.73 at which we upgraded it in August, a period in which the S&P 500 fell 2.9%. Options priced a move of roughly 21% in either direction.
  • After-hours move: shares rose about 2% in early after-hours trading and about 7.9% to $52.58 by late evening, as the call's funding commentary circulated. Premarket indications on November 11 showed gains of 13% to 16%.
  • Next-day session: opened at $55.97, a 14.9% gap up; traded a $53.81 to $62.03 range; closed at $60.77, up 24.7% and $12.04 on the day.
  • Volume: 165.1M shares against a 30-day average of 54.7M, a 3.0 times multiple.
  • Peers and benchmark: the S&P 500 rose 1.3% on the session. Nebius rose 15.4% and IREN 5.6%, so part of the move was read as sector news on funding.

This was the largest reaction-day move of the six prints we have covered, in either direction, and the second consecutive positive one. The sequence now reads negative 20.8%, negative 16.3%, negative 18.5%, negative 11.4%, positive 19.3% and positive 24.7%. The move also exceeded the roughly 21% the options market had priced.

The move priced the funding plan more than the quarter. The after-hours gain was small on the headline release and grew through the evening, and the regular session extended it from about 8% after hours to 24.7% at the close. The attribution in the coverage of the print centred on the pledge not to issue unsecured debt at current yields and on the $9B of prepayments, not on the margin beat. That is consistent with how the stock fell: the drawdown came through the credit market, so the recovery began there too.

The drawdown was sectoral, and so was part of the rebound. From October 9 to the day of the print, CoreWeave fell 40.6% while Nebius fell 39.4% and IREN 43.9%. The market treated the neocloud group as a single credit trade, triggered by a post about one bond and reinforced by a forecast that more than 80% of neocloud providers will stop operating independently by 2030, a pulled storage-company IPO, a near-40% fall in an Australian data-center operator, and the Microsoft roster report. Nebius rising 15.4% on CoreWeave's call is the same mechanism running in reverse.

The rebound recovered about a fifth of the decline. The $12.04 gain recovered 20% of the $59.00 lost from our August upgrade price, and 36% of the decline since October 9. In August we wrote that the reaction had not yet priced the new financing structure. The market then went much further than not pricing it and treated the whole capital structure as impaired. Our read is that the November 11 session was the first step in separating the contract-backed debt from the corporate debt, and that this separation has further to run if the investment-grade facility closes on the indicated terms.

Street Perspective

Debate: Is the Income Statement or the Bond Market Right?

Bull view: The bull case on the Street holds that the bond market reacted to a social-media post and to sector contagion, not to anything in CoreWeave's numbers. The company beat every guided operating line, raised the full-year profit guide 26%, quantified a renewal wave on paid-off equipment, and its secured loans trade around par. Bonds yielding 15–16% on a company guiding to adjusted net income in 2028 are mispriced, and the equity will follow the bonds back.

Bear view: The bear camp contends that the bond market is pricing exactly what the balance sheet shows: $46.9B of debt against $4.0B of equity, $10.6B due within a year, net debt up $12.3B in a quarter and free cash flow of negative $7.5B. A long-standing skeptic described the business model as the overhang regardless of the quarter, and several desks cut targets after a beat.

Our take: Both are right about different layers of the capital structure. The contract-backed debt is money good on the evidence, and the market is starting to price it that way. The unsecured parent debt and the equity behind it carry the residual, and their value depends on the income statement continuing to outrun the balance sheet. This quarter it did, decisively. We cut our multiple to reflect the higher cost of capital the bond market has imposed, and the stock still trades well below our revised value.

Debate: Can the Company Fund 2027 Without the Unsecured Market?

Bull view: The bull framing is that the unsecured market was always the most expensive source and is no longer needed. Prepayments of $9B in three months, an investment-grade term loan at 7–7.5% with an advance rate above 90%, nearly $5B of undrawn committed loans, more than $9B of infrastructure already in construction in progress, and EBITDA less interest more than doubling in 2027 together leave less than $5B to raise at the parent.

Bear view: The skeptics argue that the plan relies on prepayments, which cost price and concentrate exposure on the same customers, and on an at-the-market equity programme that would now issue about 24% below the price at which it was arranged. A funding plan with one channel closed has less room for a delay, a customer loss or a second credit shock.

Our take: The bulls have the arithmetic and the bears have the margin of safety. Our base case assumes the plan works as described. We carry at-the-market dilution in the bear case only; full use at today's price would lower the base midpoint by about 2% after crediting the proceeds. We would downgrade on evidence that the parent needs materially more than $5B, because that is the point at which dilution at current prices destroys value faster than the income statement creates it.

Debate: Is Short-Term Pricing a Durable Premium or the Top of a Cycle?

Bull view: A growing view holds that installed, paid-off capacity becomes more valuable the more the industry struggles to build. Ampere and Hopper prices rose 30–50% in a quarter, short-term capacity sells at about $40M per megawatt, and if construction slows across the sector because of credit conditions and permitting, the companies already holding capacity earn scarcity rents.

Bear view: The bear camp notes that the oversupply question was asked directly and answered with a description of today's market. Pricing at nearly twice the long-term rate is what a shortage looks like, and shortages in capital-intensive industries end with the capacity they finance. The more of the book that moves to short-term terms, the more of the margin is exposed when they do.

Our take: The bears are right that the short-term premium is cyclical and wrong to treat that as a threat to the contracted base. We value on the exit-2026 run rate, which is weighted toward long-term contracts, and we exclude the renewal opportunity from the valuation entirely. If short-term pricing holds, that is upside we have not counted.

Debate: Do Hyperscaler Consolidation and Texas Policy Cap 2028?

Bull view: The constructive reading is that both risks touch capacity the company has not yet sold. The 2027 deliveries sit on permitted sites, 1.3 GW of contracted power is outside the United States, and the diversification statistics show the customer base widening into enterprises and financial services.

Bear view: The skeptical reading is that the two risks hit the same year from two directions. The company's largest historical customer is reported to be reviewing its neocloud roster while building its own capacity, frontier labs are signalling the same, and the state that hosts much of the industry's planned capacity has a governor-elect pledged to pause new approvals.

Our take: Neither risk touches the current backlog, and both touch the growth rate after it. We do not change 2026 or 2027 numbers for either. We move the customer-concentration risk to emerging and add weight to the siting risk, and we will treat the 2027 capex guide and the 2028 contracting pace as the evidence.

Model Update & Valuation Framework

ItemPrior (Q2 2026 recap)Updated (Q3 2026 recap)Reason
FY2026 revenue$12.4–13.0B$13.0–13.2BCompany guide raised to $13.0–13.3B; we trim the top for Q4 deployment timing
FY2026 adjusted operating income$900M–1.10B$1.275–1.35BQ3 beat by $119M; modelled inside the guide because part of the beat rests on short-term pricing
Implied Q4 adjusted operating income~$676M at guide midpoint$747–822MNine-month actual of $528M; Q4 step now ~2.1x the Q3 level
FY2026 capex$35–39B$37–39BCompany guide; low end raised on earlier deliveries
Exit-2026 annualized run rate$18.5–19.5B$20–21BCompany range raised $1.5B at both ends
Exit-2027 annualized run rateMore than $30B, not reaffirmedMore than $30B, "meaningfully exceed"Terminal sanity check only
FY2026 exit net debt$38–44B (base $41B)$41–47B (base $43B)Net debt already $40.4B; Q4 capex of $6.3–8.3B partly offset by prepayments
Share count (valuation basis)551M559MQ3 weighted average; at-the-market shares in the bear case only
Base-case EV multiple on run rate5.25–6.25x4.25–5.0xHigher cost of capital: unsecured yields of 15–16% against coupons of 8.5–9.75%; parent-level market closed at current yields
Renewal revenue (2027–2028)Not modelledNot modelledGross opportunity outside backlog; net uplift not disclosed
12-month price target$124$93Multiple cut outweighs the run-rate raise
Conviction67Two of four higher-conviction conditions fired
RatingOutperformOutperformNo downgrade condition fired; 53% implied return

Valuation base unchanged in method. We continue to value on management's exit-2026 annualised run rate less an explicit exit-2026 net debt assumption, divided by the weighted-average share count. The method has not changed since February, so the inputs can be compared directly with every prior recap.

At the November 11 close of $60.77 and 559M shares, equity value is approximately $34.0B. Adding total debt of $46.87B and subtracting cash, restricted cash and marketable securities of $6.47B gives an enterprise value of roughly $74.4B. That is 5.7 times the midpoint of the FY2026 revenue guide, 3.6 times the midpoint of the exit-2026 run rate, and under 2.5 times the exit-2027 target. On the same basis the August price was 4.6 times the then-current run-rate midpoint; today's enterprise value is 3.9 times that same August midpoint of $19B. The stock entered the print at 3.3 times.

Why the multiple comes down by more than a turn. The multiple carries the cost of capital, and the cost of capital changed. When we set 5.25–6.25 times in August, the company's unsecured notes carried coupons of 8.5–9.75% and the weighted cost of debt was falling. Those notes now yield 15–16% and the parent-level market is closed to new issuance at those yields. An equity standing behind that debt has to clear a higher hurdle, whatever the income statement does. We set the base at 4.25–5.0 times, a cut of 1.125 turns at the midpoint, and we would restore part of it if the investment-grade facility closes on the indicated terms and unsecured yields retrace.

ScenarioExit-2026 run rateEV/Run rateImplied EVLess net debtImplied equity valueSharesPer sharevs. $60.77
Bear$18.5B3.0–3.5x$55.5–64.8B$47B$8.5–17.8B594M$14–30−76% to −51%
Base$20.5B4.25–5.0x$87.1–102.5B$43B$44.1–59.5B559M$79–106+30% to +75%
Bull$21B6.5–7.5x$136.5–157.5B$41B$95.5–116.5B559M$171–208+181% to +243%

Twelve-month price target: $93 (base range $79–106). The target falls $31, and the composition of the change is the honest summary of the quarter. The raised run rate adds about $15.60 per share at the August multiple. The lower multiple removes about $41.90. The $2B higher net debt assumption removes about $3.60 and 8M more shares remove about $1.30. The business added value in this quarter and the cost of capital took more of it away. $93 against $60.77 implies roughly 53% over twelve months.

Dilution sensitivity. Full use of the 35M-share at-the-market programme at the current price would lower the base midpoint by about 2% after crediting the proceeds against net debt. We carry those shares in the bear case only, where the price at which they would be issued is lower still.

What differentiates the scenarios. The bear case is a funding failure rather than a demand failure: the run rate lands below the guide at $18.5B, the parent raises equity at depressed prices, net debt reaches $47B, and the market keeps applying a distressed multiple. The base case assumes the funding plan works as described and the market applies a multiple below where it stood in August. The bull case requires the investment-grade facility to close, unsecured yields to normalise and the multiple to recover toward its August level, with no change to the operating guide. The distribution is as wide as at any point in this coverage, because the equity is now the residual claim on a capital structure the credit market has questioned.

Thesis Scorecard Post-Earnings

Thesis pointStatus at Q2Status nowWhat this quarter showed
Bull #1: Contracted demand exceeds deliverable supplyConfirmedConfirmedBacklog $132.5B (+138% YoY); 10x qualified demand vs. capacity through 2027; two price increases in one quarter. Most of the backlog growth was added in the early weeks of Q3, and 2028 capacity is being held back
Bull #2: Contracted power is the moatConfirmedConfirmedMore than 1.7 GW active, more than 4.6 GW contracted (+900 MW), 1.3 GW abroad; Q3 energisation slower at more than 200 MW; Texas policy risk to future sites
Bull #3: GPU assets retain economic value past their initial termConfirmed, externally pricedConfirmed, now quantifiedAmpere and Hopper pricing +30–50% QoQ above original contracts; 130,000 GPUs off contract in 2027 and ~170,000 in 2028; more than $3B of prepayments on recontracted, paid-off infrastructure
Bull #4: Financing capability is a competitive advantageStrengthenedAt riskUnsecured market closed at 15–16% yields; at-the-market programme established; secured market open with an investment-grade facility indicated at 7–7.5%; weighted cost of debt 7.9%
Bear #1: Interest expense outgrows operating profitMaterializing, improvingMaterializing, improving sharply2.2x from 5.0x; Q4 implied ~1.4x; adjusted net income dated to H1 2028. Interest still exceeds adjusted operating income and Q4 interest rises 24–36%
Bear #2: Customer concentrationContainedEmergingNo revenue-basis concentration figure available; Microsoft reported to be reviewing its neocloud roster; diversification statistics improved (10% of the Fortune 500, 10% of backlog from financial services)
Bear #3: Growth funded by leverage, not cash flowConfirmed, escalatingConfirmed, escalatingTotal debt +33.7% to $46.9B; net debt +43.5% to $40.4B; equity down $1.0B to $4.0B; free cash flow negative $7.5B. Offsets: operating cash flow $1,975M, $9B of prepayments due, flat 2027 capex on committed work
Bear #4: Deployment timing riskContainedContainedQ4 deliveries pulled into Q3; construction in progress down to $9.3B; Q4 step now ~2.1x rather than ~2.9x
Bear #5: Valuation absorbs the thesisRecededReceded, inverted3.6x the exit-2026 run rate against 4.6x in August; the stock trades 43% below the bottom of our August base range
Bear #6: Siting and permitting riskEmergingEmerging, heavierTexas governor-elect pledged a moratorium on new approvals; management says 2027 deliveries sit on permitted sites; Texas share of contracted power not disclosed

Grading Last Quarter's Commitments

Commitment we said we would watchOutcomeVerdict
The make-or-break: Q3 adjusted operating income of $200–260M$379M, $119M above the high endBeaten
Q3 revenue $3.45–3.6B; interest $860–940M; capex $11.5–13.5B$3,661M; $846M; $14.5BBeaten, beaten; capex exceeded
FY2026 $12.4–13.2B revenue and $960M–1.15B adjusted operating income; implied Q4 $551–801MRaised to $13.0–13.3B and $1.275–1.375B; implied Q4 now $747–847MRaised
Exit-2026 run rate $18.5–19.5B; exit power above 1.85 GW; exit-2027 framework reaffirmedRun rate raised to $20–21B; power maintained; 2027 target to be "meaningfully" exceeded, number not restatedRaised; 2027 reaffirmed qualitatively
5–10 point contribution-margin step-up repeated or quantifiedNot restated in points; newest contracts "among the most profitable we have ever underwritten"; $20–25M per MW on new long-term contractsPartially
July ~25% price increase visible in pricing or mixA further ~10% increase later in Q3; Q4 pricing higher; pricing cited as a driver of the margin beatDelivered
Size of the fleet approaching recontractingMore than 130,000 GPUs in 2027, about 170,000 in 2028Delivered
Any spot, Flex or recontracted clearing priceAbout $40M per MW a year short-term, given both as a floor and as a cap; Ampere and Hopper +30–50% QoQ, above original contractsDelivered
Managed inference ARR against the $250M exit-2026 target; platform ARRMore than $250M booked, a quarter early; platform ARR not updatedMet early
First quantified CoreWeave Omni contributionFirst customer signed; no figureNot delivered
First self-build site online later this yearNot mentionedNot addressed
Top-three revenue concentration (72%) and whether the disclosure survivesNot in the release; quarterly report not yet filedNot available
Operating cash flow recovery from $679M; deferred-revenue inflow repeatable$1,975M; $2,127M of deferred revenue inflow; ~$9B of prepayments due in three monthsMet

Grading the Rating Conditions We Set in August

Condition, as published in AugustDirectionOutcomeStatus
A third quarter at or above the high end of $200–260MHigher conviction$379MFired
Disclosure of the size of the fleet approaching recontractingHigher conviction130,000 GPUs (2027), ~170,000 (2028)Fired
A spot or Flex clearing priceHigher convictionAbout $40M per MW a year on short-term contracts, given both as a floor and as a cap: a contract rate, not a Spot or Flex clearing priceNot fired
A quantified CoreWeave Omni contributionHigher convictionNo figureNot fired
A third quarter below $200M of adjusted operating incomeHold or lower$379MNot fired
A full-year profit guide cutHold or lowerRaised 26% at the midpointNot fired
The first sequential decline in prior-generation GPU pricingHold or lower+30–50% QoQNot fired
A counterparty credit event among the top three customersHold or lowerNone reported; Microsoft roster review is not a credit eventNot fired
A financing that widens spreads on the tranches where lenders do not take residual riskHold or lowerNo such financing completed. Secured loans said to trade around par. The next facility is indicated at a 7–7.5% yield against DDTL 4.0's sub-6%, with base rate and spread not separatedNot fired; on watch

The unsecured bonds that sold off in October are parent-level debt, where holders carry the full enterprise risk, including residual value. The fifth downgrade condition was written about the contract-backed tranches, and we grade it against those. It fires if the next investment-grade facility prices at a wider spread over SOFR than DDTL 4.0 on a comparable rating. We state now how it would net: if that condition fires and nothing else does, it offsets the higher-conviction conditions that fired this quarter, and conviction returns to 6 with the rating unchanged.

Overall: Nine of thirteen commitments met, beaten, raised or delivered, one partial, and three not delivered or not available, with capex exceeded for the reason management gave. Two of four higher-conviction conditions fired and none of the five downgrade conditions did. On operating evidence this is the strongest quarter in six quarters of coverage. On the balance sheet it is the most stretched, and the market spent three months pricing the second fact while ignoring the first.

Action: Maintain Outperform, target $93 from $124, conviction 7 from 6. We said in August which conditions would raise our conviction and which would take us to Hold, and we honour them rather than re-cut them after the stock fell 55%. Two of the four positive conditions fired, and the August wording needed only one. The negative ones did not. We acknowledge plainly that the August upgrade has so far cost 43.6% against an index down 1.7% to the November 11 close. The risk we named, leverage, priced in through the credit market rather than through the income statement, and much faster than we expected.

Why conviction rises only to 7. The operating conditions fired, but the risk that hurt the stock was not one of them. The unsecured market is closed at current yields, the funding plan has less slack than it did in August, the customer-concentration figure is unavailable at the moment the largest historical customer is reviewing its suppliers, and the at-the-market programme sits ready to issue equity well below our value. An equity this levered remains an option on a specific operational outcome and should be sized as one.

What would take us to a higher conviction: fourth-quarter adjusted operating income at or above $847M, the top of the implied range; the investment-grade facility closing inside the indicated 7–7.5% with an advance rate above 90%; a revenue-basis concentration table with the top three below 72%; disclosure of the current revenue on the GPUs coming off contract in 2027, so that the renewal uplift can be measured net; or a quantified CoreWeave Omni contribution.

What would take us to Hold or lower: fourth-quarter adjusted operating income below $747M, the bottom of the implied range and of our own model; the first sequential decline in prior-generation GPU pricing; a counterparty credit event among the largest customers, or a disclosed non-renewal or reduction by Microsoft; a financing that widens spreads on the tranches where lenders do not take residual risk; or parent-level capital needs above the $5B stated for the period through 2027, including any new unsecured issuance at current yields. A higher-conviction condition and a downgrade condition firing together offset each other, and the rating holds.

Independence Disclosure As of the publication date, the author holds no position in CRWV and has no plans to initiate any position in CRWV within the next 72 hours. Aardvark Labs Capital Research maintains a firm-wide policy of not trading any security we cover. No compensation has been received from CoreWeave, Inc. or any affiliated party for this research.