CAPITAL THESIS
EQUITY RESEARCH / REVISED COVERAGE / 15 OCTOBER 2026

CoreWeaveThe platform improves.
The equity still has to earn its return.

From GPU capacity to an AI platform: the operating opportunity is improving, but financing, replacement spending and dilution determine the return to shareholders.
NASDAQ: CRWV · Technology / AI infrastructure · USD unless stated

Report price $65.78 · 14 October 2026, 4:00 pm EDT close
Latest actuals: June 2026 · Four distinct supplied research notes · Five-year conditional scenarios

CAPITAL THESIS VIEW · WATCHLIST

A better AI platform. A demanding equity equation.

The new evidence improves the strategic case: CoreWeave is combining scarce compute capacity with tools for inference, training and model improvement. Better performance, broader product adoption and older-GPU renewals could support pricing and asset returns. The financial test remains harder: convert those advantages into cash after hardware replacement, interest and dilution.

Our five-year base remains 14.1% annualised, conditional on financing access and substantial growth. It sits below a 20% hurdle; the bear case leaves no common-equity value. The strongest reason to keep researching is a widening platform opportunity. The strongest reason to wait is the absence of a reconciled, current capital stack and repeatable cash returns by capacity cohort.

This revised edition incorporates four distinct supplied research notes. It adds platform analysis, a dated broker comparison and source reconciliation. The original Capital Thesis financing scenarios remain unchanged: the notes strengthen the operating debate without resolving its cash and capital uncertainties.

$65.78Frozen report price · 14 Oct close
$104bnJune revenue backlog · conditional
−$13.66bnTTM cash flow after cash capex
14.1%Conditional base five-year CAGR

Research date: 15 October 2026, Sydney. Financial actuals through 30 June; share snapshot 31 July; material releases checked through 14 October US time. Reported financial evidence is high confidence; current funding reconstruction and long-run return assumptions are materially less certain.[S1][S2][S3][S7]

REPORTED · filingsGUIDANCE · managementCONSENSUS · external estimatesSCENARIOS · house assumptions

1. The central investment debate

The company thesis is that customers will keep paying for fast, reliable access to large GPU clusters and the software around them. The stock thesis adds a harder requirement: the return on that deployed equipment must exceed its replacement and financing cost, with enough residual value per share to compensate for customer and liquidity risk. Those are separate propositions.

At the October close the market capitalisation is about $36.28bn using the July A+B share count. Add the June contractual debt and finance leases, subtract unrestricted cash and securities, and the dated-balance enterprise-value reference is $66.51bn. It is 15.12 times trailing adjusted EBITDA. This mixes a current share price with older balance-sheet data; it is a transparent reference, not an assertion of October net debt or a complete current diluted enterprise value.[S2][S7][S9]

AnchorValue / interpretation
Security / report priceNasdaq: CRWV · Class A common · USD 65.78 · 14 October 2026, 4:00 pm EDT close
Basic economic share count551.536602m A+B shares · 31 July 2026; ClassA alone understates equity capitalisation
Dated-balance EV / TTM sales$66.51bn / $7.590bn = 8.76×
P / E, PEG and current owner-earnings yieldNot meaningful: GAAP losses and negative cash flow after equipment spending
Consensus contextFY2026 revenue $12.87bn; FY2027 $26.30bn. Adjusted EPS estimates remain negative.
Public 12-month analyst targetAverage $141.47, range $39–278; different horizon and models, not our valuation input
Market observations are dated; external estimates are not company promises. No quantitative peer-median valuation is claimed. The latest balance sheet cannot establish a same-day capital stack. [S2][S7][S8][S9]
DebateWhat could create per-share valueWhat could prevent it
Deployment advantageCluster performance, usable power and software reduce customers’ time to productive workloads.Delivery delays leave rent and interest running before billing starts.
Asset economicsEquipment retains profitable demand through renewals after debt paydown.Faster obsolescence or lower renewal prices destroy residual cash returns.
Financed scaleLong contracts and prepayments support economical capacity funding.Debt, collateral requirements and equity issuance absorb operating growth.
Analyst interpretation, not a claim that a proven mispricing exists.

Our variant question is the durability of after-financing cash per diluted share. The base revenue path is close to the public 2027 revenue estimate at first, then assumes continued expansion. It is not a unique near-term earnings forecast. A cheap-looking revenue multiple or a sharp share-price fall would not by itself establish a margin of safety.

2. Platform economics and customer risk

CoreWeave supplies AI compute, storage and networking, with orchestration and development tools around the infrastructure. Specialisation can improve delivery and performance, while larger cloud platforms can compete through financing capacity, existing customer relationships and bundled services. Some counterparties are simultaneously customers, competitors, investors or suppliers; these relationships are commercially useful but not independent sources of risk diversification.[S1][S2]

The latest quarter reported 1.5GW of active power and approximately 3.7GW contracted power at June-end; contracted power reached 4.2GW by 11 August. Contracted sites must still become delivered, energised and billable capacity. September’s Rubin deployment with Cognition and Forge launch provide execution evidence, but neither release supplies enough economics to justify treating the entire business as high-margin software. The October India announcement describes a planned 240MW campus, with its first phase expected in mid 2028 and further capacity phased, not near-term revenue.[S3][S6][S16][S17][S18][S19]

Revenue visibilityWhat the evidence saysWhat it does not establish
June RPO$103.7bn; 41% expected within 24months, 39% in months 25–48, remainder in months 49–78.Cash already collected, immediate revenue, or present equity value.
Revenue backlogApproximately $104bn atJune 30; includes estimates beyond RPO. More than $25bn new commitments earlyQ3 were separately disclosed.A reconciled current backlog of $129bn after revenue recognition and changes.
Customer concentrationQ2 customers A / B / C represented 36% / 26% / 10% of revenue:72% combined.The identities of anonymised A / B / C, or unconcentrated payment risk.
Take-or-pay structureCommitted capacity supports visibility, subject to delivery and service requirements.Immunity from performance credits, termination provisions or customer insolvency.
Primary June disclosures and company definitions. [S2][S3]

NVIDIA’s $2bn January equity investment and technology collaboration support access and execution. Its separately disclosed residual-capacity order is subject to contract conditions; it is not an unconditional guarantee of every GPU, loan or common share. Selected A100 and H200 renewal examples provide counterevidence to an immediate-obsolescence thesis, but cannot prove fleet-wide residual value. Similarly, roughly $40m annualised revenue perMW on selected 3–6month contracts must not be multiplied by the entire power pipeline to manufacture a revenue forecast.[S13][S14][S15][S16]

The competitive test is whether customers renew at returns that cover capital consumed. Useful evidence would include billed utilisation, realised pricing by GPU generation, full replacement expenditure and cash return after financing. The public material does not provide a complete cohort-level return schedule. That gap is central to the watchlist posture.

The platform opportunity: move up the customer workflow

The investment opportunity is broader than renting GPUs. CoreWeave can earn a greater share of a customer’s AI workflow if infrastructure, data movement, model evaluation and production feedback work together reliably. Deutsche Bank and J.P. Morgan describe this strategic transition after Fully Connected. We regard it as a plausible source of differentiation, not proof that the company already deserves the valuation or capital intensity of a software business.[R1][R2][S18]

LayerCustomer problem addressedPotential shareholder benefitEvidence still needed
Compute and powerDeliver productive GPU clusters at scale.Revenue starts sooner; reliable performance can support price.Time from spend to billable capacity and cash return by cohort.
Storage and orchestrationMove data and keep large workloads productive.Higher retention and more services per customer.Separately disclosed economics and service attachment.
Forge and model developmentConnect experiments, observations, evaluation and improvement.Recurring workflow adoption and potential mix benefit.Recognised software revenue, retention and margins.
Inference and agentsOperate production workloads and learn from results.Broader use cases, customer base and equipment demand.Inference mix, unit cost and realised contract cash returns.
Capital Thesis causal framework. Product scope is company disclosure; the shareholder benefits are hypotheses. [S17][S18][R1][R2]

The supplied notes cite more than $400m of non-GPU infrastructure run-rate revenue and 70% of customers using at least three products. Those are encouraging company / conference measures. They are not audited software segment revenue, a revenue-weighted adoption rate or evidence of a particular margin. We would look for consistently defined disclosure over time before allowing these observations to lift the base-case valuation multiple.[R1][R2]

Cognition’s reported Rubin results are concrete but narrow: up to 4.8× total-token throughput on SWE-2 inference and 3.8× output-token throughput on reinforcement-learning workloads against GB200 NVL72. Different workloads and denominators must stay separate. Better throughput may help customers justify pricing, but it does not translate one-for-one into CoreWeave revenue, profit or share value.[S17][R2]

Renewals support the asset-life thesis; short contracts add a trade-off

The A100 contract extending to 2029, highlighted in the UBS read-through, is useful evidence against the claim that an older GPU immediately becomes worthless. Selected renewals can create a second cash-generating life after initial financing has amortised. The missing bridge is fleet-wide utilisation, renewal prices, maintenance and replacement spending. A successful example cannot establish the residual value of all equipment.[R4][S6][S15]

Recent list-price increases and selected short-term capacity agreements support the demand argument. They also create a choice: capture a near-term premium or lock in duration and customer diversity. Shorter contracts can improve current economics while increasing future repricing and utilisation risk. A management pipeline described as more than ten times supply is a prospect measure, not signed backlog. We do not add it to RPO or project a fleet-wide price increase from selected contracts.[R1][R2][S16]

3. What the external research changes

The research notes give a useful range of operating views, not a set of interchangeable price targets. We use them to identify what must be proved, retain their original dates and separate them from our own assumptions. Positive conference observations can share the same management source; several analysts repeating them does not create several independent confirmations.

External view and dateWhat it contributesHow Capital Thesis uses it
Deutsche Bank · 1 October · Buy / $150Enterprise readiness, product adoption and a broader AI platform.Raises the importance of product-level adoption and monetisation evidence; does not supply a reconstructed target bridge.
J.P. Morgan · 30 September · OverweightInference, Rubin performance and potential services-led margins.Adds technical and pricing context. The $125 target is shown in 24 September rating history; this note supplies no complete valuation bridge.
Barclays · 12 August · Equal Weight / $105Detailed operating forecast, higher contract contribution margins and financed growth.Provides a dated estimate comparator. Its 12-month target is separate from our five-year terminal outcomes.
UBS · 12 August · Megaport read-throughOlder-GPU renewals and inference demand as industry signals.Uses CRWV commentary only. The A$24.20 target and AUD forecasts belong to Megaport and are excluded.
External research opinions, with their own disclosed potential conflicts. No affiliation with or endorsement by these institutions is implied. [R1][R2][R3][R4]

The real disagreement is margin and funding, not just revenue

USD bn except margins · dated estimates2026E2027E2028E2029E
Barclays revenue12.98925.68640.45354.966
Barclays adjusted EBITDA7.76716.90927.59238.136
Barclays adjusted EBITDA margin59.8%65.8%68.2%69.4%
Barclays adjusted EBIT1.0664.1937.48010.290
Capital Thesis base revenue12.800*26.00035.00044.000
Capital Thesis base EBITDA margin60.0% H2*60.0%60.0%59.0%
Barclays estimates dated 12 August 2026, detailed income statement p. 7. Capital Thesis values are conditional assumptions, not consensus. *2026 house revenue anchors the hypothetical opening bridge; 60% applies to H2 only and is not a directly comparable full-year margin. [R3]

Our 2027 revenue assumption starts close to Barclays, but our later revenue and EBITDA margins are lower. The broker’s 2027–2029 margin progression from 65.8% to 69.4% is a substantial operating claim. It would need durable pricing, productive deployment and enough services mix to offset infrastructure costs. Better contract contribution margins are promising; they do not automatically become consolidated margins or free cash flow. Our base is deliberately slower to credit that transition.

Barclays describes a roughly 33× CY2027 adjusted-EBIT framework, about $4.19bn of EBIT and approximately $72bn of FY2027 net debt. Its $105 target is a 12-month research output, not our terminal value. The rounded supplied inputs and weighted-average diluted share count do not reproduce $105 exactly, so we attribute the target without claiming to have reconstructed its full valuation. We do not blend it with our five-year prices or treat the mean of broker targets as fair value.[R3]

Source reconciliation changes what can safely enter the model

Issue in the supplied notesTreatment in this report
Barclays cash-flow summary versus detailed modelFY2027 summary FCF is −$27.674bn, while detailed p. 9 shows −$31.257bn. No reconciling definition is supplied. These are not imported as a unified cash model.
Company capex versus cash equipment paymentsQ2 company-defined capex near $9.4bn is different from $6.422bn of cash equipment / software purchases. Financing and timing matter; the cash-flow statement controls historical cash analysis.
UBS exit-run-rate range and Q3 operating profitUse official $18.5–19.5bn exit run-rate guidance. Q3 adjusted EBIT is $200–260m, midpoint $230m; $260m is the top end.
Broker historical figures and peer comparisonsUse primary filings for actuals. Infrastructure depreciation and funding prevent a casual comparison with capital-light software margins.
The differences are material to the equity funding bridge. They are surfaced rather than averaged away. [S2][S3][S5][R3][R4]

4. Earnings, cash and the capital stack

USD bnFY2024FY2025H1 2025H1 2026TTM June 2026
Revenue1.9155.1312.1944.6537.590
GAAP operating profit / (loss)0.324−0.046−0.008−0.193−0.231
Net loss−0.863−1.167−0.605−1.366−1.928
Adjusted EBITDA1.2193.0931.3602.6674.400
Interest expense, net0.3611.2290.5311.1761.874
D&A0.8632.4541.0032.5403.991
Expensed stock compensation0.0310.6300.3290.3180.619
Operating cash flow2.7493.058−0.1903.6636.911
Cash equipment / software purchases8.70210.3093.86014.11720.566
Calculated cash FCF−5.953−7.251−4.050−10.454−13.655
Reported statements, except company-adjusted EBITDA and calculated cash FCF / TTM. TTM = FY2025 +H1 2026 −H1 2025. For example adjusted EBITDA:3.093+2.667−1.360=4.400. [S1][S2][S3][S4]

Cash FCF here is operating cash flow less cash purchases of equipment and capitalised software. It excludes other investing outflows, principal repayments and the economic cost of stock dilution. H1 operating cash included $1.365bn from deferred-revenue growth and $607m from receivables collection. Those timing benefits are not repeatable profits. Cash equipment purchases also differ from accrual-based or financed capex: the filing reports substantial non-cash equipment additions and OEM financing. Adding every supplemental line to cash capex would double count parts of the spending bridge.[S2]

Adjusted EBITDA adds back depreciation and stock compensation, among other items. GPUs wear out economically even when depreciation is non-cash. Much of that depreciation sits in technology and infrastructure expense, so a simple revenue-minus-cost-of-revenue gross margin misses a major fleet cost. The six-year equipment accounting life is an estimate, not proof of six years of attractive customer pricing.[S1][S2]

June 30 capital claims · USD bnAmountTreatment
Debt before discounts / issuance costs35.551Includes recourse and non-recourse financing; obligations remain senior to common equity.
Debt carrying value35.068Accounting value; model starts from contractual debt table before discounts.
Finance leases0.221Included in model financing claims.
Unrestricted cash + securities5.539Deducted once.
Restricted cash1.380Excluded from freely available cash.
Net debt including finance leases30.23335.551+0.221−5.539; dated June reference.
Operating lease liabilities16.319Disclosed separately; operating rent stays in model EBITDA, avoiding double deduction.
Not-yet-commenced leases35.5Undiscounted future commitments, plus other contingent sites; not simply added as current PV debt.
Operating lease PV and future cash payments are different measures. June debt principal maturities included $4.413bn in remaining 2026 and $6.184bn in 2027, separate from leases. Undrawn facility availability is not cash. [S2]
Post-June financingVerified eventLimit
August facilities$1.2bn DDTL5.5 draw and separate $1.2bn revolver draw disclosed in subsequent events.A$2.6bn facility ceiling is not $2.6bn already borrowed.
September converts$4.2bn closed 22 September; net $4.137bn before $566.2m capped calls and other expenses.Preliminary $3.7bn pricing was superseded.
September ATMAuthorised up to 35m shares, including potential forward transactions.Actual sales, prices, settlement and proceeds not verified.
Cash available from the September notes after purchaser discount and capped calls was $3.5708bn before other expenses / use. Continued spending and repayments prevent reconstruction of an actual October balance sheet from announcements alone. [S2][S9][S10][S12]
Dilution bridgeMillion shares / rule
July 31 basic A+B551.536602
Gross options / RSUs / warrants23.641274 / 29.328373 / 11.864667
Basic plus gross award overhang616.370916; no exercise-proceeds benefit; not GAAP diluted shares
2031 / 2032 / 2033 initial conversion shares24.002685 / 33.444800 / 42.921480; aggregate 100.368965
All initial conversions plus gross overhang716.739881 before ATM or future awards; remove $10.7875bn converted principal
Authorised but ungranted plan reservesExcluded; authorisation is not existing award dilution
Loss-period diluted EPS omits anti-dilutive securities and is not an economic cap table. The later prospectus warrant count exceeds the June EPS-table count; detailed instrument reconciliation remains open. [S2][S9][S10][S11]

Initial conversion prices are approximately $107.80, $119.60 and $97.85. The scenarios add conversion shares only when terminal value supports conversion and remove the corresponding debt principal at the same time. This is an economic as-converted sensitivity, not a prediction that contractual conversion triggers will be met: the 2032 and 2033 notes mature after the terminal period and early conversion is conditional. They assume initial-ratio share settlement, ignore capped-call benefits and make-whole adjustments, and include all reported option / RSU / warrant overhang without exercise proceeds. This is an explicit gross-dilution convention, not a claim to know the exact future settlement outcome.

5. Five-year scenarios after financing

These are conditional operating-and-financing sensitivities. They value terminal enterprise earnings, subtract the financing claims left after the cash path, and divide by economic diluted shares. EBITDA is used only because current GAAP earnings and FCF are negative; the funding path and terminal cash yield constrain its interpretation. A terminal multiple is a house assumption, not an independently validated DCF or peer-derived fair value.

Terminal equity = max(0, terminal EBITDA × exit multiple − net debt after conversion)
Share value = terminal equity ÷ diluted shares
Five-year wealth CAGR = (terminal share value ÷ $65.78)1 / 5 − 1

No common dividends, buybacks or other cash distributions are assumed. Operating rent is included in EBITDA margins; finance leases are in debt. Cash tax is 25% of positive EBITDA less depreciation and cash-interest proxy, with no loss-carryforward benefit. All forecast rates below are analyst assumptions, not management guidance. Five annual model periods use FY2027–2031 scale proxies after an illustrative late 2026 funding bridge. The five-year return convention is rounded for comparability with the library; this is not a calendar-dated cash-flow IRR.

Assumption / terminal resultBearBaseBull
Revenue path · USD bn18 / 22 / 25 / 28 / 3026 / 35 / 44 / 52 / 6028 / 42 / 58 / 75 / 90
EBITDA margins48 / 46 / 45 / 44 / 43%60 / 60 / 59 / 58 / 57%62 / 62 / 61 / 60 / 60%
Cash capex / revenue85 / 70 / 55 / 45 / 40%100 / 75 / 55 / 45 / 38%100 / 70 / 50 / 40 / 33%
D&A / revenue45 / 43 / 40 / 38 / 36%42 / 38 / 35 / 32 / 30%40 / 35 / 31 / 28 / 26%
Blended cash funding rate11.0%8.5%7.0%
New annual share dilution3.0%2.0%1.5%
Terminal EV / adjusted EBITDA3×5×6×
Terminal EBITDA · USD bn12.9034.2054.00
Terminal enterprise value · USD bn38.70171.00324.00
Net debt before conversion · USD bn102.8477.6553.40
Converted principal removed · USD bn0.0010.7910.79
Diluted shares · millions755.12819.54802.08
Terminal common-equity value · USD bn0.00104.14281.39
Terminal share value$0.00$127.07$350.82
Five-year wealth CAGR-100.0%14.1%39.8%
Forecasts are house assumptions. The bear funding path is not plausibly assured: closure of financing can impair equity earlier. Zero is an equity-value floor, not a forecast of a certain restructuring.

The base assumes scaling and cash conversion improve but that a mature, capital-intensive business receives only 5× EBITDA, well below the dated 15.12× trailing reference. The bull earns 6× through better growth, cash generation and funding access; it does not rely on expansion from today’s high trailing multiple. The bear applies 3× to weaker economics and leaves no common equity after senior claims. These multiples are judgmental; cash-flow checks below show why even 5× does not automatically mean cheap.

Visible ranking evidenceValue
BearBear five-year total-return CAGR: -100.0%.
BaseBase five-year total-return CAGR: 14.1%.
BullBull five-year total-return CAGR: 39.8%.
Without reratingBase five-year total-return CAGR without rerating: 52.9%.
Reference price USD 65.78; price date 2026-10-14. Return measure is terminal wealth CAGR, with no distributions and no reinvestment. The website applies its own disclosed comparison weights; this report assigns no scenario probabilities.

Why “without rerating” is optimistic here: holding the dated 15.12× EV / TTM adjusted EBITDA reference against the base terminal EBITDA produces $549.25 per share and 52.9% annualised. The base cash path and conversion rules stay unchanged. This exceeds the bull case because it removes the large multiple compression assumed in all three operating cases. It is a mechanical snapshot sensitivity, not a conservative case or a second fair-value estimate. The older balance sheet and growth-distorted trailing earnings further weaken that comparison.

Recent website quotes can reprice these fixed terminal outcomes. They do not refresh the financing bridge, cash forecasts, valuation date or investment conclusion. The site’s default risk-adjusted formula assigns a −100% certainty-equivalent return when a case has zero terminal wealth, so this report will rank at the bottom under that setting. That is a mathematical consequence of the total-loss case, not a prediction of a −100% expected return. A high score under another selected assumption set cannot resolve the evidence gaps in this report.

6. The base-case funding path

Management outlook from 11 AugustQ3 2026FY2026
Revenue$3.45–3.60bn$12.4–13.2bn
Adjusted operating income$200–260m$960–1,150m
Company-defined capex$11.5–13.5bn$35–39bn
Interest expense$860–940mNot given in this slide
Exit annualised run-rate revenue—$18.5–19.5bn
ARR is the final month’s revenue multiplied by 12, not guaranteed annual recurring revenue. Company capex is not automatically cash PP&E paid in the same period. [S5][S6]

Before projecting five years, the model explicitly bridges the stale June funding position to an indicative late 2026 starting point. It is not reported H2 or current debt. The assumptions are:$12.8bn full-year revenue less $4.653bn H1;60% H2 EBITDA margin; $21bn H2 cash equipment spend; $1.8bn cash-interest proxy; $0.1bn taxes; $2bn net customer-prepayment inflow; and hypothetical sale of all 35m ATM shares at the frozen $65.78 less 2% fees. Actual ATM execution is unknown. Other working-capital movements are zero.

Indicative opening bridge · USD bnCash effect / balance
June net debt including finance leases30.233
H2 EBITDA−4.888
H2 cash capex / interest / taxes+21.000 / +1.800 / +0.100
Net new customer prepayments−2.000
Hypothetical ATM net proceeds−2.256
September note discount and capped-call outlay+0.629
Indicative late 2026 net debt44.62
Economic opening shares651.371m: July basic+gross overhang+assumed 35m ATM
Borrowing proceeds fund the gap; they are not added a second time to net debt. The September convert is included within financing, not treated as free cash. Existing July awards are counted once; the later annual dilution rates cover new awards. No option / warrant exercise cash is credited.

The subsequent cash engine assumes no further equity-funded cash issuance and no recurring working-capital or prepayment benefit. Deficits increase net debt; surpluses reduce it. Interest is charged on average beginning / end gross financing debt at the case rate, holding $5.539bn unrestricted cash / securities constant. Principal maturities require refinancing; net debt alone does not prove that refinancing is available. This simplified bridge omits new acquisitions, collateral calls, FX movements and detailed facility covenants.

Base path · USD bn except shares20272028202920302031
Revenue26.0035.0044.0052.0060.00
Adjusted EBITDA15.6021.0025.9630.1634.20
Depreciation proxy10.9213.3015.4016.6418.00
Cash-interest proxy4.916.066.827.167.16
Cash tax0.000.410.941.592.26
Total cash equipment investment26.0026.2524.2023.4022.80
Cash after investment and financing-15.31-11.72-5.99-1.991.98
Closing net debt, pre-conversion59.9371.6577.6579.6377.65
Shares before convert settlement · m664.40677.69691.24705.06719.17
All forecasts. Cash tax ignores SBC deductions and tax-loss carryforwards; diluted-share growth captures new equity compensation. Figures rounded; calculation code uses full precision.

Base net debt peaks at about $79.63bn before conversion. At the end, cash after investment and financing is only $1.98bn, or about 1.9% of the model’s terminal common-equity value. This diagnostic divides year-five cash flow before terminal conversion by equity value after conversion; it is not a steady-state post-conversion cash yield. The model retains that cash for debt reduction; it is not also paid as a dividend. The low yield makes the $127.07 base endpoint demanding despite the 5× EBITDA multiple: a terminal buyer still needs growth and improving cash returns. That is why the model does not establish an ownership-ready target.

Base funding sensitivityTerminal net debt · bnShare valueFive-year CAGR
Base77.65$127.0714.1%
Capex +5 percentage points of revenue each year90.13$111.5111.1%
Funding rate +2 percentage points84.94$118.1212.4%
Both stresses97.93$101.399.0%
These change the annual funding path with the same revenue, margins and 5× exit multiple. Convertible settlement changes when terminal prices cross conversion strikes. These are mild stresses relative to a funding shutdown.

Maintenance versus growth capex is not separately disclosed well enough to estimate a reliable current owner-earnings yield. The model therefore uses total cash equipment spending. Our terminal 38% base capex / revenue ratio is an assumption about the combined replacement and expansion burden, not a sourced maintenance requirement. Fleet-level replacement and contract-cohort cash returns remain the most important missing operating evidence.

7. Back-solving the 20% hurdle

At $65.78,20% annual compounding for five model years requires $163.68 per share, with no cash dividends. Under the base financing and conversion assumptions, that requires $201.00bn terminal enterprise value. The mechanical ways to reach it are shown below; none is an independent forecast.

20% hurdle: one variable changesRequired outcome
Hold base EBITDA and funding5.88× terminal EV / EBITDA versus 5× base
Hold 5× multiple and base terminal debt / shares$40.20bn terminal EBITDA versus $34.20bn base
Hold 57% EBITDA margin,5× and debt / shares$70.53bn terminal revenue versus $60bn base
Hold base terminal share value$51.07 entry price for 20% annualised; conditional arithmetic, not a buy limit
The revenue backsolve holds debt and shares fixed to isolate the valuation requirement. Extra growth would generally require extra capex and financing, so this is a lower-complexity hurdle, not a self-funding growth plan. The entry-price backsolve also holds the original assumed ATM proceeds fixed; it does not rerun equity issuance at a lower sale price.

To merely preserve the entry price with the base terminal debt and shares, the business still needs substantial terminal enterprise value. That makes terminal earning power and the financing balance jointly important: asset growth is not sufficient if lenders and new shares take most of the economics.

Terminal EBITDA margin / EV multiple3×4×5×6×8×
47%$9.66$48.88$88.09$124.87$193.69
52%$22.18$65.56$108.31$146.84$222.98
57%$34.69$82.25$127.07$168.80$252.26
62%$47.21$98.87$145.37$190.76$281.55
USD per share at $60bn revenue, holding the base funding path fixed and resolving convertible settlement in each cell. This isolates terminal valuation; margin changes over the entire five-year path would also alter debt, taxes and cash flows. It is not a complete alternate operating forecast.

PEG is deliberately absent. Dividing a multiple by earnings growth does not make sense when starting GAAP earnings are negative, and it would ignore the capital needed to produce growth. An adjusted-EBITDA multiple is also incomplete; its usefulness depends on the cash and dilution bridge being visible. More complicated arithmetic cannot substitute for missing evidence on equipment economics.

8. What would confirm or break the thesis?

Evidence to watchConstructive outcomeThesis-breaking evidence
Next quarterly resultsRevenue and adjusted operating profit meet disclosed guidance while the cash / debt bridge reconciles.Revenue delays accompanied by rising capex, collateral requirements or financing cost.
Cash return on capacityTransparent cohort pricing, utilisation and replacement spending support returns after funding.Lower renewal prices or shorter equipment lives consume the residual cash margin.
Delivery and contracted powerCommissioned capacity becomes billed revenue on schedule.Power / installation delays while rent and interest continue.
Financing and dilutionActual ATM proceeds, new shares and facility draws are reconciled; maturities refinance on workable terms.Emergency equity issuance, stalled facility access or rising obligations without proportional earnings.
Customer performanceCollections and prepayments remain robust across major buyers.Payment stress, service credits, cancellations or concentration shifts weaken receipts.
Governance and capital allocationNew projects clear an after-financing return hurdle; acquisitions do not hide equipment economics.Growth targets take precedence over liquidity or return discipline.
Analyst monitoring framework. Trigger interpretation should follow current filings, not isolated social-media claims.

The next quarterly report is the nearest practical evidence window. A public vendor estimated 9 November, but no official date was verified, so it is not treated as a confirmed catalyst. The India project’s mid 2028 delivery target is a much later execution milestone. Rubin and Forge are already announced developments; a press release is not a future earnings surprise by itself.[S7][S17][S18][S19][S21]

CoreWeave entered the Nasdaq-100 in June 2026. Index membership adds market-flow exposure; it does not validate cash economics. Public positioning data can change quickly and does not support a verified borrow, options or forced-flow model in this report. The security is an absolute-return, financing-sensitive AI infrastructure research idea, with exposure to risk appetite, interest rates and customer AI spending. No portfolio position or hedge is assumed.[S20]

9. Conclusion and limitations

The strongest constructive case combines reliable deployment, durable customer demand, profitable reuse of equipment and cheaper funding. That can produce powerful equity upside from a depressed entry value. The strongest adverse case is a cash squeeze: spending and debt service arrive before delayed receipts, refinancing becomes expensive or unavailable, and new capital dilutes existing owners. CoreWeave can grow revenue through that process while common shareholders lose money.

Our base 14.1% and bull 39.8% annualised outcomes are conditional on financing continuing. The bear’s zero terminal equity is not a probability forecast, but it makes the asymmetric capital-structure risk explicit. No weighted expected return is presented because defensible probabilities have not been established. The optimistic 52.9% constant-multiple result should receive less underwriting weight, not more, because the denominator is trailing growth-period EBITDA and the snapshot net debt is stale.

Research assessmentConclusion
Evidence confidenceHigh for cited historical statements and executed financing terms; medium for public market / consensus snapshots; low-to-medium for forward cash / financing assumptions.
Underwriting statusWatchlist initiation. No completed positive ownership conclusion or verified fair-value target.
What could improve the viewCurrent capital stack, reconciled cash / accrual spending, realised cohort returns and a credible path to cash per diluted share.
What would worsen itCapacity delays, weak renewals, higher funding cost, customer stress or materially more dilution than the model.
Data cut-off15 October 2026 Sydney; regular-session price 14 October; actual financials throughJune; sharesJuly; later announcements separately identified.
Analyst judgment. The same scenario outputs can support different decisions under different risk tolerances; this report provides conditional research, not personalised advice.

Limits: the model is not a fully integrated three-statement forecast, project finance schedule or asset-level DCF. Its late 2026 opening bridge is hypothetical. It assumes funding access and refinancing, holds cash constant, simplifies tax and working capital, uses a blended funding rate and forecasts no new cash equity after the assumed ATM. Gross awards ignore exercise proceeds; convert treatment ignores capped-call benefits, adjusted conversion rates and issuer settlement choices. Future grants are separate assumed dilution. The annual periods are stylised scale proxies and the five-year CAGR is not an exact dated IRR. Australian investors also face FX, tax and fees outside these USD calculations.

Source differences are resolved in favour of filings for historical figures. For example, the public statistics page showed a different historical operating-income / EV convention; the report uses the audited 2025 operating loss of $46m and the filing-based $231m TTM loss, and explicitly excludes operating-lease PV from a rent-inclusive EBITDA multiple. September final note terms supersede earlier proposed / priced amounts. A newer quote can update the site’s ranking, but only new research can update this operating and funding thesis.[S1][S2][S7][S10]

Source register

All sources accessed 15 October 2026, Australia / Sydney. Filing financials cover the stated periods; company forecasts and examples are claims, not independently proven future outcomes. Historical filings remain historical. Market and consensus pages are timestamped snapshots.

S1 · FY2025 Form 10-K

Primary filing · year ended 31 December 2025; filed 2 March 2026. Statements, equipment lives, business risks and governance.

https://www.sec.gov/Archives/edgar/data/1769628/000176962826000104/crwv-20251231.htm
S2 · Q2 2026 Form 10-Q

Primary filing · six months ended 30 June 2026; August 2026. Statements and Notes 2, 8, 10, 12–16; July share count. Latest verified actuals.

https://www.sec.gov/Archives/edgar/data/1769628/000176962826000366/crwv-20260630.htm
S3 · Q2 2026 earnings release

Company release · 11 August 2026. Q2 / H1 figures, non-GAAP reconciliations, capacity and backlog definitions.

https://investors.coreweave.com/news/news-details/2026/CoreWeave-Reports-Strong-Second-Quarter-2026-Results/default.aspx
S4 · FY2025 earnings release

Company release · 26 February 2026. Historical adjusted EBITDA and financial comparison.

https://investors.coreweave.com/news/news-details/2026/CoreWeave-Reports-Strong-Fourth-Quarter-and-Fiscal-Year-2025-Results/
S5 · Q2 2026 outlook presentation

Company guidance · 11 August 2026, slide 3. FY2026 / Q3 outlook; definitions and non-GAAP limitations.

https://s205.q4cdn.com/133937190/files/doc_financials/2026/q2/CoreWeave-Q2-26-Earnings-Outlook-Presentation.pdf
S6 · Q2 2026 corrected call transcript

Company-hosted transcript · 11 August 2026. Management commentary, capacity plans, cash and deployment claims.

https://s205.q4cdn.com/133937190/files/doc_financials/2026/q2/CRWV-US-CORRECTED-TRANSCRIPT-CoreWeave-Q2-2026-Earnings-Call-11August2026.pdf
S7 · Public CRWV price and statistics

Market-data page · checked 15 October 2026, Sydney. Regular-session close $65.78 at 4 pm EDT, 14 October; later after-hours price excluded from frozen anchor.

https://stockanalysis.com/stocks/crwv/statistics/
S8 · Public analyst consensus

Stock Analysis / S&P Global · page updated 14 October 2026. Public aggregate estimates; broker models and revision timestamps not individually audited.

https://stockanalysis.com/stocks/crwv/forecast/
S9 · ATM offering prospectus

Primary SEC prospectus · 17 September 2026. Up to 35m shares; July 31 basic / award counts and fees. Capacity is not confirmed execution.

https://www.sec.gov/Archives/edgar/data/1769628/000162828026062362/coreweave-424b5.htm
S10 · September convertible closing 8-K

Primary filing · September 2026, closing 22 September. Final $4.2bn issuance, initial conversion terms and capped-call costs.

https://www.sec.gov/Archives/edgar/data/1769628/000176962826000432/crwv-20260917.htm
S11 · 2031 convertible closing 8-K

Primary filing · December 2025. $2.5875bn principal, initial conversion ratio and settlement terms.

https://www.sec.gov/Archives/edgar/data/1769628/000176962825000105/crwv-20251208.htm
S12 · August DDTL 5.5 financing

Company release · 10 August 2026. $2.6bn facility and contract / loan duration. Actual subsequent draws are in Q2 Note 16.

https://investors.coreweave.com/news/news-details/2026/CoreWeave-Closes-2-6-Billion-Loan-Facility-Expanding-Financing-Flexibility-for-AI-Infrastructure/default.aspx
S13 · NVIDIA strategic collaboration

Company release · 26 January 2026. $2bn equity investment and development collaboration; not an equity guarantee.

https://investors.coreweave.com/news/news-details/2026/NVIDIA-and-CoreWeave-Strengthen-Collaboration-to-Accelerate-Buildout-of-AI-Factories/default.aspx
S14 · NVIDIA residual-capacity order

Primary filing · September 2025. Initial $6.3bn order to April 2032, subject to delivery, availability and termination conditions.

https://www.sec.gov/Archives/edgar/data/1769628/000176962825000047/crwv-20250909.htm
S15 · September investor presentation

Public SEC exhibit · September 2026. Selected A100 / H200 renewal examples; preliminary financing figures superseded by S10.

https://www.sec.gov/Archives/edgar/data/1769628/000176962826000429/crwv2026convertpresentat.htm
S16 · Short-contract pricing update

Company release · 17 September 2026. Particular 3–6 month contracts, not realised fleet-wide pricing.

https://investors.coreweave.com/news/news-details/2026/CoreWeave-Continues-to-Contract-New-Compute-Capacity-at-Higher-Prices/default.aspx
S17 · Vera Rubin production deployment

Company release · 30 September 2026. Cognition production deployment; contract economics not quantified.

https://investors.coreweave.com/news/news-details/2026/CoreWeave-Delivers-NVIDIA-Vera-Rubin-NVL72-Performance-at-Production-Scale-Starting-With-Cognition/default.aspx
S18 · CoreWeave Forge launch

Company release · 30 September 2026. Integrated development workflow; no standalone software-margin inference.

https://investors.coreweave.com/news/news-details/2026/CoreWeave-Forge-Launches-Turning-the-AI-Loop-Production-Run-Into-a-Better-Model-and-Agent/default.aspx
S19 · India expansion with AdaniConneX

Company IR release · 7 October 2026. Planned 240MW campus; first phase expected mid 2028, further capacity phased. Not current capacity.

https://investors.coreweave.com/news/news-details/2026/CoreWeave-Enters-India-Expanding-AI-Cloud-Platform-With-AdaniConneX/default.aspx
S20 · Nasdaq-100 June quarterly changes

Exchange notice · 11 June 2026; inclusion effective 22 June. No assumed future index-flow benefit.

https://ir.nasdaq.com/node/110541/pdf
S21 · Investor relations events and releases

Company index · accessed 15 October 2026. No official next earnings date verified in this research.

https://investors.coreweave.com/overview/
R1 · Deutsche Bank — Fully Connected takeaways, 1 October 2026

User-supplied research, pp. 1–3; Buy / $150 dated view. Conference and management observations, not audited results. The two supplied copies are identical. Source PDF retained privately; not redistributed.

Supplied research note · title, date and pages identified above
R2 · J.P. Morgan — Fully Connected technology and demand, 30 September 2026

User-supplied research, pp. 1–2; Overweight. The $125 target appears in 24 September rating history on p. 4, not as a new target calculation. Source PDF retained privately; not redistributed.

Supplied research note · title, date and pages identified above
R3 · Barclays — Q2 2026 recap, 12 August 2026

User-supplied research, pp. 1–9; Equal Weight / $105 12-month target. Estimate exhibit uses the detailed income statement, p. 7. Cash-flow and target-bridge conflicts are disclosed in the report. Source PDF retained privately; not redistributed.

Supplied research note · title, date and pages identified above
R4 · UBS — Megaport: CoreWeave Q2 read-throughs, 12 August 2026

User-supplied research, p. 1. This is a Megaport report: its A$ target and company forecasts are not CRWV research inputs. The exit-run-rate typo is corrected using primary guidance. Source PDF retained privately; not redistributed.

Supplied research note · title, date and pages identified above