The contracts are signed. The capital is not.
CoreWeave has ~$104B of contracted backlog, revenue growing 112% and positive operating cash flow — yet the stock closed 1.4% above its 52-week low after a 36% slide from Sep 30. The market is pricing one question: can the money to energize that backlog be raised at a cost below what the contracts earn? Six analyst lenses, three scenarios, four horizons. At ~$75B of enterprise value the stock trades near 9.9× 2026E and 4.5× 2027E consensus adjusted EBITDA (derived).
Gray line = CoreWeave’s actual closes since its Mar 28, 2025 IPO, built only from dated anchor closes (Mar ’25 $37.08, Jun 20 ’25 $183.58, Oct 28 ’25 $139.93, Dec ’25 $71.61, Mar ’26 $77.47, Apr ’26 $111.60, Jun ’26 $99.54, Sep 30 $87.12, Oct 12 $80.26, Oct 14 $65.78, Oct 22 $55.60, Oct 23 $54.24); intra-period swings are not drawn. History and forecast panels use different px-per-month scales on a log price axis. Colored paths = synthesized scenario midpoints, probability-weighted (bear 30% · base 45% · bull 25%); open clay circle = Street 12-month consensus ≈$130 (range $39–$250). Scenario prices are illustrative frameworks, not forecasts.
Re-weight the scenarios
Those probabilities are a judgment call — so make them yours. Drag to set how likely the bear and bull cases are (base takes the remainder); the blended target below, the dotted line on the chart, and the prob-weighted row of the scenario cards all update live.
Six analyst lenses, six answers
Each lens is a synthesized expert perspective with its own 12-month target and the math behind it. Panel mean ≈ $74 vs. a Street consensus near $130 — the desk sits well below the sell-side.
The Capacity Compounder
Backlog of $104B (+246%) plus >$25B of early-Q3 adds, 2026 revenue guided to $12.4–13.2B and consensus 2027 revenue near $26.5B. Prices were raised 25% in July and 10% in October. Operating leverage is appearing: adjusted operating income $128M vs $21M in Q1.
The Credit Underwriter
TTM free cash flow is −$13.7B and net debt is ~$46B; there is no cash floor to underwrite. The equity is a levered option on contract conversion, so only a price that already discounts a 15% cost of capital is attractive.
The Funding Gap
Bernstein-style math: ~$85B to fund the committed backlog, ~$9B covered, ~$76B outside. Interest is $640M a quarter and rising; short interest of 23.7% shows the debate is crowded but the credit tape agrees with the bears.
The Contract Book
Take-or-pay contracts with Microsoft, Meta, OpenAI and Anthropic are real moat-like visibility. But HSBC estimates ~$12M revenue per MW in 2027 vs $17–18M at hyperscalers, so the neocloud earns less per unit of power.
The Falling Knife
Beta near 3.5, a new 52-week low on Oct 22 on 2x normal volume, price below its 200-day average and a 36% fall from Sep 30. Trend is down; a base needs the Q3 print (~Nov 11) and a calmer bond tape.
The Cost of Money
Rates, not demand, are the repricing: Nvidia calls neocloud demand “off the charts” yet ~15% unsecured yields mean credit is the binding constraint. Goldman argues borrowing costs are overestimated; a calmer tape could re-rate the stub quickly.
Wall Street 12-month price targets
Selected sell-side targets, sorted low to high. The dashed line is today’s $54.24; the Street is split between a $39 Reduce and a $250 Street-high.
Targets as reported Oct 19–23, 2026: HSBC Reduce $39; Bernstein Underperform $51 (cut from $74 on Oct 22); MoffettNathanson Neutral $89; Goldman Sachs upgraded to Buy, $106 (from $139) on Oct 19; BNP Paribas Outperform $117; Evercore ISI Outperform $150; Rosenblatt Buy $250. Consensus sources differ: $129.90 (Benzinga), $134.84 (TipRanks), $136.21 (Yahoo), $140.58 (S&P Global, older). Firm ratings are the real sell-side calls; the analyst lenses above are not.
Where the road leads
Synthesized scenario midpoints (mid-year), returns vs. today’s $54.24. Each price = NTM adjusted EBITDA × EV/EBITDA − net debt, divided by diluted shares. Illustrative frameworks, not predictions; a highly levered equity moves far more than the business does.
1 Year
Mid-20272 Years
Mid-20283 Years
Mid-20295 Years
Mid-2031▸ Bull case — show the assumptions & math
▸ Base case — show the assumptions & math
▸ Bear case — show the assumptions & math
Revenue, capex, free cash flow & net debt ($B)
Where the money actually goes. The gap between the sky bar and the clay bar — and the olive bar below zero — is the whole funding debate.
Cash-flow capex ($2.9B → $8.7B → $10.3B → $20.6B TTM) outruns operating cash flow each year, so free cash flow (olive, drawn below zero) widens to −$13.7B TTM; net debt including leases (slate) reaches ~$46B. Management’s Q2 capex of $9.4B is accrual-based vs. $6.42B in the cash-flow statement, and the 2026 guide is $35–39B. Sources: StockAnalysis / company filings; TTM revenue $7.59B.
Energized vs. funded — the funding-gap gauge
Backlog is only revenue once capacity is built. Bars are scaled to Bernstein’s ~$85B need; estimates use different assumptions and are not like-for-like.
Adjusted EBITDA path underpinning the targets ($B)
CoreWeave has no positive GAAP EPS to capitalize, so the targets are built on this adjusted EBITDA ladder times an exit multiple, less net debt.
Gray = reported ($1.22B 2024, $3.09B 2025). Olive = consensus ($7.57B 2026E, $16.63B 2027E, $28.68B 2028E per MarketScreener; Motley Fool cites $7.6B and $16.7B). Faded olive = desk extension for 2029–31E, consistent with the base case. Consensus EPS is negative through 2027 (about −$4.33 for 2027, down from −$3.37 90 days earlier), so an EPS ladder would not underpin a price.
Demand is growing — so are the costs
Q2 FY26, year-over-year. Read these against a stock at its 52-week low: the top line is intact, but interest and capex grow as fast or faster.
Adj. EBITDA $1.51B vs $753M; revenue $2.575B vs $1.212B; net interest $640M vs $267M; Q3 revenue guide $3.45–3.6B vs $1.365B (+158% at midpoint); cash capex $6.42B (+162% YoY); backlog $104B (+246%), excluding >$25B added in early Q3. Costs (terracotta) compounding with revenue is the bear’s point; backlog and guidance (clay) are the bull’s.
Bull vs. Bear
The valuation argument compresses into one disagreement: is the backlog a financeable annuity, or a promise that must be funded at distress-level rates?
▲ THE BULL CASE
- Backlog keeps growing. $104B at June 30 (+246%) with >$25B more committed in early Q3; Q3 revenue guided to $3.45–3.6B.
- Pricing power. Compute prices raised 25% in July and 10% in October, per Goldman; Jensen Huang calls neocloud demand “off the charts.”
- Operating leverage appears. Adj. operating income $128M vs $21M in Q1; operating cash flow +$679M vs −$251M a year ago.
- Credit still open. Q2 interest landed at the low end of guidance; unsecured debt can still be placed, and NVIDIA invested $2B.
- Cheap on contracts. ~$75B EV is ~4.5× 2027E consensus adjusted EBITDA and ~2.9× 2027E sales.
- Crowded short. Short interest is 23.7% of float; a financing headline can reverse the tape fast.
▼ THE BEAR CASE
- Funding gap. Bernstein: ~$85B needed, ~$76B outside capital; Goldman’s lower $23B is still large against a ~$30B market cap.
- Cost of capital. The 2032 notes yield ~14.8% vs a 9.625% coupon; HSBC sees 2026–30E returns of 8.9%, below borrowing cost.
- Cash burn. TTM free cash flow −$13.7B, FY26 capex $35–39B vs revenue $12.4–13.2B, net debt ~$46B.
- Concentration. Microsoft was ~67% of FY25 revenue; backlog growth ex-OpenAI was 25%.
- Structural disadvantage. HSBC: ~$12M revenue per MW in 2027 vs $17–18M for hyperscalers.
- Losses and dilution. Q2 net loss $626M; consensus 2027 EPS slid to about −$4.33; equity may fund part of the gap.
Risk map — likelihood × impact
Over a 3–5 year horizon. The hot corner — likely and high-impact — is financing cost; the serious tail risks sit one row down.
- Depreciation drag
- Equity dilution
- Concentration
- Financing cost
- Power delays
- GPU obsolescence
- Credit shutoff
- Counterparty default
- Demand air-pocket
Financing cost
Unsecured paper trades near 15% against a 9.625% coupon; HSBC puts 2026–30E returns (8.9%) below the cost of borrowing — growth stops creating value.
Credit shutoff
The refinancing window closes before the $35–39B 2026 capex plan is funded; builds stall and backlog converts late.
Counterparty default
A large AI-lab customer cannot pay: Microsoft was ~67% of FY25 revenue and OpenAI is a major backlog name.
Equity dilution
Bernstein sees ~$76B of outside capital needed; any equity share of it dilutes holders at depressed prices.
Concentration
A handful of hyperscaler and AI-lab contracts drive the backlog; one renegotiation moves the whole model.
Demand air-pocket
LLM revenue growth moderates, GPU rental prices roll over and take-or-pay terms get tested — low odds, but the stub equity reprices fast.
Power delays
Active power lags contracted power; late energization delays revenue while interest keeps accruing.
GPU obsolescence
Faster NVIDIA cycles shorten useful life, raising depreciation and lowering residual value versus five-year contracts.
Depreciation drag
GAAP losses persist (Q2 net loss $626M) even as adjusted EBITDA margins run near 59%.
The jargon, decoded
Hover the dotted terms in the metrics, or scan the desk’s working definitions here.
- Revenue backlog
- Contracted future revenue not yet recognized. It measures demand, not cash in hand; it converts only as capacity is delivered.
- Take-or-pay
- Contracts where the customer pays for reserved capacity whether or not it uses it — the basis for lenders financing CoreWeave builds.
- Adjusted EBITDA
- Earnings before interest, tax, depreciation, stock comp and one-offs. For CoreWeave it is ~59% of revenue but excludes the GPU depreciation and interest that drive GAAP losses.
- Net debt (incl. leases)
- Debt plus lease obligations minus cash — ~$46B at June 30, 2026 (StockAnalysis). It is the subtraction between enterprise value and equity value.
- Funding gap
- Outside capital needed to build what is already contracted. Bernstein: ~$76B after ~$9B of facilities; Goldman: ~$19.5–23B including prepayments.
- EV / EBITDA
- Enterprise value divided by EBITDA. Targets here are NTM EBITDA × an exit multiple − net debt, divided by shares — the stub equity is highly sensitive.
- Neocloud
- A specialist GPU-rental cloud (CoreWeave, Nebius, IREN) competing with hyperscalers on speed to capacity rather than breadth of services.
- Prob-weighted
- Each scenario price times its probability, summed into one expected value across bear, base and bull.