Our call: avoid CRWV. The selloff is exposing a growth plan that still needs tens of billions from capital markets. A cheaper share price does not make the next data centre cheaper to finance.
CoreWeave can have extraordinary AI demand and still deliver a miserable outcome for shareholders. The lenders, the equipment and the buildings must be paid before the equity enjoys the upside.
01 / Follow the cash
The $36bn cheque still has to clear.
Our reconciliation points to $27–36bn of additional consolidated financing through December 2027, after crediting the liquidity and subsequent financing we can identify. The Bloomberg case lands at $36.2bn. [1]
Starting from June 30, the arithmetic is brutal: $47.4bn of forecast cash burn, plus $10.6bn of scheduled principal, against $21.7bn of identified resources. Those resources already include June’s cash and unused facilities, August’s new commitment, and September’s convertible proceeds after the capped-call cost. [2]
This gives CoreWeave substantial credit: identified facilities are assumed usable and there is no retained cash buffer. Amortization on subsequent borrowing is excluded. A facility with draw conditions and an expiry date still has to become cash.
02 / Even the friendlier numbers need funding
A Buy rating doesn’t write the cheque.
Run the same bridge with the other published forecasts and the requirement is approximately $27.0bn using BofA and $31.4bn using Visible Alpha. These are our calculations from their reported cash-flow forecasts. [3]
Financing the next phase is an assumption inside the bullish case. A huge order book helps attract lenders. It does not establish that every required dollar is committed.
03 / The equity is the shock absorber
The SPV story offers less shelter than advertised.
At June 30, approximately 90% of gross debt was classified as recourse, including corporate debt and project borrowing with recourse. Project companies organize collateral and cash flows; they do not universally remove parent exposure. [2]
The buildings add another dependency. Morgan Stanley’s selected landlord-bond tracker identifies $14.1bn linked to CoreWeave tenancy—about 31% of that universe. These are landlords’ bonds, not extra CoreWeave principal. But a landlord’s funding problem can delay the powered halls CoreWeave needs to deploy equipment and earn revenue. [4]
If the next lender says no,
the shareholder pays.
Our bear case is that financing pressure forces expensive capital, dilution or delayed deployment. Slower deployment then postpones the revenue needed to support the debt. The growth engine becomes harder to finance precisely when it needs more cash.
04 / The call
Don’t buy the financing risk
just because the stock fell.
We would reconsider after closed financings and an updated cash bridge demonstrate sufficient liquidity after debt service, required project contributions and reserves. The price and terms of that capital must leave a credible return for existing shareholders.
Until then: CRWV—don’t buy that dip.
How the $36.2bn estimate is calculated
Horizon: 1 July 2026 to 31 December 2027. US$bn. Bloomberg consensus pulled on 15 October 2026.
| 2026 forecast cash burn, less $10.454bn incurred in H1 | 16.743 |
|---|---|
| 2027 forecast cash burn | 30.614 |
| Principal on June debt: H2 2026 + 2027 | 10.597 |
| Total uses | 57.955 |
| June cash, securities and undrawn facilities | −15.553 |
| August new facility commitment | −2.600 |
| September net proceeds after capped calls | −3.571 |
| Additional consolidated financing | 36.231 |
Cash FCF is operating cash flow less cash equipment purchases. Bloomberg’s estimate-comparable historical FCF matches this definition in 2025 and Q1–Q2 2026. Interest, operating rent and customer advances are already reflected in operating cash flow. Principal is added separately.
August draws are not added again to full commitments. The bridge assumes historical facilities were or can be drawn within their availability windows. Unused expired commitments would reduce resources. No unconfirmed ATM proceeds, retained cash target, other investment uses, additional reserve funding or subsequent-loan amortization is included. Broker forecast conventions and actual spending or collections can change the result. The funding requirement includes refinancing and is not an equity-issuance forecast.
The June balance-sheet date is used because a complete updated October cash and undrawn-capacity reconciliation is not public. These estimates do not establish a current payment default or a fixed parent-only funding gap.
Inspect the calculation data