Two major Wall Street institutions have offered answers that differ by more than threefold regarding the same company's funding gap, pushing AI compute leasing firm CoreWeave, Inc. (NASDAQ: CRWV) back into the center of market controversy.
Bernstein analyst Madison Rezaei published a report on Thursday, October 22, cutting the price target on CoreWeave, Inc. (NASDAQ: CRWV) from $74 to $51 and estimating that the company would need to raise $85 billion just to fulfill its existing contractual commitments, resulting in a net funding gap of $76 billion after deducting existing financing arrangements. This differs by more than three times from Goldman Sachs' estimate of $23 billion issued three days earlier.
Shares of CoreWeave, Inc. (NASDAQ: CRWV) fell another roughly 10% on Thursday to $55.70, erasing all gains recorded after Goldman Sachs upgraded its rating.
Goldman Sachs analyst Gabriela Borges upgraded CoreWeave, Inc. (NASDAQ: CRWV) to "Buy" on Monday and released a follow-up report on Wednesday, characterizing the funding gap as "approximately $23 billion." However, Bernstein's report quickly pointed out that Goldman Sachs' model contains multiple optimistic assumptions, and ironically, the incremental debt implied by Goldman Sachs' own cash flow model is actually closer to Bernstein's estimate.
Meanwhile, Goldman Sachs disclosed in its report that it is acting as manager or co-manager for a pending underwriting engagement for CoreWeave, Inc. (NASDAQ: CRWV), and holds more than 1% of the company's shares.
Bernstein: $85 Billion Gap, Losses Continuing Through 2029
Bernstein's report title is blunt and sharp: "Core-We've got bills to pay." Analyst Rezaei wrote that CoreWeave, Inc. (NASDAQ: CRWV) needs to raise $85 billion to fulfill its existing committed backlog, of which $9 billion is already covered by existing financing arrangements, leaving a net gap of $76 billion. This $76 billion represents the capital needed solely to deliver already-sold compute capacity without signing any new contracts.
Bernstein also dismissed the fallback option of "slowing growth and harvesting cash." The report explicitly states that even if CoreWeave, Inc. (NASDAQ: CRWV) stops pursuing new contracts, it would still need to fund deployment obligations embedded in existing contracts as well as maintenance and replacement needs. The "no new backlog" scenario would not immediately improve profitability.
On earnings forecasts, Bernstein significantly lowered its estimates: adjusted loss per share for 2026 widened from $4.56 to $5.14, and for 2027 from $4.77 to $6.81, with adjusted losses expected to persist at least through 2029.
On capital expenditures, Bernstein estimates 2027 growth capex at approximately $46 billion, plus about $6 billion in maintenance capex, totaling $52.5 billion, above market consensus. Even under the "no new backlog" scenario, 2027 capex would still reach $43.5 billion.
Goldman Sachs' $23 Billion: A Number Built on Three Key Assumptions
Goldman Sachs' $23 billion conclusion is built on a layered set of optimistic assumptions.
According to the analytical framework, Borges' estimates show that CoreWeave, Inc. (NASDAQ: CRWV) would require approximately $60 billion in capital expenditure to fulfill its $129 billion backlog, of which 80% could be debt-financed through DDTL (Delayed Draw Term Loan). After deducting $10.6 billion in committed but undrawn DDTL capacity and $10.4 billion in "cash options," the incremental debt requirement comes to approximately $36.7 billion. Further assuming customer prepayments of 20% of total contract value (approximately $15.7 billion), the net funding gap is ultimately calculated at $22.9 billion.
Bernstein challenged each of these points:
First, prepayment trends are moving in the opposite direction. Goldman Sachs itself acknowledged that CoreWeave, Inc.'s (NASDAQ: CRWV) customer prepayment ratio has declined over the past six months. Goldman Sachs' explanation is that when debt financing visibility decreases, customers become more willing to prepay. But Bernstein argues that when customers see CoreWeave, Inc.'s (NASDAQ: CRWV) bonds trading at yields above 15%, they may not be willing to prepay funds to a counterparty whose debt is at distressed levels.
Second, the "cash options" are overly optimistic. Goldman Sachs' $10.4 billion in "cash options" includes $2.2 billion from selling 35 million ATM shares at $64, while the stock currently trades about 9% below that level. Additionally, Goldman Sachs assumes access to a $2 billion revolving credit facility, but according to Bernstein's debt data, $1.2 billion of the $2.5 billion revolving credit has already been drawn, leaving an actual available balance of only about $1.3 billion.
Third, the largest scenario's numbers are deliberately downplayed. Goldman Sachs' Scenario 3 (covering 4.2GW of power capacity) assumes customer prepayments of $52.7 billion. Without such prepayments, the required incremental debt would reach $84.5 billion, with a debt-to-loan ratio of 94%, far exceeding Goldman Sachs' own 80% ceiling for the DDTL structure. This figure did not appear in the report's headline.
Fourth, the report avoids the cost of capital throughout. Goldman Sachs calculated how much CoreWeave, Inc. (NASDAQ: CRWV) needs to borrow but did not specify the borrowing rate, only citing the 10% incremental debt cost assumption from its earlier upgrade report. Meanwhile, CoreWeave, Inc.'s (NASDAQ: CRWV) unsecured bonds currently yield approximately 15%, meaning the former is about two-thirds of the latter.
Goldman Sachs' Model Contradicts Itself
The most damaging argument in Bernstein's report may come from Goldman Sachs' own financial model.
Turning to the cash flow statement on page two of Goldman Sachs' report, the bank forecasts that CoreWeave, Inc. (NASDAQ: CRWV) will consume $28.2 billion in free cash flow in 2026, another $28.2 billion in 2027, and $15.8 billion in 2028. To fund this, debt in Goldman Sachs' model increases by $26.7 billion this year, $26.6 billion in 2027, and $20 billion in 2028—totaling $46.6 billion in new debt over the next two years, double the "funding gap" in its report title. This does not even include lease liabilities that Goldman Sachs projects will reach $50 billion by 2028 (which are essentially also debt).
According to Goldman Sachs' own model, CoreWeave, Inc.'s (NASDAQ: CRWV) long-term debt will grow from $14.9 billion at the end of 2025 to $90.8 billion by 2028. The net debt-to-equity ratio will peak at 3,059% in 2027. Book cash will dwindle to just $1.2 billion by the end of 2027, while annual interest expense will reach $6 billion. The interest coverage ratio will be only 0.3x this year, 0.7x in 2027, and barely rise to 1.3x by 2028.
DDTL Maturity Pressure: The Bills Are Already Coming Due
Bernstein's report also specifically examined CoreWeave, Inc.'s (NASDAQ: CRWV) debt structure and near-term repayment pressure. As of now, CoreWeave, Inc. (NASDAQ: CRWV) has total outstanding debt of $42.1 billion with annual interest expense of $3.3 billion and a blended interest rate of 7.8%—a figure significantly lowered by $10.8 billion in convertible bonds carrying rates of only 2%-3%, with all other debt priced at junk-grade rates.
According to Bernstein's DDTL repayment calendar, monthly repayments on DDTL 5.0 have already begun as of approximately October 22, 2026 (the current point in time), with the first 1.35x Debt Service Coverage Ratio (DSCR) covenant test approaching around November 30. DDTL 5.5 will begin amortizing next January. The $8.5 billion DDTL 4.0 will begin amortizing next July.
On new financing costs, CoreWeave, Inc.'s (NASDAQ: CRWV) unsecured bonds yield approximately 15%, while the latest secured financing DDTL 5.5 (priced in August this year) carries a spread of SOFR+5.50%, the widest spread since the first DDTL in 2023. Bernstein believes subsequent financing could be priced even higher.
This forms the "dilemma" at the core of Bernstein's thesis: to remain competitive, CoreWeave, Inc. (NASDAQ: CRWV) needs to shift toward shorter-term contracts with enterprise customers. Short-term contracts carry higher unit pricing but pose greater risk to creditors and do not help reduce financing costs. The more strategically correct the choice, the higher the financing cost.
A 9% Business, a 32% Story
Bernstein conducted a return analysis on CoreWeave, Inc.'s (NASDAQ: CRWV) best contract—the agreement with Meta (including a $21 billion expansion)—arriving at an unlevered internal rate of return (IRR) of 9%-12% in the best case (9% in the base model). With leverage, the IRR jumps to 32%, but the composition of this figure is quite fragile:
CoreWeave, Inc. (NASDAQ: CRWV) invests $10.8 billion in capital expenditure, borrows $10 billion of it, contributes only $756 million in equity, and generates $129 million in levered free cash flow annually. The 32% return depends almost entirely on high leverage and a bet on the residual value of six-year-old GPUs in 2031. Meanwhile, the 9% unlevered return happens to match the cost of new secured financing (approximately 9%) and falls far short of the cost of unsecured financing (15%). HSBC previously reached the same conclusion with an 8.9% return on invested capital (ROIC).
Management has stated that contracts signed since the second quarter of 2024 have "significantly improved economic terms." Bernstein's response is that new contracts may be too small to meaningfully change the overall blended return profile, and its judgment leans toward medium-term economics being closer to older contracts.
Nvidia Is a Floor, Not a Cure
Bernstein did not predict that CoreWeave, Inc. (NASDAQ: CRWV) would go to zero. The report notes that the company still has a massive backlog and delivery visibility for powered-up data center capacity. Existing contracts, while not attractive enough, are not yet loss-making.
The real floor support comes from Nvidia. Bernstein believes that if CoreWeave, Inc. (NASDAQ: CRWV) faces a genuine downturn, Nvidia would step in, "not necessarily to revive the stock price, but to raise the floor." This is good news for CoreWeave, Inc. (NASDAQ: CRWV) shareholders, but somewhat awkward for Nvidia CEO Jensen Huang—who just last week publicly stated that CoreWeave, Inc.'s (NASDAQ: CRWV) cost of capital is "actually declining," when real-world data clearly does not support that claim.
Bernstein also believes that the potential shock from CoreWeave, Inc. (NASDAQ: CRWV) is "somewhat manageable"—if it disappeared tomorrow, its colocation contracts would be absorbed by the market, and the rest of the market would continue to function. But the report also notes that CoreWeave, Inc. (NASDAQ: CRWV) remains the core benchmark by which the market measures AI credit risk, as the tenant for 7 of the 20 project-financed data center bonds in the junk bond market. Emerging cloud computing companies that tried to follow its listing have also suffered successive setbacks.
Whose Math Is Right?
Combining both reports, how much more CoreWeave, Inc. (NASDAQ: CRWV) needs to borrow depends on whose spreadsheet you believe and your assumptions about customer behavior: Goldman Sachs' answer is $19.5 billion to $22.9 billion (including customer prepayments), or $36.7 billion to $84.5 billion without prepayments; Bernstein's answer is $76 billion.
Goldman Sachs stated at the end of its report that "workpapers are available upon request, and we welcome feedback on our assumptions." Bernstein's report may be exactly that feedback.
It is worth noting that Goldman Sachs also disclosed that it is acting as manager or co-manager for a pending underwriting engagement for CoreWeave, Inc. (NASDAQ: CRWV), holds more than 1% of the company's shares, and expects to seek or receive investment banking services compensation within the next three months. As Bernstein put it, the bills ultimately must be paid—and the first DDTL 5.0 repayment is due right now.
Comments