Goldman Sachs Upgrades CoreWeave to "Buy": Market Overestimates Financing Costs and Underestimates Computing Fee Capabilities

Goldman Sachs believes that concerns about CoreWeave’s funding payment methods are overstated, while its potential for compute fee generation is seriously underestimated.

According to Pioneer Trading Desk, Goldman Sachs analysts Gabriela Borges and Maura Hager on Monday upgraded CoreWeave's rating from “Neutral” to “Buy” with a price target of $106. Since the stock has fallen a total of 43% since its Q2 earnings release on August 11, including a separate drop of 24% over the past five trading days, closing at $61.39, the $106 price target still implies about 73% upside.

In the report, Goldman Sachs directly references the “Finance 101” framework, noting that CoreWeave’s enterprise value is highly correlated to the spread between its unlevered internal rate of return (uIRR) and weighted average cost of capital (WACC), arguing that the market is underestimating the extent of possible uIRR upgrades while overestimating the upward potential for WACC.

This upgrade has a direct impact on market sentiment.

In the past week, CoreWeave’s unsecured bond yields surged to 13.5%-15%, credit default swaps (CDS) hit record highs, and Deutsche Bank estimates there is a funding gap of about $35 billion over the next 12 to 18 months; these factors collectively weighed on the stock price. Goldman Sachs’ core counterpoint is: the unsecured bond market does not represent CoreWeave’s true cost of new funding, as this market has been closed to new CoreWeave issuance for months. CoreWeave’s main funding tool is asset-backed delayed draw term loans (DDTL) secured by GPUs and customer contracts.

Cost of Funding: Goldman Sachs Believes Market Got the Benchmark Wrong

The core controversy in the Goldman Sachs report is whether the metrics the market uses to gauge CoreWeave’s funding stress are accurate.

Goldman Sachs points out that CoreWeave's main funding is not traditional corporate bonds, but DDTL—a type of asset-backed project finance secured by specified customer contracts, GPU infrastructure, and related cash flows. Lenders underwrite not CoreWeave’s overall credit quality, but the cash flow quality of specific contracts and collateral value of underlying assets. This structure shifts funding risk away from CoreWeave’s corporate balance sheet to specific contracts and infrastructure, while reducing refinancing risk.

Based on this, Goldman Sachs estimates CoreWeave’s current weighted average cost of debt at about 800-900 basis points (using annualized Q2 interest expense of $2.56 billion divided by roughly $30 billion total debt). For new funding, Goldman Sachs assumes DDTL pricing for investment-grade customers is around SOFR + 300 bps (such as for financings tied to Meta contracts), and for high yield customers about SOFR + 600 bps, with an overall blended range of 750-1050 bps. Goldman ultimately uses 1000 bps as a conservative baseline, implying new funding costs rising just 100-200 bps above current levels.

Goldman Sachs also cites market data as evidence: DDTL 5.0 and 5.5 are both trading at or above the original issue discount, showing that the asset-backed funding market’s confidence in CoreWeave has not collapsed.

Fee-Generating Ability: Revenue Per Megawatt Is Key to the Bull Thesis

The other half of Goldman Sachs’ bull case is based on expectations for rising CoreWeave revenue per megawatt of compute power.

Goldman estimates CoreWeave’s current blended monetization rate is roughly $10 million in annual recurring revenue (ARR) per megawatt, but believes this number has upside potential over the next 24 months.

Supporting evidence includes: short-term contracts (3-6 months) are being signed at up to $40 million per megawatt; CoreWeave raised prices by 25% in July and another 10% in October; contracts signed in Q2 have profit margins 5-10 percentage points higher than recent ones. Factoring in the next generation of GPUs and higher-margin software products, Goldman Sachs views $13 million per megawatt as a “reasonable starting assumption” for newly signed long-term contracts.

Inputting $13 million/MW into Goldman’s example model (5-year term, 200 MW contract, 80% debt financing, 10% cost of debt, zero terminal value) yields a 15% unlevered IRR and a positive spread of roughly 3 percentage points over the 11.8% WACC.

However, Goldman’s sensitivity analysis also reveals the fragility of this logic: if the current actual blended level of $10 million/MW is used, uIRR drops to 6.3%, and the spread turns negative at minus 6 percentage points given similar funding costs. In other words, the entire bull thesis requires CoreWeave’s revenue per megawatt to increase about 30% from current levels.

Goldman’s supplementary argument here involves terminal value. The base case assumes GPUs are worth zero after five years, which is described as conservative, referencing A100 contract renewals through 2029 and H100 contracts being renewed at a premium. If a full sixth year’s terminal value is included, uIRR jumps to about 20%-25%.

Capital Expenditure and Free Cash Flow: Self-Sufficiency Not Expected Until 2030

Goldman also argues that CoreWeave's most intensive phase of capital expenditures has passed, and the proportion of new capacity additions relative to installed base has peaked, which will limit the financial drag from undeployed capacity.

But Goldman’s own model data reveals the long and challenging path ahead: free cash flow is projected at negative $28.2 billion in both 2026 and 2027, and negative $15.8 billion in 2028; long-term debt is expected to rise from $44 billion this year to $91 billion by 2028; expected interest coverage is only 0.3x in 2026, 0.7x in 2027; net debt/equity ratio peaks above 3000% in 2027. Goldman expects capital expenditures to stabilize in 2028, free cash flow to turn positive in 2030, and for CoreWeave to become financially self-sustaining at that point.

Goldman also presents a counterintuitive point: if the cost of capital rises across all compute service providers, supply growth will slow, delaying the onset of capacity oversupply; this means existing capacity could become “more valuable in the near term, not less”—in effect, credit tightening is described as a potential positive for CoreWeave.

Price Target and Valuation: Above Consensus, Discounted vs. Hyperscale Cloud Providers

Goldman Sachs set the price target at $106, based on an 18x Q5-Q8 EBIT multiple (previous target of $139 used 22x), stating this equates to roughly 10x EBIT in three years—about a 30% discount relative to Oracle, and more than 50% discount compared to hyperscale cloud providers, reflecting CoreWeave’s “higher absolute debt level.”

In terms of earnings projections, Goldman Sachs is notably above consensus: its 2028 revenue forecast is 11% higher than the average, 2028 EBIT is 24% above average, largely driven by higher per-megawatt ARR assumptions. Goldman Sachs also raised its 2028 EPS forecast to $3.36 (from a prior $1.15 loss), but cut its 2026 EPS forecast to a loss of $3.80 (previously $3.64 loss).

Important Footnote

Goldman Sachs acknowledges the main risks in the report: once supply catches up with demand, pricing could collapse, making fourth-to-sixth year contract renewals unable to cover WACC, while CoreWeave will remain locked into 7- to 16-year data center leases; hyperscale cloud providers such as Microsoft (estimated by Goldman to account for about 12%) could bring compute demand in-house; additionally, current spot pricing of $40-50 million per megawatt matches capital expenditure levels which mean a one-year payback, something Goldman admits is far above normal returns, and the market may doubt the sustainability of such ROIC.

It’s worth noting that Goldman Sachs disclosed in the report’s appendix that it is serving as lead or joint-lead underwriter on a pending CoreWeave underwriting mandate, holds over 1% of CoreWeave shares, has received investment banking fees from CoreWeave in the past 12 months, and expects to seek related business over the next three months.

In addition, according to previous reports, CoreWeave founders cashed out about $3.3 billion before the stock’s sharp decline.