The Zhitong Finance App learned that as the core manufacturer of AI computing power infrastructure, CoreWeave (CRWV.US) holds 100 billion on-hand orders and the business is expanding rapidly, but capital expenditure pressure and financing gaps have become the focus of the market's biggest concerns. Goldman Sachs released an in-depth research report on October 21, 2026, which maintains the purchase rating on the basis of dismantling the company's capital requirements. The target price for 12 months is 106 US dollars, and there is 71.5% upside compared to the stock price of 61.82 US dollars on the reporting date. At the same time, it warns that the implementation of financing is the primary investment risk.
As of August 2026, CoreWeave's current order reserves reached 129 billion US dollars, while the company's actual electricity put into operation in the second quarter of 2026 was only 1.5 GW, and the scale of electricity that has already been locked in for construction is as high as 4.2 GW, and the task of expanding production capacity in the future is heavy.
The company uses a DDTL special trust loan with a financing model for SPV special purpose entities. The capital expenditure can achieve up to 80% debt financing. The remaining 20% must be covered by own cash. 70% of the business contract comes with customer advance payments. The advance payment accounts for 15% to 25% of the total contract value range. The advance payment ratio is the core variable affecting the funding gap.
The research report sets up multiple sets of scenarios to calculate funds. It only serves the current $129 billion of on-hand orders. If GPU devices follow a 4.5-year usage cycle without customer advance payments, the company will need to add 37 billion US dollars in debt and 12 billion US dollars in own cash. The total internal usable cash is 10.4 billion US dollars. There is still a capital gap of 1.6 billion US dollars. The gap needs to be filled by customer prepayments, supplier credit, and existing business operating cash flow.
Under the optimistic assumption that 20% of the total contract value is taken into account customer advance payments and that the GPU usage period will be extended to 6 years, the minimum demand for new debt can be narrowed to 19.5 billion US dollars. If the entire 4.2 GW electricity production capacity is completed, the company can accept orders of about 263 billion US dollars, which is nearly double the current order, and corresponding additional debt of about 38 billion US dollars will be required.
At the financial level, CoreWeave is in a phase of rapid money-burning expansion. Revenue is forecast to be $12.96 billion, $27.53 billion, and $47.54 billion respectively from 2026 to 2028. Although the revenue growth rate remains high, the company continues to be in a state of loss. It is expected that profit correction will not be achieved until 2028, and earnings per share will reach $3.36 billion.
Dragged down by huge capital expenses of computing power facilities, free cash flow continued to be negative from 2025 to 2028, the size of debt on the balance sheet expanded rapidly, the level of leverage was high, and the business operation was maintained through continuous external financing. On the valuation side, with the release of scale effects, the EV/EBITDA ratio will drop 10.6 times in 2026 to 4 times in 2028.
In terms of stock price performance, as of the time of reporting, CRWV's stock price had retreated by more than 50% in the past 12 months, significantly outperforming the S&P 500 index. Goldman Sachs's logic for giving a target price is to value the company's Q5-Q8 standardized EBIT 18 times. At the same time, the research report clarifies a number of core risks: whether huge financing can be successfully implemented is the biggest risk; in addition, it also includes falling short of expectations in data center production capacity construction, mismatch between contract orders and construction cost inflation, and changes in AI computing power supply and demand, and declining prices for new contracts and contract renewals due to changes in AI computing power supply and demand.