高盛上调CoreWeave评级至买入
高盛上调CoreWeave评级至买入
Conclusion
Goldman Sachs’ move to Buy on CoreWeave is a bullish call on the economics of its AI-computing assets—not a finding that its balance sheet is low-risk. Goldman upgraded the stock from Neutral on October 19, 2026, while cutting its price target from $139 to $106. Its thesis is that tight AI-capacity supply, execution and performance advantages, and higher revenue per megawatt could more than offset moderately higher borrowing costs. The counterweight is substantial: rapid debt growth, large lease obligations, heavy investment spending, customer concentration, and GAAP losses.
Data as of October 19, 2026. The latest quarterly financials reviewed are for the quarter ended June 30, 2026; the later financing update is from September 22.
The bullish case
-
Demand and scale are growing quickly. Q2 revenue was , versus a year earlier—growth of . CoreWeave reported of active power and approximately of contracted power. The release also noted more than in new customer commitments early in Q3, not included in its June backlog figure.
-
The forward demand pool is large, though not all of it is firm near-term revenue. The SEC filing reported of remaining performance obligations (RPO) at June 30; was expected to be recognized in the first , with the rest spread over later periods. The company’s broader “revenue backlog” measure also includes estimated amounts beyond RPO and is subject to delivery and service-availability conditions.
-
Adjusted operating performance is expanding in dollars, but not yet in margin. Q2 adjusted EBITDA was , compared with , a increase. However, adjusted EBITDA margin fell from to —a -percentage-point decline. Goldman’s case depends on capacity additions translating into stronger returns per megawatt, rather than growth alone.
-
Goldman’s quality argument is about execution and moving up the stack. CNBC’s account of Goldman analyst Gabriela Borges’ note says the firm sees CoreWeave as relatively well positioned among “neocloud” providers because it can deploy and scale high-performance clusters and has scope to build software and platform advantages beyond GPU-as-a-service. Goldman expects revenue per megawatt and unlevered returns to improve over the next two years, and argues that asset-backed project financing may make market prices on unsecured debt a poor guide to CoreWeave’s marginal funding cost. These are analyst expectations, not reported outcomes. CNBC, October 19; Investing.com, October 19.
What the financials and filings put at risk
-
The business is not yet producing positive GAAP earnings, and interest is a major drag. Q2 had a GAAP operating loss of , net interest expense of , and a net loss of . Adjusted EBITDA is useful for assessing operations, but it does not remove the cash and refinancing burden of interest.
-
Investment spending far exceeds internally generated cash. In the first half, operating cash flow was , while purchases of property and equipment, including capitalized software, were . That implies calculated free cash flow of billion. The 10-Q also shows debt issuance and equity proceeds helping fund the investment cycle. This makes continued access to financing—and the returns earned on each new tranche of infrastructure—central to the case.
-
Funded debt grew sharply, and leases add another large fixed obligation. At June 30, calculated gross funded debt was billion, compared with billion at December 31, 2025, an increase of . After cash and marketable securities, and excluding restricted cash, calculated net funded debt was billion. Operating and finance lease liabilities were another billion; funded debt plus those lease liabilities totaled billion. These are distinct obligations, not all conventional borrowings. The balance sheet separately reports substantial recourse debt as well as non-recourse debt, so the asset-backed-financing argument should not be mistaken for all borrowings being isolated from the parent.
-
Customer concentration is material. Two anonymized customers represented and of Q2 revenue— combined. The same filing says those two customers each accounted for of accounts receivable. Large contracts support visibility, but customer dependence makes renewals, utilization and payment behavior unusually consequential.
-
Backlog has execution and measurement risk. The filing explains that RPO is net of estimated variable consideration, including possible service credits, delivery delays and capacity the company has the right to resell. A large backlog is therefore not equivalent to cash on hand or guaranteed revenue at the stated timing. CoreWeave must bring power and data-center capacity online, install and operate equipment, and meet customer delivery terms.
-
Some commitments sit outside the consolidated balance sheet. The 10-Q discloses up to of funding commitments for two data-center joint ventures, expected to be satisfied during 2026, with further contributions required under the JV agreements. That adds to the need to track capital commitments alongside reported debt and capex.
What changed after the June balance sheet
On September 22, CoreWeave completed a offering of senior unsecured convertible notes bearing interest and maturing in 2033. This is evidence of access to long-dated, lower-coupon funding, but it was issued after June 30 and is not reflected in the June balance-sheet totals above. The notes are convertible; capped calls are intended to reduce dilution only up to their contractual cap, not eliminate the possibility. The same filing identifies existing senior notes due 2030–2032 with stated coupons as high as , underscoring why the overall mix and cost of capital matter. SEC Form 8-K, September 22.
How to assess the balance
Treat this as an execution-and-financing thesis, not a simple “AI demand is strong” story. The key test is whether realized revenue and cash contribution per megawatt rise enough to cover equipment depreciation, leases, interest and the cost of adding capacity. Favorable evidence would include backlog converting on schedule, improving EBITDA margins, capex growing more slowly than operating cash flow, stable customer concentration, and new capacity earning attractive returns without worsening financing terms.
The bear case is a timing mismatch: large obligations and upfront infrastructure spending arrive before enough customer revenue is delivered, or financing costs rise while asset utilization, pricing or residual equipment values disappoint. Goldman’s upgrade is a credible argument that the market may be discounting those risks too heavily; the target cut, negative free cash flow and rising obligations show that the risks remain unresolved. The balance is bullish on operating potential, but financially fragile and highly dependent on execution.