CoreWeave Revenue More Than Doubles to $2.58 Billion as Backlog Reaches $104.2 Billion
Key Takeaways
- •CoreWeave's Q2 revenue of $2.58 billion more than doubled from the prior-year period and modestly exceeded the $2.56 billion analyst consensus.
- •The company's contracted but unbilled revenue backlog surged 246% year-over-year to $104.2 billion, excluding approximately $25 billion in additional commitments secured in early Q3.
- •CoreWeave reported a $626 million net loss for the quarter, primarily driven by $640 million in net interest expenses tied to debt financing for its GPU fleet and data center expansion.
- •CEO Michael Intrator emphasized AI inference as a larger market opportunity than model training and a key strategy to maximize returns on the company's $46.7 billion in property and equipment.
- •Nvidia, which holds close to a 13% stake in CoreWeave, announced a separate $500 billion AI infrastructure financing initiative with major asset managers one day before CoreWeave's earnings report.

CoreWeave, a specialized neocloud provider delivering AI infrastructure services, reported second-quarter revenue of $2.58 billion, more than double the prior-year period and slightly ahead of the $2.56 billion analysts had projected.
"Our near-term capacity remains effectively sold out," CEO Michael Intrator told analysts on a Tuesday earnings call.
The company's revenue backlog—contracts signed but not yet billed—surged 246% year-over-year to $104.2 billion in the second quarter, a figure roughly ten times CoreWeave's current annual revenue pace. CoreWeave also noted that this figure does not include approximately $25 billion in net new customer commitments secured in early Q3.
Following the earnings release, the stock jumped more than 14% in after-hours trading on Tuesday. The gain came after a volatile stretch during which share prices fluctuated as investors weighed concerns about a potential AI bubble against the enormous capital expenditures flowing into AI infrastructure. CoreWeave, which went public earlier in 2025, has seen its shares trade in a wide range since its market debut.
CoreWeave is part of a growing segment of neocloud providers that construct and operate data centers purpose-built for running AI models, competing with established cloud leaders Amazon Web Services, Microsoft Azure, and Google Cloud. Unlike those hyperscalers, which offer a broad portfolio of cloud services, neocloud operators like CoreWeave focus almost exclusively on GPU-intensive computing—raising substantial debt to finance large-scale GPU acquisitions and then renting that capacity to AI developers.
The debate around potential overinvestment in AI gained fresh intensity one day before CoreWeave's earnings. On Monday, Nvidia—a company that holds close to a 13% stake in CoreWeave—announced plans to mobilize $500 billion in financing for AI infrastructure in partnership with Apollo Global Management Inc., Blackstone Inc., BlackRock Inc., and Brookfield Asset Management. Nvidia shares rose approximately 1% following that announcement and edged higher again after CoreWeave published its results.
During the earnings call, Intrator said the prices CoreWeave charges for access to Nvidia's cutting-edge Blackwell and Vera Rubin chips are "setting new highs," while even older-generation inventory is renting at price levels last seen years ago.
Despite robust demand, CoreWeave posted a net loss of $626 million in the second quarter, driven largely by a $640 million net interest expense tied to the heavy debt financing underpinning its GPU fleet and data center buildout. That compared with a net loss of $260 million in the same quarter a year earlier. Excluding stock-based compensation and other items, the company reported $128 million in adjusted operating income.
Investor caution around CoreWeave has partly stemmed from concerns about the useful life of older inventory, particularly the high-value Nvidia GPUs housed in its data centers. The company carries $46.7 billion in property and equipment on its balance sheet, primarily consisting of GPUs that some fear could depreciate rapidly as newer models enter the market.
Company executives highlighted AI "inference"—in which CoreWeave's computing infrastructure runs AI models rather than training them—as a strategy to extract maximum value from its chip and infrastructure investments. Inference represents a distinct and potentially larger market segment than model training, as deploying AI applications in commercial use requires ongoing computational capacity rather than one-time training runs.
"The market is very deep," Intrator said. "We think that we have an embedded advantage because of our control over the silicon, and we think that we're going to be very successful in that market over time."
Intrator also sought to allay concerns that local opposition and moratoriums on data center construction could impede CoreWeave's growth. Currently, 18 states have either enacted or are weighing significant restrictions on data center building. Intrator acknowledged this will influence where facilities are sited but said it will not diminish overall demand. He advised companies to collaborate with local communities and offer incentives such as funding grid upgrades to shield residents from costs and foster long-term job creation.
CoreWeave has set a target of reaching eight gigawatts of power capacity by 2030.
"None of those numbers will be impacted by the regulatory pushback; as of today we are comfortable with it," Intrator said.