CoreWeave Doubles Revenue, Reports $104 Billion Backlog as Enterprises Sign On

CoreWeave more than doubled its revenue in the second quarter and reported a revenue backlog of about $104 billion, the AI cloud provider said on August 11. The company also named a new set of enterprise customers, including Caterpillar and Bentley Systems. For CIOs weighing where to run AI workloads, the quarter shows specialized GPU clouds moving beyond AI labs into mainstream enterprise buying, along with the financial strain behind that growth.

What did CoreWeave report?

Revenue reached $2.58 billion, up from $1.21 billion a year earlier. The backlog figure, measured at June 30, excludes more than $25 billion in net new customer commitments CoreWeave said it signed early in the third quarter. Active power rose by nearly 500 megawatts to 1.5 gigawatts, and contracted power reached about 3.7 gigawatts.

Profitability is another story. CoreWeave reported a net loss of $626 million, compared with $290 million a year earlier, and net interest expense of $640 million, up from $267 million. Adjusted EBITDA was $1.51 billion, a 59% margin. The company spent $6.4 billion in cash on property and equipment in the quarter.

On the earnings call, CoreWeave raised its full-year revenue outlook to $12.4 billion to $13.2 billion and set capital spending at $35 billion to $39 billion, according to call summaries from American Companies and INDmoney. INDmoney reported that shares rose about 15% in after-hours trading on August 11.

Which enterprises are buying from CoreWeave?

The customer list is the most telling part for enterprise readers. CoreWeave's release named Bentley Systems, Caterpillar, Grammarly, Isomorphic Labs and Sunday Robotics as new customers, and said it expanded work with Cognition, Databricks, Hudson River Trading, Periodic Labs, Rescale and Runway. American Companies' call summary added Flow Traders, IMC and Leidos across financial services and government.

CEO Michael Intrator said demand is rising "as enterprise adoption broadens," as quoted by Pulse 2.0. He also described near-term capacity as effectively sold out, according to American Companies.

CoreWeave is also moving up the stack. Its managed inference business grew from about $1 million to more than $100 million in annualized recurring revenue within months of launch, according to both call summaries, with a goal of at least $250 million by year-end. For enterprises, that matters more than raw GPU rental. Most companies do not want to manage clusters. They want to run models reliably at a known cost.

What are the risks for enterprise customers?

The same numbers that impress investors should prompt questions from buyers. CoreWeave is funding its buildout with heavy borrowing. In the quarter it raised a $3.1 billion term loan, took a $1 billion strategic investment from Jane Street, and issued more than $10 billion in unsecured debt and convertible bonds. INDmoney's analysis said the company's two largest customers accounted for about 65% of first-quarter revenue, and that cash economics remain unproven.

None of that means enterprises should avoid neoclouds. It means they should treat them like any strategic supplier with a concentrated, leveraged business. AI Connect speakers have raised this risk before. In a AI Connect panel on future-proofing AI infrastructure, SandboxAQ VP of Engineering Stefan Leichenauer cautioned that a chosen vendor "may not even exist two years from now," and advised open frameworks and testing several vendors.

Cost discipline matters too. At COLLIDE 2024, Varun Tottempudi, Head of Data & Platforms Engineering for Global Analytics at Lockton, warned that without monitoring and alerts, "in six months you might see a bill" as large as the past five years' spending, a point covered in our piece on AI ROI across the project life cycle.

What should CIOs do with this news?

Check capacity commitments. If near-term capacity is sold out, enterprises that need GPUs for inference or fine-tuning in the next year should confirm delivery dates in writing, with remedies if capacity slips.

Keep workloads portable. Containerized inference, open model formats and standard orchestration let teams move between hyperscalers, neoclouds and on-premises systems if pricing or reliability changes.

Price in the full stack. GPU rates are only part of the cost. Data egress, storage and the engineering time to move AI into production at scale often decide whether a project pays off.

CoreWeave's third-quarter guidance calls for revenue of $3.45 billion to $3.6 billion and interest expense of $860 million to $940 million, according to the call summaries. Whether revenue growth can outpace that interest bill will shape how comfortable enterprises feel building on specialized AI clouds.

Frequently asked questions

What is CoreWeave's revenue backlog?

CoreWeave reported a revenue backlog of about $104 billion as of June 30, 2026, excluding more than $25 billion in new commitments signed early in the third quarter.

Is CoreWeave profitable?

Not on a net basis. It reported a $626 million net loss for the second quarter of 2026, largely reflecting heavy interest and depreciation costs, though adjusted EBITDA was $1.51 billion.

Should enterprises use neoclouds like CoreWeave for AI?

They can be a useful source of GPU and inference capacity, but buyers should treat them as strategic suppliers: confirm capacity in writing, keep workloads portable and monitor costs closely.

Filed under: Data & AI Infrastructure, AI Strategy & ROI