CoreWeave’s Q2 2025 revenue surged 207% year-over-year to $1.21 billion, driven by rapid adoption of its AI-optimized cloud infrastructure. The company closed the quarter with a record $30.1 billion revenue backlog, up 86% from a year ago, reflecting major contract wins including a $4 billion OpenAI expansion, in addition to an $11.9 billion deal signed earlier this year. A newly signed hyperscaler customer also expanded commitments in the quarter. Adjusted EBITDA reached $753 million, up 3x from Q2 2024, with a 62% margin, while adjusted operating income rose to $200 million. GAAP net loss narrowed to $291 million from $323 million last year, with interest expense at $267 million.
Capital expenditures climbed to an all-time high of $2.9 billion, supporting aggressive data center and GPU infrastructure expansion. CoreWeave ended Q2 with approximately 470 MW of active power and 2.2 GW of contracted power across 33 data centers. The company is also developing a new greenfield campus in Kenilworth, New Jersey, with up to 250 MW capacity in its first purpose-built AI data center project, scheduled for initial delivery in 2026. Year-to-date, capital spending has reached $4.8 billion as CoreWeave ramps to meet long-term AI infrastructure demand.
Technology milestones included becoming the first provider to offer NVIDIA’s complete Blackwell GPU portfolio at scale and delivering the largest-ever MLPerf Training v5.0 submission—34x larger than other entries and 4.5x more performant than the next-best GB200 result. The acquisition of Weights & Biases adds model lifecycle tools such as cluster health monitoring, real-time inference access, and online evaluation capabilities to CoreWeave’s platform. Customers announced this quarter included BT Group, Cohere, Hippocratic AI, Hologen, LG CNS, Mistral, Moonvalley, Novel, and Woven by Toyota.
“Our strong second quarter performance demonstrates continued momentum across every dimension of our business,” said Michael Intrator, Co-Founder, Chairman and CEO of CoreWeave. “We are scaling rapidly as we look to meet the unprecedented demand for AI.”
🌐 Why it Matters: CoreWeave’s results underscore the capital-intensive race to build AI-ready data center capacity, with hyperscalers and AI labs locking in multi-year commitments. The combination of GPU leadership, deep AI customer integration, and aggressive power scaling positions CoreWeave as a top-tier player—but sustaining this trajectory will depend on balancing high debt service costs with infrastructure monetization.
| Quarter | Revenue ($M) | CAPEX ($B) | Active Power (MW) | Contracted Power (GW) |
|---|---|---|---|---|
| Q2’24 | $395 | $2.6 | — | — |
| Q3’24 | $584 | $2.2 | — | — |
| Q4’24 | $747 | $2.4 | — | — |
| Q1’25 | $982 | $1.9 | 350 | 1.6 |
| Q2’25 | $1,213 | $2.9 | 470 | 2.2 |
| Source: CoreWeave Q2’25 earnings materials. | ||||
Major Announcements in Q2 2025
CoreWeave’s second quarter was marked by a series of high-profile contracts, technology rollouts, and infrastructure expansions:
- Contract Wins & Expansions
- Signed a $4 billion expansion with OpenAI, in addition to the previously announced $11.9 billion deal.
- Secured and expanded a new hyperscaler customer agreement.
- Added enterprise and AI lab clients including BT Group, Cohere, Hippocratic AI, Hologen, LG CNS, Mistral, Moonvalley, Novel, and Woven by Toyota.
- Named the Official AI Cloud Computing Partner of the Aston Martin Aramco Formula One® Team.
- Technology Leadership
- Became the first company to bring NVIDIA GB200 NVL72 systems online for customers at scale.
- Delivered the largest-ever MLPerf Training v5.0 submission, 34x larger than other entries and 4.5x faster than the next-best GB200 submission.
- Announced general availability of NVIDIA B200-based instances.
- Product & Platform Expansion
- Completed acquisition of Weights & Biases (W&B), integrating AI development tools directly into CoreWeave’s platform.
- Launched Mission Control Integration, a real-time AI cluster health management system, now available through W&B Models.
- Introduced W&B Inference for direct access to open-source AI models and W&B Weave Online Evaluations for real-time performance monitoring of AI agents.
- Hosted its largest-ever developer conference, Weights & Biases by CoreWeave Fully Connected.