CoreWeave reported third-quarter adjusted EBITDA of $838 million, representing margins in excess of 60%. Management attributes this performance to an AI-optimized infrastructure design, long-term contracts covering a significant portion of capacity, and robust demand from enterprise and hyperscaler customers. These results exceed consensus forecasts by 5 percentage points on margin and underscore the scalability of CoreWeave’s bare-metal GPU clusters in delivering high-performance computing services without virtualization overhead.

Despite 133% year-over-year revenue growth to $1.36 billion, CoreWeave posted a net loss of $110 million in Q3, translating to a $0.22 per-share loss. Operating expenses surged to $1.31 billion, compressing operating margin to 4% from 20% in the prior-year period. The company’s capital-intensive model relies heavily on GPU hardware financed through debt, resulting in $10.3 billion in non⁠-current borrowings (up 89% year-over-year) and $310.6 million in quarterly interest expenses. Depreciation costs on aging accelerators further pressure free cash flow, while the lack of proprietary silicon limits pricing power against GPU vendor Nvidia and hyperscale rivals developing custom AI chips.

Investor concerns over leverage have overshadowed CoreWeave’s positioning to capitalize on the emerging agentic AI compute floor. Persistent inference workloads from autonomous software agents could drive a sustained increase in GPU utilization, expanding the total addressable market toward $200 billion with a 43% compound annual growth rate over the next several years. CoreWeave currently operates 590 megawatts of active power capacity, with a 2.9-gigawatt contracted pipeline expected to come online within 12 to 24 months. This build-out is poised to accelerate 2026 revenues significantly above consensus estimates as new capacity ramps under long-term commitments.