CoreWeave reported third-quarter adjusted EBITDA of $838 million, representing margins above 60 percent thanks to its AI-optimized infrastructure and a book of contracted customer demand. Management highlighted that investments in custom cooling systems and power-delivery upgrades enabled higher utilization rates for its GPU clusters, supporting more efficient AI training workloads. The company’s resilient outlook is underpinned by long-term service agreements covering over 80 percent of capacity through 2026, providing clear visibility on recurring cash flows.

Despite revenue growth of 133 percent year-over-year to $1.36 billion in the third quarter, CoreWeave’s operating expenses climbed to $1.31 billion, compressing operating margins to just 4 percent and resulting in a net loss of $110 million. Interest expenses totaled $310.6 million, driven by $10.3 billion in long-term debt used to finance hardware purchases and data center expansions. Depreciation charges on thousands of high-end GPUs further weighed on earnings, underscoring challenges in translating top-line growth into sustainable profitability.

Investor focus is shifting to the potential of continuous inference workloads from agentic AI applications to create a durable compute floor. CoreWeave currently operates 590 megawatts of active power and has contracted an additional 2.9 gigawatts of pipeline capacity expected online within 12 to 24 months. As autonomous AI agents drive nonstop demand for GPU cycles, management projects that revenues could accelerate significantly in 2026, potentially surpassing current consensus forecasts by leveraging its bare-metal cluster design and direct partnerships with technology developers.