CoreWeave has outlined capital expenditures of $12–14 billion for 2025, with deferred projects pushing total capex toward $30 billion in 2026. Management is locking in next-generation GPUs and high-speed networking hardware from Nvidia and other vendors to support surging AI compute demand. Investors are focused on the useful lifecycle of these servers, since any shortening of expected hardware depreciation schedules could delay the company’s path to positive free cash flow. Current forecasts assume multi-year GPU lifespans, but evolving energy-efficiency innovations and the rapid release of successor chips may force accelerated refresh cycles.

CoreWeave secured one of the first cloud-provider deployments of Nvidia’s Rubin AI platform following CEO Jensen Huang’s October production announcement. Nvidia’s 5 percent equity stake in CoreWeave underscores the strategic alliance, and management says Rubin will improve cost-efficiency for large-scale training and inference workloads. This endorsement has driven a 25 percent share rally over the past week and reassured investors that CoreWeave can maintain technology leadership among AI-focused cloud services.

At the end of Q3 2025, CoreWeave reported a revenue backlog of $56 billion—nearly four times year-ago levels—and operated 41 dedicated AI data centers totaling 590 megawatts of active power capacity. The company holds 2.9 gigawatts of contracted power for future builds and plans to activate at least 1 gigawatt more over the next 12–18 months. Analysts forecast revenue of $5.1 billion in 2025 (up 165 percent year-over-year) and nearly $12 billion in 2026, suggesting the business could more than double its top line next year and justify a valuation expansion as capacity comes online.