Neocloud GPU rental demand is surging with agentic AI, allowing CoreWeave to reach traditional cloud-like operating margins at scale.
Nvidia's GB200 delivers 3x better token throughput per watt than the H100, making it the scarce bottleneck asset in a $700B+ capex buildout.
AI data center economics are stronger than feared; hyperscalers earn cloud-like margins on gigawatt-scale GPU infrastructure.