Crusoe signs a $65M annual inference deal

Crusoe and Thinking Machines Lab signed a $65 million annual inference agreement for production workloads on dedicated NVIDIA HGX B200 systems. Crusoe’s partnership announcement names Inkling and GLM 5.2 and 5.3 workloads and says Managed Inference has passed $100 million in contracted ARR less than a year after launch; those are vendor-reported contract figures, not usage revenue. Teams choosing between owned and outsourced serving capacity should compare SLA-backed throughput against reservation risk; Crusoe’s financing-versus-delivery analysis shows why contracted capacity still needs delivery evidence.