Thinking Machines signs $65m-a-year inference deal with Crusoe

Crusoe will run inference for Thinking Machines Lab's open models on Crusoe Cloud under a $65m annual agreement. Workloads, including Inkling and GLM 5.2/5.3 models, will run on a dedicated Nvidia HGX B200 cluster with Quantum-2 InfiniBand networking. The deal pushes Crusoe Managed Inference past $100m in contracted annual recurring revenue.

• Crusoe-Thinking Machines deal is worth $65m per year
• Inference runs on a dedicated Nvidia HGX B200 cluster
• Crusoe Managed Inference tops $100m in contracted ARR
• Mira Murati's lab is in talks to raise $6bn