Crusoe has broadened its managed artificial intelligence platform by introducing serverless fine-tuning and self-service inference deployments, available via its Intelligence Foundry platform, aimed at enterprise AI teams customizing open-weight models without managing GPU infrastructure. This launch signifies a shift in the AI data center market from merely providing GPU access to facilitating a seamless transition from experimentation to production while managing costs and ensuring model ownership, according to Dave McCarthy, research vice president at IDC. Erwan Menard, senior vice president of product at Crusoe, stated that "the appetite for fine-tuning is accelerating faster than we expected," as enterprises seek ownership of their models. The platform supports a curated library of open-weight models including Qwen, DeepSeek, Gemma, and GPT-OSS, employing dynamic scheduling for fine-tuning jobs to reduce idle GPU spending. The new Self-Serve Deployments service operates on Nvidia H100 and H200 GPUs, enabling customers to deploy managed inference endpoints without direct infrastructure provisioning. The serverless fine-tuning and self-serve deployments are set to become generally available next week, with pricing based on a per-million-token basis for fine-tuning and per GPU hour for inference deployments.