Cedana (YC S23) brings hyperscaler and frontier-lab orchestration capabilities for AI workflows, with GPU live migration as their core capability. Described as "VMware for GPUs," Cedana enables enterprises to orchestrate AI workloads precisely, reliably, and efficiently.

The company builds live migration and checkpoint/restore infrastructure for GPU-heavy workloads, enabling stateful reliability even through catastrophic GPU failures. Their technology interposes at the Driver API to reliably, transparently, and deterministically capture and restore GPU state for multi-GPU workloads.

Performance: Increases cost savings up to 80%, accelerates time to first token 2-10x, and restored cold starts are an order of magnitude faster than native cold starts.