Memory infrastructure for AI agents and applications.
Persistent VMs with 100 CPUs each, charged by usage, never for what they sit on.
