The use of AI has moved from the PoC phase to the phase of systematization, service creation, and actual operation. In this context, when it comes to hardware in an AI operational environment, attention tends to be focused on hardware equipped with low-power embedded CPUs or AI-specific chips, but it is also possible to build an operational environment for AI inference using general-purpose machines equipped with x86 architecture CPUs. Machines equipped with such general-purpose CPUs offer a wide range of options and freedom of configuration, giving them advantages over dedicated hardware. However, since it is not dedicated hardware, various ingenuity is required to use AI. In this presentation, I will present techniques for building an AI inference environment on such general-purpose machines.