Hosting a private AI chatbot with an NVIDIA GPU requires balancing raw compute power with a secure, low latency net