Nvidia has announced the Personal AI Router (PAIR), a free, open-source tool that links a household's idle computers together to run local AI inference tasks such as agentic workflows. Rather than being a piece of hardware, PAIR is software that detects compatible devices on a home network, connects them and puts their spare processing power to use only when they are not otherwise busy, avoiding disruption to tasks like gaming. This matters because it offers a way for people to run their own local AI models using equipment they already own, rather than relying on cloud services.

PAIR works with Nvidia GeForce RTX 20-series cards and newer, RTX Pro GPUs, DGX Spark systems, and Apple devices with M4 chips or later, and pairs with tools such as Ollama and LM Studio. Nvidia says a household with several high-end devices could have around 165 teraflops of unused compute, though the company expects most real-world users to combine just one laptop with one gaming PC. Security is handled via a six-digit pairing code and mutual TLS encryption between devices, and the beta is available now on Windows, Linux and macOS. Alongside PAIR, Nvidia also announced that AI agent apps Perplexity Portable Computer, Hermes Agent and OpenClaw will get simplified local setup on Windows GPUs.