Nvidia Launches PAIR to Turn Idle PCs Into Local AI Clusters

AI agents increasingly divide complex jobs among concurrent tasks, creating bottlenecks when every inference request competes for one GPU. Nvidia’s Personal AI Router, or PAIR, addresses that constraint by directing separate requests across compatible computers on the same local network. The software is designed to keep prompts, files and agent context on premises, reducing reliance on cloud computing. It does not pool GPU memory, create a single virtual GPU or split one inference request across machines.

Nvidia unveiled the free, open-source PAIR beta at IFA 2026 on Sept. 3, 2026. The software automatically discovers participating devices and routes jobs according to engine availability, model availability and current workload, while presenting Ollama-compatible and OpenAI-compatible endpoints. It supports Ollama and LM Studio across Windows, Linux and macOS. Eligible hardware includes GeForce RTX 20 Series or newer GPUs, RTX PRO cards based on Turing or later, DGX Spark systems and Apple M4 or newer silicon. Nvidia lists 8GB of RAM as the minimum and recommends at least 20GB of storage.