NVIDIA has unveiled Personal AI Router (PAIR) at IFA 2026, a free, open-source tool that lets households pool the processing power of multiple idle PCs to run AI workloads more efficiently. The tool addresses a common bottleneck: while AI agents already split complex tasks into smaller jobs, those jobs often end up competing for the same GPU, slowing everything down. By tapping into spare computers sitting unused around the home, PAIR aims to speed up AI tasks without hogging the resources of the machine someone is actively using.

PAIR works by scanning a local network for idle PCs and routing AI task requests to whichever machine has spare capacity, so subagents handling a job—such as sorting a cluttered email inbox—aren't stuck queuing for a single GPU. NVIDIA notes that more than half of US households own two or more PCs, most of which sit idle for large parts of the day. The tool is currently in beta and supports Windows, macOS and Linux via graphical and terminal interfaces, working with NVIDIA GeForce RTX 20 Series and newer GPUs, RTX PRO workstation cards from Turing architecture onward, NVIDIA DGX Spark, and Apple M4 or newer chips.