Nvidia launches PAIR to pool idle computers for local AI

Nvidia has introduced Personal AI Router (PAIR), a free, open-source software tool designed to cluster local computers to handle AI inference tasks. Available in beta for Windows, Linux, and macOS, PAIR connects compatible machines over a local network to distribute workloads for tools like Ollama and LM Studio. The software works with Nvidia GeForce RTX 20-series cards and newer, RTX Pro GPUs, DGX Spark systems, and Apple M4 chips or newer.

Instead of relying on a single GPU, PAIR utilizes idle processing power across multiple devices to run agentic workflows in parallel. The system dynamically adapts as computers join or exit the network, ensuring that if a user launches a game on a connected desktop, the AI workload seamlessly shifts. To secure these distributed networks, PAIR requires a six-digit pairing code and encrypts communication channels using Mutual Transport Layer Security (mTLS).

Nvidia product manager Seth Schneider highlighted that a household with multiple modern devices—such as an RTX Spark laptop, a DGX Spark desktop, an RTX 5090 laptop, a gaming PC, and a MacBook Pro—could harbor around 165 teraflops of underutilized compute. Schneider described this idle capacity as a "treasure trove of free tokens" even when factoring in average American electricity costs. However, Nvidia expects a more typical setup to consist of just one laptop and a single gaming PC.

For AI practitioners, PAIR lowers the barrier to running complex local models without renting expensive cloud instances. Alongside PAIR, Nvidia announced that three major AI agent applications—Perplexity Portable Computer, Hermes Agent, and OpenClaw—now feature simplified local setup on Windows. This integration allows developers to deploy local agents with Nvidia GPUs in just a few clicks, bypassing tedious manual configurations.