Nvidia's Free PAIR Software Lets Every PC in Your Home Share the AI Workload

Key points

• Nvidia released a free, open-source software tool called PAIR (Personal AI Router) in beta on Windows, Linux, and macOS.
• PAIR links multiple computers on a home network so they share AI processing tasks, using spare power only when each machine is idle.
• Compatible devices include Nvidia GeForce RTX graphics cards from the RTX 20-series onwards, plus Apple M4 chips or newer.
• Nvidia estimates a well-equipped household could have roughly 165 teraflops, a measure of raw computing speed, of untapped processing power available.
• Three AI agent apps, Perplexity Portable Computer, Hermes Agent, and OpenClaw, are also getting simplified one-click setup for local use on Nvidia hardware.

Your gaming PC sits idle most of the day. Your laptop is charging on the kitchen counter. Your partner's MacBook is asleep on the sofa. Nvidia now wants all of those machines to quietly cooperate on AI tasks while nobody is using them.

The company announced PAIR (Personal AI Router) this week, a free software tool that discovers computers on your home Wi-Fi network, links them together, and turns their spare graphics-card muscle into a shared AI processing pool. First reported by The Verge AI, the tool works with popular local AI apps like Ollama and LM Studio, which let people run AI chatbots and tools on their own hardware instead of sending data to a cloud server.

What does PAIR actually do?

PAIR is software, not a physical device. Once installed, it quietly runs in the background and hands out AI processing jobs to whichever machine has spare capacity at that moment. If you start a game, PAIR moves the work elsewhere. If a laptop leaves the network, the remaining devices carry on.

That matters because local AI tasks can be slow on a single machine. A so-called agentic workflow, where AI software breaks a big job into many smaller steps and runs them one after another, can easily overwhelm one graphics card. Spreading the load across several machines keeps things moving.

Nvidia product manager Seth Schneider gave press a hypothetical family: two adults with high-end laptops, a teenager with a gaming desktop, and a kid with a MacBook Pro. Together, Schneider said, that household holds about 165 teraflops of unused computing power. In plain terms: a teraflop is a measure of how many calculations a chip can do per second; 165 teraflops is a serious amount of number-crunching sitting dormant.

Is it safe to link all your computers together?

Security is handled through a six-digit pairing code, similar to pairing a Bluetooth device. After that, all communication between machines travels over an encrypted channel using mTLS (Mutual Transport Layer Security), a standard that makes each computer verify the other's identity before sharing anything. Your home AI traffic stays inside your home.

Who is this actually for?

When asked, Schneider said Nvidia's realistic target user is someone with one Windows or MacBook laptop and one gaming PC, not a tech-packed household of five.

Compatible hardware covers Nvidia GeForce RTX 20-series graphics cards and newer, Nvidia RTX Pro and DGX Spark professional systems, and Apple silicon chips from the M4 generation onwards.

The PAIR beta is available now. Alongside it, Nvidia announced that three AI agent apps, Perplexity Portable Computer, Hermes Agent, and OpenClaw, will offer simpler one-click local setup on Windows with Nvidia graphics cards, cutting the fiddly manual configuration those tools previously required.

For most households, the practical upside is straightforward: faster, more capable local AI without spending a penny on new hardware.