Can AI run without the cloud? Nvidia’s new system surprises everyone

NVIDIA’s new Personal AI Router (PAIR) introduces a way to handle local AI tasks by pooling the power of various computers in your home. Here is what you need to know about NVIDIA PAIR.

Relying on the cloud for AI won’t always be necessary; the various computers in your home can work together to handle AI tasks. NVIDIA announced its new Personal AI Router (PAIR) at IFA 2026. Notably, PAIR is not a hardware router but free, open-source software that connects home PCs to execute local AI inference tasks. It works with tools like Ollama and LM Studio. Let’s explore how NVIDIA PAIR works.

How does NVIDIA PAIR work?

PAIR automatically discovers compatible computers on the local network and connects them for AI workloads. Its standout feature is its ability to utilize the processing power of these computers; if one PC is busy with gaming or other tasks, PAIR can offload AI processing to other available systems.

In Agentic AI, large operations can be broken down into smaller tasks. PAIR distributes these tasks across computers on the network, preventing the entire processing load from falling on a single GPU. According to NVIDIA, PAIR adapts dynamically as devices join or leave the network.

Which computers support PAIR?

PAIR primarily works with systems equipped with NVIDIA GeForce RTX GPUs, including the RTX 20 Series and newer, RTX Pro GPUs, and DGX Spark Systems. Additionally, systems powered by Apple M4 chips or newer can also use it. According to NVIDIA, the PAIR beta is available for Windows, Linux, and macOS, and can be used via both graphical and terminal interfaces.

Is PAIR secure?

NVIDIA utilizes a 6-digit pairing code to connect devices via PAIR. Subsequently, communication between computers is secured using mTLS (Mutual Transport Layer Security), establishing an encrypted communication channel between the two systems.

A New Boost for Local AI

According to NVIDIA, PAIR will accelerate local AI tasks by pooling the unused computing power of idle home PCs. New optimizations for llama.cpp and vLLM have also been introduced to enhance local AI inference performance. Additionally, the local setup process has been simplified for Hermes Agent, OpenClaw, and Perplexity Portable Computer.

RTX Spark PCs Arriving in October

NVIDIA’s new RTX Spark Windows PCs are set to launch in October 2026. The company is designing them for AI agents and local AI workloads, providing users with new options to run AI directly on their own systems.