Nvidia said RTX Spark devices powered by its N1X chip will begin shipping in October. According to ChainCatcher, the N1X supports up to 128GB of unified memory, while the full-spec version includes a Blackwell GPU with 6,144 CUDA cores and a 20-core Grace CPU, and the other version has a 5,120-core GPU and an 18-core CPU.

Nvidia also introduced NVIDIA PAIR, an open-source tool that connects multiple RTX devices, DGX Spark systems, and Apple devices on a home network to coordinate idle computing power for AI agent tasks. According to ChainCatcher, compatible devices include PCs with Nvidia GPUs from the RTX 20 series and later, RTX Pro GPUs, DGX Spark, and Apple devices with an M4 chip or newer.

The company said PAIR prioritizes unused compute and automatically adjusts as devices join or leave the network. In a media briefing example, a household had about 165 TFLOPS of underused compute, and the Qwen 3.6 35B A3B model completed an agent task in an average of 18 minutes on a single Spark laptop and 8 minutes 48 seconds on a three-device PAIR cluster. Perplexity Portable Computer, Hermes Agent, and OpenClaw will also get simpler local deployment support.