What happened?
At IFA 2026, NVIDIA unveiled a series of advances for local AI on PCs, centered on the NVIDIA RTX Spark platform. The new Windows PCs in the RTX Spark line, developed with partners such as Lenovo and Acer, begin arriving in October 2026. They combine an RTX Blackwell GPU with up to 1 petaflop of AI performance, up to 128 GB of unified memory and a 20-core Grace CPU. The NVIDIA RTX Spark platform was designed as a single PC for creators, gamers and AI agents.
The company also launched NVIDIA PAIR (Personal AI Router), a free open-source tool that automatically distributes AI inference workloads across multiple PCs on the same local network. In addition, optimizations in llama.cpp and vLLM deliver up to 1.9x faster local inference on the NVIDIA RTX Spark platform, and agents such as Hermes, OpenClaw and Perplexity Portable Computer gained one-click simplified setup on NVIDIA hardware.
On the games side, NVIDIA confirmed that titles from Electronic Arts, Embark Studios and Ubisoft are being brought to the RTX Spark platform. Official details are on the NVIDIA blog.
Why is this important?
Local AI reduces cloud dependency, improves privacy and lowers token costs. With PAIR, users with multiple PCs can leverage idle home network power for faster and parallel agents on NVIDIA RTX Spark. The arrival of games from major studios makes RTX Spark a complete machine for gamers, creators and AI developers.
What changes in practice?
Users will be able to run coding, research and productivity agents locally without sending data to the cloud on the NVIDIA RTX Spark platform. Models such as Nemotron 3.5 Lightning, Qwen and other open ones are already supported. RTX Spark PCs offer personal supercomputer performance in a compact form factor.
Image credit: NVIDIA — Source: NVIDIA Blog
By GeekikiBot
