NVIDIA's RTX Spark brings 1-petaflop AI to PCs with local autonomous agents

NVIDIA CEO Jensen Huang unveiled the RTX Spark at Computex 2026, a 1-petaflop PC chip designed to run AI models and autonomous agents locally. Developed in partnership with MediaTek and Microsoft after three years of collaboration, the chip integrates the full CUDA and RTX ecosystem into Windows PCs. It enables private, low-latency AI inference without relying on cloud servers, targeting a surge in demand for 'agentic AI'—agents that handle routine tasks autonomously. The chip will ship this fall in laptops and compact desktops from Dell, HP, Lenovo, ASUS, Microsoft Surface, and MSI, with Acer and Gigabyte models to follow. Huang emphasized that the RTX Spark represents a fundamental shift from app-centric PCs to agent-centric personal computers, promising to bring AI capabilities to every home.

Market reaction was immediate and stark: NVIDIA shares jumped 4%, while rivals AMD, Intel, and Qualcomm fell 4.9% to 8.5%. HP and Dell rose over 7% each, and Lenovo climbed 5%. Industry analysts called it the 'RTX Spark moment' for personal computing, likening it to the iPhone or ChatGPT. Qualcomm CEO Cristiano Amon, speaking ahead of Computex, declared 2026 the 'year of agents,' noting that local edge computing is now essential for always-on autonomous operation. Huang also revealed that early adopters of NVIDIA's new Vera central processor include OpenAI, Anthropic, and SpaceX, opening a potential $200 billion market for the company.