NVIDIA CEO Jensen Huang Officially Launches RTX Spark Series SoC Chips at GTC Taipei 2026, Redefining Personal Computers in Collaboration with Microsoft
NVIDIA CEO Jensen Huang officially launched the RTX Spark series SoC chips at GTC Taipei 2026, collaborating with Microsoft to redefine personal computers.
The SoC features TSMC's 3nm process and a dual chiplet design: MediaTek CPU (high-end 20-core Arm architecture) + Blackwell architecture RTX GPU (6144 CUDA cores, 1 PFLOP AI computing power), equipped with 128GB unified memory, with high-end performance comparable to desktop-level RTX 5070.
Huang stated that the new PC operating system will consist of a traditional OS + large language model (LLM), with LLM becoming the modern DirectX, and each PC will have its own autonomous agent. The first products are expected to be launched in the fall of 2026.
In terms of market dynamics, the PC industry is accelerating its transformation towards local AI supercomputing devices, with funding shifting from traditional laptops to high-performance AI SoC platforms. NVIDIA and Microsoft benefit from the integrated hardware-software ecosystem, while traditional PC chip manufacturers face pressure due to performance and ecosystem lag.
Source: Public Information
ABAB AI Insight
NVIDIA has previously laid out its edge AI computing strategy through the Jetson and Grace series, and the release of the RTX Spark SoC continues its path of penetrating from data centers to personal terminals. The unified memory architecture of CPU + GPU significantly reduces power consumption and latency, enabling local trillion-parameter models and agents to run continuously.
In terms of capital strategy, NVIDIA and Microsoft are concentrating resources on SoC design, agent runtime, and ecosystem development, motivated by the goal of bringing AI from the cloud to personal devices, forming a dual-driven model of home AI hubs and developer workstations to seize dominance in the next generation of personal computing platforms.
Similar cases include Apple's M series chips achieving performance leaps through unified memory, and NVIDIA CUDA establishing ecological barriers in gaming and computing; the current PC industry is at a critical juncture of transforming from traditional computing devices to AI agent supercomputers.
Essentially, this is a technological replacement: personal computers are shifting from a CPU + discrete GPU architecture to a highly integrated AI SoC platform, driven by the demand for LLMs and agents pushing for local high computing power, making unified memory and full-stack software the new standard, thereby reconstructing the hardware form, software ecosystem, and pricing power distribution in the PC industry.
ABAB News · Cognitive Law
The more unified the chips, the closer AI is to users.
LLM is not a function; it is the soul of the new operating system.
Excellent hardware sells agents, while traditional hardware sells computing power.