NVIDIA founder and CEO Jensen Huang (黃仁勳) unveiled a sweeping vision for the personal computer at GTC Taiwan on June 1, announcing a partnership with Microsoft to redesign the PC around artificial intelligence agents — a move with significant implications for Taiwan's PC supply chain.
Speaking at the keynote, Huang said that 40 years on, Microsoft and NVIDIA would reinvent the PC — a machine that would no longer be a passive tool for launching applications and typing commands, but an active platform capable of understanding user intent, executing tasks, and running AI agents locally.
PC Becomes an AI Agent Platform
Huang traced the original success of the PC industry to Microsoft's creation of an open, extensible computing platform — spanning BIOS, chipsets, operating systems, drivers, and multimedia APIs — that allowed PCs to migrate from enterprise settings to consumer households. He said agentic AI would now trigger another architectural shift in the PC.
Huang said AI agents will operate not only in the cloud and inside corporate data centers, but directly on personal computers. Users would be able to speak their requirements to a machine, with agents then reading files, conducting research, operating software tools, generating content, and assisting with design and creative workflows. Huang compared the vision to the robot companions R2-D2 and C-3PO from the Star Wars films — AI assistants that stay with a user continuously and handle a broad range of tasks.
RTX Spark Platform and the MediaTek M1X Chip
NVIDIA introduced its RTX Spark platform at the event, covering laptops, desktops, and workstations. According to Huang, RTX Spark is built around a Blackwell RTX GPU with 6,144 CUDA cores and one petaflop of AI performance. It is paired with a 20-core Grace CPU co-developed with MediaTek (聯發科, ticker: 2454), 128 GB of unified memory, and is manufactured on TSMC's 3-nanometer process node.
Huang singled out the M1X chip — co-developed by NVIDIA and MediaTek — calling it "a very beautiful chip" and describing it as the product of NVIDIA's years of accumulated software stack and AI computing capabilities, combined with the Windows ecosystem.
He said the platform could run NVIDIA's full software stack — spanning digital biology, seismic data processing, astrophysics, AI, computer graphics, and CUDA applications — while simultaneously running existing Windows software and adding the ability to execute AI agents.
To illustrate the point, Huang demonstrated an architectural design workflow in which an AI agent on an RTX Spark system took a site plan, a sketch, style references, and a project brief, then operated Rhino to generate a spatial layout, imported the result into Blender for material adjustments, and used generative AI to produce photorealistic renderings.
From AI PC to Home AI Supercomputer
Huang suggested that households may eventually own a dedicated AI supercomputer in the same way many homes today have home theater systems, gaming consoles, and household appliances. Such a device would run AI agents around the clock, connect to other appliances in the home, and grow more capable as underlying models are upgraded.
NVIDIA also showed a Windows version of its DGX Station workstation at the event, featuring 768 GB of memory and support for trillion-parameter models, targeting developers working on large language models and AI agents.
Huang framed the announcement as the first genuine redesign of the PC in 40 years. For Taiwan's PC supply chain — encompassing brand vendors, ODMs, chip designers, memory makers, thermal management suppliers, and software ecosystem players — the shift toward AI-native personal computing signals a new upgrade cycle across the board.
（Related：
NVIDIA Declares the Age of Physical AI Has Begun
｜
Latest
）
