NVIDIA Reshapes the Future of Computing with the True AI PC
For decades, the Windows PC ecosystem has operated on a relatively stable division of labor: Microsoft handled the operating system, Intel and AMD managed the x86 processor platform, and NVIDIA dominated graphics and, more recently, AI acceleration. However, at COMPUTEX 2026, NVIDIA CEO Jensen Huang unveiled a strategy that moves the company from a component supplier to the core of a new computing paradigm, centered around a single, powerful concept: the AI Agent.
The central theme of the presentation was a fundamental shift in how we interact with computers. The PC is envisioned to evolve from a tool for manually launching applications into a personal AI platform designed to run intelligent agents natively. Huang compared this transformation to how the telephone evolved into the smartphone, where making calls is no longer its primary function. In this new era, the PC will become an AI supercomputer for personal and professional workflows, with AI agents at its heart.
Agent, Agent, and More Agent.
At the core of this new PC ecosystem is the NVIDIA RTX Spark processor. This chip is a powerhouse, featuring a Blackwell RTX GPU capable of 1 petaflop of FP4 AI performance, a custom 20-core Grace CPU co-developed with MediaTek, and 128 GB of unified memory connected via NVLink C2C with 600 GB/s of bandwidth. This hardware is designed not just for lightweight AI assistants but to bring data-center-level AI, high-end gaming graphics, and professional creative power into mainstream laptop form factors—as thin as 14mm and weighing around 3 pounds.
The vision for RTX Spark extends beyond a single device. Huang showcased a product line that includes laptops, desktops, and workstations, all compatible with the same Windows, CUDA, and AI software stacks. A desktop, for instance, could function as a personal AI host, running agents 24/7 to connect and manage everything from laptops and cameras to home security systems and other smart devices. For developers, the DGX Station for Windows, with its 748 GB of memory and 20 petaflops of compute, allows for the local development and testing of trillion-parameter models.
This new paradigm of agentic AI requires a complete redesign of the underlying computing infrastructure, from the user's desk to the data center. To power this, Huang announced Vera Rubin, NVIDIA's next-generation AI superchip platform. This system, designed for "AI Factories," integrates the Rubin GPU, the new Vera CPU, and advanced networking and security components to handle massive-scale agentic workloads. The logic is simple: agentic AI involves constant, low-latency tasks like tool calling, data retrieval, and code execution, which can easily bottleneck traditional systems.
The Vera CPU is specifically engineered to address this challenge. Unlike traditional CPUs designed for human interaction, the Vera CPU is built for the rapid-fire demands of countless AI agents. Featuring 88 proprietary Olympus cores in a monolithic design to reduce latency, it aims to eliminate performance bottlenecks and maximize the utilization of the expensive GPUs in an AI factory. This ensures that the entire system is optimized for the new currency of the AI economy: generating revenue through tokens.
Beyond the hardware, NVIDIA is providing the tools for enterprises to build and deploy their own agents. This includes Nemotron 3 Ultra, a new, highly efficient open model whose training scripts and data will be made available to the public. This is part of a complete four-part stack for agent development: the model (Nemotron), an orchestration system (OpenShelf), tools and skills (CUDA X libraries), and the runtime environment (AI platform). This comprehensive approach is reinforced by NVIDIA DSX, a blueprint for designing and operating entire AI factories.
In essence, NVIDIA's latest announcements represent a cohesive, top-to-bottom strategy to build the foundational infrastructure for the age of AI. The same core computing model—model, harness, tools, and runtime—is designed to scale across every environment, from personal PCs (RTX Spark) and enterprise systems to data centers (Vera Rubin), autonomous vehicles, and robotics. NVIDIA is no longer just in the business of selling chips; it is building the entire computational framework for a future driven by AI agents.
