Nvidia on Monday, June 1, made a slew of announcements, including a new chip that puts AI capabilities directly into laptops and desktop computers.
The RTX Spark PC chip was unveiled by Nvidia CEO Jensen Huang in his keynote speech at the Computex conference in Taipei, Taiwan. The PC trade show runs from June 2 to June 5, with some of the world’s largest tech companies expected to showcase their latest innovations in the Intel-Windows ecosystem.
Nvidia’s RTX Spark PC chip is part of a three-year collaboration with Microsoft to “reinvent the PC” for the AI era. It is designed to run AI agents locally rather than relying solely on cloud computing. It was also developed in partnership with Taiwan’s MediaTek.
“The PC is being reinvented. For forty years, you launched apps. Click. Type. With RTX Spark and Microsoft Windows, you ask — and the PC does the work. RTX Spark brings everything NVIDIA has built — CUDA, RTX, our AI platform — into a single superchip. Local agents. Frontier models. Creative workflows. RTX games. All on a laptop. This is the new PC. The personal AI computer,” Huang said in his keynote speech at Taipei Music Hall on Monday.
The RTX Spark chip will power a new wave of Windows laptops and desktop PCs launching this fall. The devices, which aim to deliver all-day battery life and premium displays, will be made and sold by major brands such as Asus, Dell, HP, Lenovo, Microsoft Surface, and MSI. Models from Acer and gigabyte will follow, as per Nvidia.
Huang also spent much of his keynote address on Monday highlighting the new Vera CPU (central processing unit). The new hardware could potentially transform how users engage with AI. It also marks Nvidia’s first major push into the Windows PC market. So far, Qualcomm had exclusive license to develop Arm-designed system-on-chip (SoCs) for Windows 11 devices.
Here is a quick look at everything Nvidia announced at Computex 2026.
Story continues below this ad
RTX Spark PC chip
The RTX Spark chip features an NVIDIA Blackwell RTX GPU with 6,144 CUDA cores and fifth-generation Tensor Cores with FP4 precision, connected via Nvidia’s NVLink to a high-performance, 20-core NVIDIA Grace CPU.
To meet the processing demands of on-device agents, RTX Spark has been designed to deliver up to 1 petaflop of AI compute and 128GB of unified memory.
This will enable users to carry out a wide range of compute-intensive tasks such as render ultralarge 90GB 3D scenes, edit 12K 4:2:2 video, run 120-billion-parameter large language models (LLMs) with 1 million tokens context, and play AAA games at 1440p resolution and over 100 frames per second, as per the company.
Nvidia also said it is working with Adobe to optimise Adobe Premiere and Photoshop for RTX Spark-powered laptops.
Story continues below this ad
In an attempt to address security risks posed by AI agents, Nvidia said that the new RTX Spark-powered devices will come with new Windows security primitives and Nvidia OpenShell, which will provide “additional policy capabilities for the user to define what agents can and cannot do…”
Vera CPU for agents
Nvidia on Monday announced Vera CPUs to drive diverse workloads across industries, including agentic AI, reinforcement learning, and data processing. Its early adopters include OpenAI, Anthropic, and SpaceX. Anthropic is evaluating adding Vera to scale CPU-intensive agentic workloads, Nvidia said.
The central processor gives Nvidia access to a new $200 billion market. “This (Vera CPU) is going to be our new major growth driver,” Huang said in an earnings call last month.
Vera features 88 Olympus cores, Spatial Multithreading and a LPDDR5X memory subsystem that delivers up to 1.2TB/s of bandwidth. This helps “deliver higher end-to-end throughput and faster time to solution for users, improving responsiveness and efficiency across training, inference and agentic execution,” Nvidia said. It will be available to system builders and cloud partners starting this fall.
Story continues below this ad
DGX Station for Windows
DGX Station is an AI supercomputer designed to build, run and connect always-on AI agents to Windows applications and workflows. It is capable of running frontier AI models of up to 1 trillion parameters locally.
It is powered by the NVIDIA GB300 Grace Blackwell Ultra GPU connected to a 72-core NVIDIA Grace CPU via the NVIDIA NVLink. It features up to 748GB of coherent memory and up to 20 petaflops of FP4 performance.
In addition, DGX Station for Windows comes with support for networking at up to 800Gb/s thanks to the ConnectX-8 SuperNIC that enables extremely fast network data transfers for AI workloads and high-speed connectivity of multiple DGX Station systems for even larger workloads.
DGX Station for Windows is expected to be available from Asus, Dell, Gigabyte, HP, MSI, and Supermicro in the fourth quarter of this year.
Story continues below this ad
Open-source agent tools
Nvidia has announced a major collection of open source physical AI tools that are designed for developers to turn complex robotics, autonomous vehicle (AV), vision AI and industrial digital twin workflows into agent-automated tasks.
It is part of the company’s Agent Toolkit to turn physical AI development processes into repeatable instructions that coding agents can follow. This includes which tools to call, what outputs to produce and how developers can validate results.
The new skills can be found at GitHub and skills.sh for use with any coding agent. Microsoft, CoreWeave and Nebius are also integrating these agent skills and tools with their cloud services to enable developers to streamline and scale synthetic data generation and deployment, as per the company.
Cosmos 3 foundational AI model
Nvidia unveiled a new open-weight model for physical AI with a mixture-of-transformers architecture that combines vision reasoning, world generation, and action prediction in a single system.
Story continues below this ad
The company claimed Cosmos 3 is the world’s first fully open omnimodel that can natively understand and generate text, images, video, ambient sound and actions with leading physics accuracy. It is capable of reducing physical AI training and evaluation cycles from months to days.
Alpamayo 2 Super reasoning AI model
Nvidia introduced a 32-billion-parameter reasoning‑based vision language action (VLA) model called Alpamayo 2 Super. It is part of the company’s family of open AI models used to generate simulation frameworks and physical AI datasets for robotaxi development.
Alongside the model, the company announced new tools, models, and agent skills for real-world data capture, closed-loop training, and in-vehicle deployment by robotaxis.
