Nvidia used Computex in Taiwan to present a sweeping set of AI products and software, with the most attention going to Vera, a new CPU the company says is built for agentic workloads.

The chipmaker said Vera is intended for tasks associated with AI agents, including using tools, writing code and processing data. Nvidia claims the processor can handle those jobs 1.8 times faster than conventional x86 CPUs. The company also said customers including OpenAI, Anthropic, SpaceXAI, CoreWeave, Lambda, Dell, HPE and Lenovo plan to adopt the chips.

CEO Jensen Huang framed the launch as part of a broader shift in computing. In a statement, he said AI agents will become the biggest users of compute and described Vera as the first CPU designed specifically for that future.

The CPU announcement was part of a larger Computex rollout that signaled Nvidia’s push beyond graphics processors into a more expansive AI platform strategy. The company introduced DSX, a software platform for planning, simulating, building and running AI factories. Nvidia said one component, DSX MaxLPS, is designed to improve power and cooling efficiency so operators can fit 40% more GPUs into the same power envelope. Another piece, DSX OS, is being positioned as an open-source operating system for AI factories.

Nvidia also unveiled a new desktop product, DGX Station for Windows, which it says can run AI models with as many as 1 trillion parameters locally. The company is targeting engineers, researchers, developers and enterprises with the system, pitching on-device execution as a way to improve speed, security and cost control.

On the robotics side, Nvidia announced Cosmos 3, a world model designed to work across text, images, video, audio and actions. The company said it aims to support physical AI and robotics applications. It also revealed Alpamayo 2 Super, an open reasoning model for robotaxis, along with H2 Plus, an open reference design for humanoid robots that combines hardware, software and onboard computing.

Nvidia said its Vera Rubin infrastructure platform is moving into full production. The system combines GPUs, CPUs, networking, storage and security, and the company said it is now ramping across 350 factories in 30 countries. Nvidia claims the platform delivers 10 times the agent throughput of its earlier Grace Blackwell generation.

The company’s latest announcements underline how far it has moved from its original identity as a GPU specialist. While graphics chips remain central to its business, Nvidia is increasingly building around the full AI stack, from processors and systems to models and software. That expansion comes as the company works to extend its influence beyond the CUDA software ecosystem that helped make it a dominant force in the current AI wave.

For Nvidia, Computex was less a single product launch than a statement of direction. The company is betting that the next phase of AI will require new kinds of infrastructure, new computing architectures and more tightly integrated systems built for agents, robotics and large-scale model deployment.