Jensen Huang’s GTC Taipei Keynote: Vera Rubin Enters Full-Scale Production; AI Agents Are a Key Focus

Author: WallStreetCN

On June 1, Jensen Huang, founder and CEO of NVIDIA, delivered the keynote address for NVIDIA GTC Taipei 2026 in Taipei, Taiwan.

During the keynote, Huang unveiled several new products and platforms, including the Vera Rubin architecture, the Vera CPU designed for AI agents, the Nemotron 3 Ultra AI model, and the DSX AI factory platform.

NVIDIA also signaled its intent to enter the next-generation AI PC market—targeting the entire Windows ecosystem.

At the event, NVIDIA revealed that leading Chinese automakers and autonomous driving companies—including BYD, Geely, Zeekr, Xiaomi, and Pony.ai—have adopted or are developing intelligent driving systems based on the NVIDIA Hyperion platform.

NVIDIA’s U.S. stock rose 2.7% in after-hours trading.

Vera Rubin Architecture Now in Full Production

Huang stated that the Vera Rubin architecture has entered full production, with OpenAI, Anthropic, and SpaceX becoming the first customers for NVIDIA’s Vera chips.

AI agents will be a key strategic focus going forward, and NVIDIA will launch the Vera CPU specifically designed for AI agents.

Huang emphasized that the NVIDIA Vera BlueField-4 STX delivers agent-level AI storage processing capabilities with chip-level security.

Aiming for the Next-Generation AI PC Entry Point

Notably, NVIDIA announced a new processor for Windows-based personal computers, challenging Intel.

NVIDIA will collaborate with MediaTek to develop the RTX SPARK PC chip, which will be manufactured by TSMC. Huang revealed that NVIDIA’s PC chips will launch this fall, supporting both laptops and desktops.

Huang added that NVIDIA will release a new generation of PC chips alongside each new generation of AI processors.

Earlier, NVIDIA’s official social media account teased the launch with a three-word slogan: “A new era of PC”—signaling its ambition to capture the next-generation AI PC entry point.

Analysts note that NVIDIA aims to integrate the CPU, GPU, and AI accelerators into a single SoC, delivering complete core computing platforms directly to OEMs such as Dell and Lenovo—fundamentally shifting its position within the value chain. This vertical integration strategy mirrors Apple Silicon—but targets the entire Windows ecosystem.

NVIDIA: Collaborating with Unitree and Global Humanoid Robot Makers

Beyond partnering with China’s Unitree Technology, NVIDIA plans to collaborate with humanoid robot manufacturers across the U.S., Europe, and South Korea on robotics research initiatives.

Launching a New AI Model + “Action Guide” for AI Factories

Additionally, NVIDIA announced the new Nemotron 3 Ultra AI model.

NVIDIA introduced the NVIDIA DSX platform—a comprehensive “action guide” for infrastructure builders to construct AI factories. Huang stated:

“With the DSX platform, you can simulate an entire factory at zero cost, validate performance before installing even a single rack, and operate with the reliability required for production-grade AI.”

The Hyperion Autonomous Driving Platform Ecosystem Emerges

NVIDIA DRIVE Hyperion has become the global autonomous taxi platform.

As disclosed at the event, leading Chinese automakers and autonomous driving companies—including BYD, Geely, Zeekr, Xiaomi, and Pony.ai—have adopted or are developing intelligent driving systems based on the NVIDIA Hyperion platform.

NVIDIA also launched Alpamayo 2, a super open inference model designed for autonomous taxis.

Building a Developer Ecosystem for the Robotics Era

In the humanoid robotics domain, NVIDIA is positioning itself as an “infrastructure provider.”

At this event, NVIDIA unveiled the NVIDIA Isaac GR00T humanoid robot reference platform for academic research, along with a large open-source toolkit and skill suite for Physical AI.

During the keynote, NVIDIA selected Unitree’s H2 Plus as the reference robot to demonstrate the Isaac GR00T reference humanoid robot platform. This platform integrates the Jetson Thor compute chip, the GR00T software stack, dexterous hands, and full-body control systems—offering robot manufacturers an end-to-end development framework spanning hardware and software.