Jensen Huang took the stage at the Taipei Music Center on June 1, announcing full production of Nvidia's next-generation AI infrastructure system, Vera Rubin, by fall 2026. The company projects $1 trillion in cumulative orders for its Blackwell and Rubin systems by 2027.
The NVL72 systems bundle Rubin GPUs, Vera CPUs, Groq 3 LPX inference trays, Spectrum-6 networking, and BlueField-4 storage into a cohesive stack using Nvidia's 'extreme co-design' approach. This optimization delivers up to 10 times lower cost per token compared to Blackwell, with significantly greater throughput per megawatt when paired with Groq components. Production shipments begin autumn 2026.
Huang emphasized Taiwan's critical role, noting the company collaborates with over 350 factories across 30 countries, with 150 in Taiwan. The supply chain for Vera Rubin is twice the size of Blackwell's.
The platform is designed for agentic AI workloads-systems that take autonomous actions and run multi-step reasoning continuously. This positions Nvidia to capture enterprise demand as AI moves from experimental to operational.
The $1 trillion order projection reflects the full arc of current and next-gen product lines. However, with 150 partner factories concentrated in Taiwan, any disruption to the island's manufacturing capacity would ripple through Vera Rubin's production timeline, introducing execution risk for investors.