Nvidia launches Vera Rubin platform into full production

Nvidia Vera Rubin Platform Enters Full Production for Agentic AI Factories

Nvidia CEO Jensen Huang announced at GTC Taipei on Sunday that the company’s Vera Rubin platform is ramping into full production, marking what Huang called “the largest product launch, probably in the history of Taiwan.”nvidia

The five-rack platform — encompassing Vera Rubin NVL72 systems, the Vera CPU, Groq 3 LPX inference trays, Spectrum-6 SPX Ethernet racks and Vera BlueField-4 STX storage — is being manufactured by more than 350 factories across 30 countries, with 150 ecosystem partners in Taiwan alone.nvidia

“The supply chain we created for Vera Rubin is twice as large as Grace Blackwell,” Huang told the audience at Taipei Music Center. “We need it all to support the demand.”nvidia

A Platform Built for Agents

Vera Rubin NVL72 connects 36 Vera CPUs and 72 Rubin GPUs, delivering up to 10x higher inference performance per watt and 10x lower cost per token compared to the previous Grace Blackwell generation. When paired with the Groq 3 LPX, the system achieves up to 35x higher throughput per watt for trillion-parameter models.nvidia

The platform also introduces Spectrum-X Ethernet Photonics — which Nvidia describes as the world’s first 200Gb/s SerDes Ethernet switch with co-packaged optics — now in production for million-GPU AI factories. The technology replaces traditional pluggable optics with integrated silicon photonics, reducing power consumption per port by roughly 5x. CoreWeave, Lambda and Oracle Cloud Infrastructure are among the first adopters.financialcontent

Production shipments are set to begin this fall. Dell Technologies ↘2.34% has already delivered the first Vera Rubin NVL72 rack to CoreWeave, according to Dell CEO Michael Dell, who shared images of the system on social media.barchart

Taiwan at the Center of the AI Supply Chain

Huang used the keynote to underscore Taiwan’s role in the global AI buildout, calling it “the richest ecosystem, the world’s best supply chain ecosystem.” Each Vera Rubin system consists of almost two million parts, he said, with 150 different ecosystem partners in Taiwan required to assemble it.nvidia

The Vera CPU itself represents a new market opportunity, Huang argued, designed not for human users but for autonomous AI agents that “are going to be using the CPUs with very little patience.” The chip features 88 cores, 1.2TB/s of LPDDR5X bandwidth, and a 3.6TB/s on-chip fabric.nvidia

“Compute is revenue,” Huang told the crowd. “If you have 1 gigawatt of power, then throughput per watt is revenue.”nvidia