Nvidia’s Vera Rubin Aims to Outrun AI’s Self-Built Chip Threat—But Is It Too Late?

Nvidia’s Vera Rubin AI superchip, set to slash costs and efficiency barriers in AI, arrives as the industry grapples with self-built silicon by rivals.

Nvidia CEO Jensen Huang announced Vera Rubin is in 'full production' at CES, with delivery scheduled for late 2024.

"Today, I can tell you that Vera Rubin is in full production," said Huang.

The chip reduces AI model operational costs by 90% compared to Blackwell and requires 25% fewer chips for training. Partners Microsoft and CoreWeave will launch Rubin-powered services this year, with Microsoft's Georgia and Wisconsin data centers incorporating thousands of Rubin chips.

Austin Lyons of Chipstrat noted:

"This CES announcement around Rubin is to tell investors, ‘We’re on track.’"

The system includes six chips (Rubin GPU, Vera CPU) built on TSMC's 3nm process and advanced memory tech, with Nvidia's sixth-gen interconnects.

"Nvidia is evolving into a full AI system architect, spanning compute, networking, memory hierarchy, storage, and software orchestration," says Lyons.