Nvidia is launching a brand new Vera Rubin computing platform in early 2026. It is going to embrace six chips that can kind one AI supercomputer and can present 5 occasions extra computing assets for AI coaching than Blackwell.
Nvidia is beginning 2026 with an early launch of its new Vera Rubin computing platform, after a report 12 months for Rubin's GPU predecessor, Blackwell, fueled by an AI increase (or "bubble"), The Verge experiences, writes UNN.
Particulars
Throughout a press briefing forward of his CES 2026 keynote, Dion Harris, Nvidia's Senior Director of Excessive-Efficiency Computing and AI Infrastructure Options, described Vera Rubin as "six chips that make up one AI supercomputer."
These six chips embrace the Vera CPU, Rubin GPU, sixth era NVLink change, Join-X9 community card, BlueField4 DPU, and Spectrum-X 102.4T CPO. The platform will help third era confidential computing and, in response to Nvidia, would be the first dependable rack-scale computing platform.
Nvidia claims that the Rubin GPU is able to offering 5 occasions extra computing assets for AI coaching than Blackwell. The Vera Rubin structure can typically prepare a big AI mannequin utilizing a "combination of specialists" (MOE) construction in the identical period of time as Blackwell, whereas utilizing 4 occasions fewer GPUs and at one-seventh the fee per token.
Rubin's launch was initially anticipated later this 12 months. Its earlier launch comes simply a few months after Nvidia reported record-high information heart revenues, which grew 66 p.c year-over-year. This development is pushed by demand for Blackwell and Blackwell Extremely GPUs, which set a excessive bar for Rubin's success and served as an indicator of the "AI bubble." Services and products powered by Rubin can be accessible from Nvidia companions beginning within the second half of 2026.
Nvidia information report income of $57 billion and reassures about AI 'bubble'20.11.25, 11:59 • 3486 views
Leave a Reply