Skip navigation
✨ The #NVIDIARubin platform is engineered to master multi-step problem-solving and massive…
0:00 / 0:54

✨ The #NVIDIARubin platform is engineered to master multi-step problem-solving and massive…

NVIDIA AI Infrastructure

@NVIDIAAIInfra85K followers

Visit on X
526 likesDownloadOriginal post

10,493 views #NVIDIARubin

✨ The #NVIDIARubin platform is engineered to master multi-step problem-solving and massive long-context reasoning workflows at scale. Built using extreme co-design, Rubin delivers more AI tokens per watt and lower cost per token compared to the previous architecture. ✔️ 50 PFLOPS of NVFP4 inference ✔️ 22 TB/s of HBM4 bandwidth ✔️ 3.6 TB/s of NVIDIA NVLink bandwidth per GPU Learn more about NVIDIA Rubin ➡️

5 Comments

Comments from the original post on X (Twitter)

  • AJ Investment Research's profile picture

    @alojoh

    Aren’t these stats now obsolete? After this 👇. I guess you’ll be announcing a similar optimisation for Rubin

    12 likes

  • Mohamed Anis's profile picture

    @Mohamedanis

    Curious how the TCO math changes when hyperscalers are already sitting on billions in Hopper inventory they need to depreciate.

    1 like

  • Tim Willis's profile picture

    @gipsytim

    Ok, it a processor.

  • SynthesisLedger's profile picture

    @SynthesisLedger

    extreme co-design tames the entropy explosion in long-context chains. rubin's token-per-watt edge makes scaled reasoning workflows economically inevitable now.🇳🇴

  • Mostofa |AI Tools & News's profile picture

    @mostofakamal00

    This is a measured and well-considered point. Focusing on core principles early often reduces friction and leads to more durable outcomes.