Particle.news

Download on the App Store

Nvidia’s Next AI Platform Advances as Big Money Tilts to Hardware and Rivals Secure OpenAI Megadeals

General availability for Nvidia’s rack-scale Vera Rubin platform is targeted for late 2026.

Overview

  • Nvidia says its Vera Rubin CPU+GPU platform has taped out and is in fabrication at TSMC, with the rack-scale NVL144 CPX slated to deliver multi-exaflop performance and 100 terabytes of fast memory.
  • Recent 13F disclosures show billionaire-run funds trimming Palantir and adding Nvidia, with examples including Citadel cutting its Palantir stake by 48% and Duquesne and Bridgewater exiting positions.
  • Broadcom announced a multiyear partnership with OpenAI to co-develop accelerators targeting 10 gigawatts of deployment by 2029, while reports detail an OpenAIAMD agreement for six gigawatts of MI450 GPUs with performance-linked warrants.
  • HSBC upgraded Nvidia to Buy with a $320 target, citing a growing GPU market and potential China demand recovery, and Piper Sandler raised Palantir’s target to $201 on accelerating commercial bookings and strong contract visibility.
  • Microsoft, BlackRock and Nvidia are reported to be working on a potential $40 billion acquisition of Aligned Data Centers to speed AI buildouts, a transaction that remains subject to completion.