News
Nvidia Vera Rubin GPU architecture to feature 288 GB HBM4 and 22 TB/s bandwidth
Wednesday, February 25, 2026 at 10:45 PM
Nvidia has detailed its Vera Rubin GPU architecture, featuring 288 GB of HBM4 memory and a bandwidth of 22 TB/s. The platform aims to deliver 50 PFLOPs of AI compute, representing a significant jump in performance and memory capacity for data center infrastructure.
Context
At GTC 2026, Nvidia officially unveiled the technical specifications for its next-generation Vera Rubin GPU architecture, confirming a massive leap in memory performance. The flagship Rubin GPU will feature 288 GB of HBM4 memory and an unprecedented 22 TB/s of bandwidth, representing a 2.8x increase over the current Blackwell generation. This performance jump is designed to overcome the "memory wall" for agentic AI and trillion-parameter models, delivering 50 petaflops of NVFP4 inference horsepower.
Strategic partner Micron has already entered high-volume production of the HBM4 36GB 12H modules required for the platform. The architecture is built on a 3nm-class process and integrates the new 88-core Vera CPU, which uses NVLink 6 to provide a unified memory pool between processors. With shipments expected in the second half of 2026, the platform aims to reduce inference costs by up to 10x, further solidifying Nvidia's dominance in the AI supply chain.
Sources (10)
NVIDIA Vera Rubin Opens Agentic AI Frontier | NVIDIA NewsroomMicron in High-Volume Production of HBM4 Designed for NVIDIA Vera Rubin, PCIe Gen6 SSD and SOCAMM2 - Investor RelationsNVIDIA Vera CPU Delivers High Performance, Bandwidth, and Efficiency for AI Factories | NVIDIA Technical BlogMicrosoft’s strategic AI datacenter planning enables seamless, large-scale NVIDIA Rubin deployments | Microsoft Azure BlogNvidia delivers first Vera Rubin AI GPU samples to customers — 88-core Vera CPU paired with Rubin GPUs with 288 GB of HBM4 memory apiece | Tom's HardwareNvidia’s Vera-Rubin Platform Obsoletes Current AI Iron Six Months Ahead Of Launchtimes-online.com - NVIDIA Rubin Architecture Triggers HBM4 ... - UserNVIDIA Rubin Architecture Deep Dive: The End of Blackwell?
Related Companies
Nvidia
NVDA