🏗️ AI Infrastructure

NVIDIA Unveils Vera Rubin AI Factory Platform at GTC 2026 — Dedicated Inference Chips, 35x Performance Boost, Open Model Alliance

2 min read1 views

At GTC 2026 (March 16-19, San Jose), NVIDIA CEO Jensen Huang unveiled a comprehensive expansion of the Vera Rubin AI factory platform, marking a strategic pivot from training-focused to inference-focused AI infrastructure — directly targeting the agentic AI workload explosion.

Key Announcements:

  1. Groq 3 LPX — Dedicated Inference Hardware: For the first time, NVIDIA is adding dedicated inference chips to its platform. The new Groq 3 LPX (Logical Processing Unit) claims a 35x inference performance boost. This is a significant architectural shift — NVIDIA previously relied on GPUs for both training and inference, but the scale of agentic AI inference demands purpose-built silicon.

  2. Vera Rubin Platform Expansion:

  • 7-chip AI factory platform architecture
  • Custom CPU racks optimized for inference workloads
  • New storage architecture for agent state management
  • Inference operating system for orchestrating agent workloads
  1. DGX Station with GB300 Superchips: First DGX Station systems powered by GB300 superchips have shipped to pioneering developers, enabling deskside development with frontier-scale models.

  2. Open Model Alliance: NVIDIA launched a multi-lab open-source model coalition, partnering with external research labs to create shared model infrastructure.

  3. Broader Ecosystem:

  • Cisco expanding Secure AI Factory strategy across data center, telco edge, and enterprise
  • Hitachi Vantara expanding Hitachi iQ for responsible agentic AI
  • Multiple partners showcasing agentic AI infrastructure at GTC booths

Jensen Huang's Five Arguments for Continued AI Build-Out: Huang positioned AI infrastructure spending as essential, noting that prospective employees in Silicon Valley are already asking how many tokens come with a job offer — compute access as a talent signal.

The conference agenda has decisively shifted from model training to practical applications: inference, autonomous AI agents, and infrastructure capable of serving them in real time.

Analysts estimate the AI chip market is approaching $1 trillion, with GTC 2026 resetting expectations for infrastructure investment.

Share this article

🧠 Stay Updated on AI Agents

Get weekly insights on agentic AI, networks and infrastructure. No spam.

Join 500+ AI builders. Unsubscribe anytime.

Deploy Your AI Agent Today

Launch a managed OpenClaw instance in minutes

Request demo →

More from AI Infrastructure