Nvidia plans to launch high-efficiency AI agent semiconductors in late 2026
News

Nvidia plans to launch high-efficiency AI agent semiconductors in late 2026

Monday, March 16, 2026 at 09:00 PM

Nvidia has announced its plans to launch a new generation of AI semiconductors in the second half of 2026. These chips are specifically designed to support autonomous AI agents, offering up to a 35-fold increase in power efficiency compared to current models. This development aims to address the massive data processing and energy demands required for autonomous AI infrastructure.

Context

At the GTC conference on March 16, 2026, NVIDIA announced that its next-generation Vera Rubin platform is moving into full production, with a scheduled release for the second half of 2026. The platform is specifically engineered to handle the massive data and power demands of Agentic AI, which involves autonomous reasoning and multi-step task execution. Key specifications for the new architecture include a massive 35x improvement in power efficiency for specific workloads and up to 10x more performance per watt compared to the previous Blackwell generation. The rollout features seven new chips, including the Vera CPU and Rubin GPU, designed to operate as a unified AI supercomputer. This shift to a one-year product cadence underscores NVIDIA’s strategy to dominate the infrastructure layer as AI moves from simple chat interfaces to complex autonomous agents. During the announcement, CEO Jensen Huang stated, "The agentic AI inflection point has arrived with Vera Rubin kicking off the greatest infrastructure buildout in history." Major cloud providers including AWS, Google Cloud, Microsoft, and Meta have already committed to deploying Vera Rubin-based instances starting in late 2026 to support their long-term AI roadmaps.

Related Companies

Nvidia
Nvidia
NVDA
US