AWS launches Trainium3 chips to break the AI compute bottleneck
The new UltraServer promises 4x performance and 40% energy efficiency gains over previous hardware.
Vertical integration targeting efficiency
AWS escalated the silicon wars at re:Invent on Wednesday with the formal launch of the Trainium3 UltraServer. The new custom silicon stack is designed specifically for training massive AI models, directly challenging Nvidia’s dominance in the data center. AWS claims the Trainium3 chip delivers a 4x performance increase over its predecessor while achieving a 40% reduction in energy consumption.
This focus on energy efficiency addresses a critical constraint in the AI sector: power availability. As data centers struggle to secure sufficient electricity grid connections, the performance-per-watt metric has become as important as raw speed.
Archive Access
This article is older than 24 hours. Create a free account to access our 7-day archive.
Archive Access
This article is older than 24 hours. Create a free account to access our 7-day archive.
Share this article
Related Articles
Adobe Partners with Runway to Bring Gen-4.5 Video to Firefly
Adobe partnered with Runway to integrate the new Gen-4.5 video model into Firefly, combining professional …
Amazon Reportedly in Talks to Invest Ten Billion Dollars in OpenAI
Reports on Dec. 17 indicate Amazon is negotiating a $10 billion investment in OpenAI, potentially …
Tech Giants Launch Agentic AI Foundation to Standardize Workflows
The Agentic AI Foundation formed on Dec. 16 with contributions from OpenAI and Anthropic to …