Amazon представляет Trainium3: Новый ИИ-чип и ультраэффективные решения для бизнеса Translation: Amazon Unveils Trainium3: New AI Chip and Ultra-Efficient Solutions for Businesses

Amazon Web Services (AWS) has unveiled the latest version of its proprietary AI chip, Trainium3, alongside announcing the development of its successor, Trainium4.

The company announced the launch of the UltraServer system, which is built on the cutting-edge 3nm Trainium3 processor and utilizes an internal networking technology.

Both innovations showed a remarkable increase in performance during the training and inference of AI, compared to the second-generation semiconductors.

The system exhibits a fourfold enhancement in performance and features four times the memory. This capability not only allows for AI training but also supports AI applications during peak demand periods.

UltraServer is composed of 144 Trainium3 chips, and systems can be combined to accommodate up to 1 million semiconductors in total.

The energy efficiency of this new solution has improved by 40%. The company reported that its clients, including Anthropic, LLM Karakuri, SplashMusic, and Decart, are already utilizing third-generation chips, which have significantly reduced their computing costs.

Amazon has outlined a roadmap for semiconductor development. The next-generation chip, Trainium4, is already in the works and promises «another significant leap» in performance, while also supporting Nvidia’s high-speed NVLink Fusion connectivity technology.

Trainium4-based systems will have the capability to scale computational power by interacting with Nvidia graphics processors, while still employing Amazon’s own, more cost-effective server rack technology.

AWS has also introduced three new AI agents. One of these can learn user preferences to autonomously operate for several days.

Each digital assistant is designed for distinct tasks:

Kiro maintains «persistent context between sessions»—its memory is virtually endless. This agent can operate for several hours or days with minimal human intervention.

Another significant launch from Amazon is AI Factories. This solution enables large corporations and government agencies to deploy AI systems directly within their data centers.

«Clients provide the energy and data center, while AWS installs the AI system, manages it, and can connect it to other cloud services offered by the company,» the firm stated.

The aim is to meet the demands of corporations and governments that seek complete control over their data.

It is worth noting that in November, Amazon requested Perplexity to remove its browser featuring an integrated AI agent from its online store.