Trainium2, Graviton4 and Q: Amazon’s Three AI Chips That Promise Performance and Innovation

Amazon Web Services (AWS) has recently announced the launch of its latest generation of AI chips during the re:Invent conference in Las Vegas. These new chips represent a significant step forward for the company in the field of artificial intelligence and reveal its long-term ambitions in the industry.

AWS Trainium2: Improved Performance and Greater Energy Efficiency

Among the novelties presented by AWS, the chip Trainium2 stands out, designed specifically for the training of AI models. This chip offers performances up to 4 times superior and double energy efficiency compared to its previous version. Thanks to these characteristics, programmers can train models more quickly and economically, thanks to a reduction in energy consumption. Anthropic, a competitor of OpenAI supported by Amazon, has already announced its intentions to use Trainium2 chips to develop its own models.

Graviton4: Improved General Performance and Lower Cloud Computing Costs

The Graviton4 chip, meanwhile, has been designed for more general use. These processors, based on the Arm architecture, consume less energy compared to Intel or AMD chips. Amazon promises a 30% increase in general performance when using an AI model trained on a Graviton4 processor. This allows organizations that regularly deploy AI models to process larger data sets, reduce their labor costs, improve execution times, and cut comprehensive costs. The Graviton4 chip is available in preview starting today, with a full release planned in the coming months.

Collaboration with NVIDIA: Cloud Access to the Latest AI GPUs

Instead of abandoning third-party suppliers like NVIDIA, a leading player in the enterprise AI field, Amazon has chosen to strengthen the collaboration by offering enterprise clients cloud access to NVIDIA’s latest AI GPUs, called H200. Additionally, Amazon makes available over 16,000 NVIDIA GH200 Grace Hopper super chips for NVIDIA’s research and development team. This approach is similar to that adopted by Amazon’s principal rival in the AI sector, Microsoft, which announced a significant partnership with NVIDIA at the same time as it introduced its own proprietary AI chip, May 100.

Amazon Q: An AI Assistant Focused on Business

Beyond AI chips, Amazon has also announced a new AI assistant focused on business called Q. This assistant is described as a ‘new type of generative personal assistant based on artificial intelligence,’ and is specifically designed to simplify work projects and client assistance activities. Q can be personalized to adapt to any type of business and offers relevant answers to more common questions. Moreover, it can autonomously generate content and undertake actions based on client requests.

Q will be available on communication platforms like Slack and in commonly used text editing applications by programmers. Additionally, it will be capable of integrating with over 40 other business systems, including Microsoft 365, Dropbox, Salesforce, and Zendesk. Amazon Q is currently available in preview, with a full release expected soon. The cost varies from $20 to $30 per user per month, depending on the features available.

Amazon’s Big Bet on AI

From the recent announcements by Amazon, it is apparent that the company is heavily investing in artificial intelligence, much like many other industry players. Particularly, it is seeking to become the reference point for enterprise AI, challenging its old rival in the cloud computing field, Microsoft. Amazon is leveraging AI to maintain its dominant position in the cloud computing sector, aiming to limit the market share growth of Microsoft and other competitors such as Google and Alibaba.