Amazon has made significant headway in the AI sector by securing a deal with Meta, which will utilise millions of Amazon’s AWS Graviton chips to bolster its AI capabilities, as announced on Friday. These Graviton chips are ARM-based central processing units (CPUs), designed for general computing tasks, contrasting with the more commonly utilised graphical processing units (GPUs) in AI model training.
Although GPUs are preferred for training large AI models, the demand for CPUs like Graviton is growing as AI agents execute complex tasks such as real-time reasoning, coding, and managing multi-step processes. Amazon claims its latest Graviton iteration is tailor-made to cater to these compute-heavy requirements.
This partnership effectively reroutes more of Meta’s financial resources back to Amazon Web Services (AWS), competing against platforms like Google Cloud. Meta previously inked a six-year contract with Google Cloud worth $10 billion but had historically leaned on AWS, along with Microsoft Azure.
Notably, Amazon’s announcement came right after the conclusion of the Google Cloud Next conference, hinting at a competitive edge in the cloud arena. While Google also develops AI chips, Amazon counters with its proprietary AI GPU known as Trainium, designed for both training and inference phases of AI.
However, a recent arrangement with Anthropic has diverted much of the Trainium stock. Anthropic consented to a $100 billion deal with AWS over ten years for AI workloads, prompting Amazon to commit an additional $5 billion investment in the company.
Through the Meta agreement, Amazon is positioning its proprietary CPUs as credible alternatives in the burgeoning AI market, challenging the likes of Nvidia’s ARM-based Vera CPU, which also targets similar workloads. A key distinction is that while Nvidia sells its chips to businesses and cloud services, AWS exclusively provides access to its chips via its cloud platform.
In his annual shareholder letter earlier this month, Amazon CEO Andy Jassy emphasised the demand for superior price-performance ratios in AI solutions, announcing his intent to secure more deals based on this merit. This highlights the increased pressure on Amazon’s internal chip development team to deliver high-performance products, amidst intensifying competition from industry giants.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
