Amazon has achieved a significant strategic victory, leveraging its proprietary chip technology, as Meta has committed to utilizing millions of AWS Graviton processors to support its expanding artificial intelligence infrastructure. This landmark agreement was disclosed by Amazon on Friday.
It is important to clarify that the AWS Graviton is an ARM-based Central Processing Unit (CPU), designed for general computing operations, rather than a Graphics Processing Unit (GPU).
While Graphics Processing Units (GPUs) are predominantly favored for the intensive task of training large AI models, a discernible shift is occurring in the required chip architecture once these models are developed. AI agents, built upon these trained models, generate demanding computational workloads such as real-time reasoning, code generation, search functionalities, and the intricate coordination of multi-step tasks. Amazon Web Services (AWS) explicitly states that its newest Graviton iteration was engineered precisely to address these specific AI-centric computational demands.
This agreement effectively diverts a substantial portion of Meta's expenditure towards AWS, rather than rival cloud providers such as Google Cloud. It is noteworthy that Meta had previously entered into a six-year, $10 billion partnership with Google Cloud last August, despite having primarily relied on AWS and also utilizing Microsoft Azure prior to that commitment.
The timing of AWS's announcement, coinciding precisely with the conclusion of the Google Cloud Next conference, was not overlooked, appearing to be a calculated move aimed at its cloud competitor. Google, for its part, also develops proprietary AI chips and unveiled updated versions during its recent event.
It is also true that Amazon produces its own AI-focused Graphics Processing Unit, named Trainium. Despite its moniker, Trainium is utilized for both the training phase of AI models and the subsequent inference stage, where a trained model actively processes new prompts and data.
However, Anthropic had already secured a significant agreement, publicized earlier this month, which effectively reserved a substantial portion of these Trainium chips for several years. The creator of the Claude AI model committed to investing $100 billion over a decade to operate its workloads on AWS, with a specific emphasis on Trainium. In a reciprocal arrangement, Amazon pledged an additional $5 billion investment into Anthropic, bringing its total investment to $13 billion.
Fundamentally, the Meta collaboration provides Amazon with a prominent AI client, serving as a powerful validation for its proprietary CPUs. These processors are positioned as direct competitors to Nvidia's new Vera CPU, which is also ARM-based and engineered for AI agentic workloads. A key distinction lies in Nvidia's business model, where it sells its chips and AI systems directly to enterprises and other cloud providers (including AWS), whereas AWS exclusively offers access to its chips via its cloud services.
Earlier this month, Amazon CEO Andy Jassy, in his annual shareholder letter, specifically targeted Nvidia and Intel, asserting that businesses are seeking superior price-performance ratios for AI solutions and that Amazon aims to secure deals by meeting this demand. This strategic focus places immense pressure on Amazon's internal chip development team to perform, a team whose laboratories we had the exclusive opportunity to tour last month.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.