Challenge to Nvidia: The New Era of AI Agents is Changing Hardware Requirements Tech giant Amazon has scored a significant strategic victory over its cloud rivals.
Meta (the parent company of Facebook and Instagram) has signed a deal to use millions of AWS Graviton processors to power its rapidly growing AI infrastructure. Why Graviton? The Shift from GPUs to CPUs for AI Until now, the focus has been on GPUs (like Nvidia’s) for training large models. However, once the models are trained, the second stage is inference (executing queries). This is where AI Agents come into play.
Real-time reasoning: AI agents perform complex tasks like writing code, advanced search, and coordinating multi-step tasks. Optimization: The latest version of Graviton (ARM architecture) is specifically designed to handle these workloads with a much better cost-performance ratio than traditional processors. Cloud Battle: AWS Regains Ground Against Google This deal is a direct blow to Google Cloud, with which Meta signed a $10 billion contract last year.
By choosing Graviton, Meta is turning to the Amazon ecosystem, proving that “homegrown” chips have become a key competitive advantage. Competition: Amazon vs. Nvidia and Intel Amazon CEO Andy Jassy has been clear: enterprises want high performance at lower prices. Nvidia Vera: Amazon is now competing directly with Nvidia’s new CPU, Vera, which also targets AI agent workloads.
Exclusivity: Unlike Nvidia, which sells its chips to everyone, Amazon offers access to Graviton and Trainium only through the AWS platform, creating a technological “fence” that is difficult to break. Market Analysis: With Anthropic securing Trainium chip capacity in a $100 billion deal, and Meta choosing Graviton,
Amazon is positioning itself as the undisputed leader of future AI infrastructure, where computational efficiency is as important as raw power.
