Meta signed a deal with Amazon Web Services to use millions of AWS Graviton ARM-based CPUs for AI agentic workloads, marking a shift in the AI chip race toward specialized processors beyond traditional GPUs. The deal brings Meta's spending back to AWS after the company signed a six-year, $10 billion agreement with Google Cloud in August. AWS strategically announced the Meta partnership as Google Cloud Next conference concluded, with Amazon CEO Andy Jassy positioning AWS's custom chips as offering better price-performance ratios than Nvidia and Intel. The move reflects growing demand for CPUs optimized for AI agents that handle compute-intensive tasks like real-time reasoning and multi-step coordination, while GPUs remain dominant for training large AI models.
Meta's Multi-Million Chip Deal with AWS for AI Agent Workloads
- ▪Amazon announced Meta's deal to use millions of AWS Graviton chips on Friday
- ▪Meta signed a deal to use millions of AWS Graviton chips to power Meta's growing AI needs
- ▪Meta signed a six year, $10 billion deal with Google Cloud in August of the previous year
- ▪AWS Graviton is an ARM-based CPU that handles general computing tasks
- ▪Meta had primarily been an AWS customer that also used Microsoft Azure until the Google Cloud deal
- ▪The Meta deal with AWS brings more of Meta's cash back to AWS instead of competitors like Google Cloud
Shift from GPUs to CPUs for AI Agent Computing Needs
- ▪Nvidia's new Vera CPU is ARM-based and designed to handle AI agentic workloads
- ▪AI agents built on top of trained models are causing a shift in the type of chip needed
- ▪GPUs remain the chip of choice for training large AI models
- ▪AI agents create compute-intensive workloads like real-time reasoning, writing code, search, and coordination involved in managing agents through multi-step tasks
- ▪AWS's latest version of Graviton was designed specifically to handle AI-related compute needs
Amazon's Competitive Strategy Against Nvidia and Cloud Rivals
- ▪Amazon CEO Andy Jassy took aim at Nvidia and Intel in Andy Jassy's annual shareholder letter earlier in the month
- ▪Amazon agreed to invest another $5 billion into Anthropic, bringing Amazon's total investment to $13 billion
- ▪Anthropic agreed to spend $100 billion over 10 years to run Anthropic's workloads on AWS with a particular focus on Trainium
- ▪Andy Jassy stated that enterprises want better price-performance ratios for AI and that Andy Jassy intends to win deals on that basis
- ▪Amazon makes its own AI GPU called Trainium, which is used for both training and inference
- ▪Nvidia sells its chips and AI systems to enterprises and cloud providers including AWS
- ▪AWS only sells access to AWS chips through AWS's cloud service
Perspective of Amazon/AWS
- ▪AWS timed the Meta deal announcement strategically to coincide with the conclusion of the Google Cloud Next conference
Perspective of Meta
- ▪Meta is pursuing a multi-cloud strategy by maintaining relationships with AWS, Google Cloud, and Microsoft Azure simultaneously
Story comments
Loading comments…