In a massive boost for Amazon’s custom silicon ambitions, data cloud provider Snowflake has just announced a staggering $6 billion, five-year infrastructure agreement with Amazon Web Services. This isn’t an experimental pilot program. It is Snowflake’s largest cloud commitment ever, and it signals a fundamental shift in how the enterprise world plans to execute everyday artificial intelligence.
The core of this massive $6 billion deal revolves heavily around AWS’s homegrown Graviton ARM-based CPU instances, alongside standard GPU setups. The infrastructure will power Snowflake’s Cortex AI platform, a system designed to help enterprise clients build and deploy intelligent software agents that can reason, coordinate tasks, and actually get work done.
The Move Toward Agentic AI
The industry is rapidly realizing that while you need expensive, high-powered GPUs to train massive foundational models, you don’t necessarily need them for the day-to-day execution of business logic. Snowflake CEO Sridhar Ramaswamy made it clear that this deal is about moving beyond simple chatbots and into the era of the “agentic enterprise.”
These agents require cost-effective, scalable computing power to coordinate internal systems and drive tangible business outcomes. By leaning heavily on AWS Graviton chips, Snowflake is proving that CPU-bound inference is becoming the dominant engine for this kind of operational AI. The commitment, which officially scales Snowflake’s lifetime AWS Marketplace sales past the $7 billion mark, was announced alongside a stellar fiscal first-quarter earnings report.
Snowflake reported $1.39 billion in revenue, up 33% year-over-year, sending its stock surging over 33% in extended trading. To put this explosive growth in perspective, Snowflake’s initial AWS commitment at the time of its 2020 IPO was $1.2 billion. It jumped to $2.5 billion in 2023 before making this massive $6 billion leap in 2026.
The global footprint is expanding alongside the raw compute power. As part of the pact, Snowflake is rolling out into 10 new global AWS regions, including the highly regulated AWS European Sovereign Cloud, giving enterprise clients the ability to navigate strict data residency laws while deploying complex AI agents.
What the Graviton Shift Means for Nvidia’s Dominance
This $6 billion agreement mounts serious, tangible competitive pressure on Nvidia. For years, Nvidia has controlled the AI narrative with its dominant GPUs. But as cloud providers like Amazon prove that their homegrown Graviton CPUs can successfully handle the bulk of AI inference and agentic workloads at a significantly lower cost, the landscape changes.
This directly challenges Nvidia CEO Jensen Huang’s aggressive push to dominate the new $200 billion market for enterprise inference with Nvidia’s Vera CPU. The Snowflake deal proves that enterprise software giants are finding AWS’s in-house silicon to be both capable and incredibly cost-effective at scale.
This isn’t an isolated incident, either. Meta struck a remarkably similar large-scale Graviton deal with AWS just last month. The market is witnessing a critical inflection point where the sheer volume of everyday AI execution is shifting away from specialized GPUs and toward optimized, cloud-native CPU architectures.