Snowflake and AWS Reach $6 Billion Agreement for AI Infrastructure
The five-year deal focuses on Amazon's custom Graviton CPUs to power Snowflake’s Cortex AI and automated agents.
Primary source: TechCrunch AI. Full source links and update notes are below.
Fast summary
Start here
- Snowflake committed $6 billion over five years to AWS, nearly matching its total historical spending with the provider since 2012.
- The agreement emphasizes access to AWS's ARM-based Graviton CPUs, which are optimized for AI inference and automated agents.
- Snowflake’s customers doubled their spending on AWS-based services to $2 billion in 2025, driven by demand for generative AI tools.
What happened
Snowflake, the cloud data storage leader, signed a new $6 billion, five-year agreement with Amazon Web Services (AWS) on Wednesday. The contract represents a massive escalation in the partnership, as Snowflake has generated a total of $7 billion in revenue via the AWS Marketplace since its 2012 inception. This new commitment signals that Snowflake is doubling down on AWS as its primary infrastructure provider for its next generation of data services.
What's new in this update
The core of this new deal is Snowflake's need for AWS’s proprietary ARM-based Graviton CPUs. While Nvidia's GPUs remain the standard for training massive models, Snowflake is pivoting toward CPUs to power its Cortex AI tool and automated agents. As AI moves from the training phase to daily reasoning and automation, CPU usage for agents is expected to skyrocket, making specialized silicon a necessity for cost-effective scaling.
Key details
The expansion is fueled by a surge in enterprise AI adoption. In 2025 alone, Snowflake customers spent $2 billion on AWS-hosted services, a 100% increase from the previous year. Amazon CEO Andy Jassy has positioned the homegrown Graviton chips as offering better price-performance than standard offerings, passing these savings to customers who utilize tools like Cortex AI for text interfaces and summary reports.
Background and context
Snowflake has historically run its operations on AWS, though it also offers services through Microsoft Azure and Google Cloud. This deal follows a similar multi-billion-dollar agreement between AWS and Meta for Graviton chips. These partnerships serve as a competitive signal to Nvidia, as cloud giants increasingly attempt to capture the hardware market for inference and automation tasks that do not strictly require heavy GPU training.
What to watch next
Competition in the AI chip market is intensifying as Google and Microsoft continue to deploy their own silicon, such as TPUs and the Maia chip respectively. Meanwhile, Nvidia is preparing to defend its market share with its new Vera CPU, which CEO Jensen Huang anticipates will address a brand new $200 billion market. The industry will be monitoring whether these custom cloud chips can successfully erode Nvidia’s dominance in the enterprise AI sector.
Why it matters
This deal underscores the growing importance of custom-built cloud chips as alternatives to Nvidia GPUs for the inference and automation phase of AI development.
Read next
Follow this story through the topic hub, more ai coverage, and the latest updates.
Weekly briefing
Get the week's key developments in one concise email.
Get a fast catch-up on the biggest stories, the context behind them, and the links worth your time.
Cadence
Weekly, for a quick catch-up
Coverage
AI, business, world, security, sports
Format
Clear takeaways and useful context
Request the briefing
Leave your email to open a prepared request and get on the list for the weekly briefing.
Author
See who assembled this story and follow more of their work.
Sources and methodology