Snowflake just committed $6 billion to Amazon Web Services over the next five years. For context, Snowflake has earned $7 billion in total through Amazon's platform since the company was founded in 2012. This new contract nearly matches that entire 13-year figure in a single deal.
Snowflake stores and manages data for thousands of large organizations globally: manufacturers, insurers, retailers, logistics firms, and more. The companies that use it are paying more because they are now doing more with that data, specifically using AI tools that run directly inside the platform.
Snowflake's AI product, Cortex AI, has been available for roughly two years. It allows any employee to ask questions about company data in plain, everyday language, get automated summary reports, and query databases without technical skills. More recently, Snowflake expanded the tool to act as a kind of intelligent agent, one that can take actions on data automatically rather than just answering questions. Companies like Siemens Energy use it to turn paper records into searchable digital files; others use it for fraud detection, sales analysis, and compliance checks.
Customer spending on Snowflake through AWS doubled in 2025 to $2 billion, and Snowflake raised its full-year revenue forecast for fiscal 2027 to $5.84 billion after beating analyst expectations in its most recent quarter. The numbers reflect a real shift in how enterprises use the platform: from storing data to actively working with it through AI.
The specific focus of this deal is Amazon's own custom chips, called Graviton. These chips handle the ongoing, repetitive tasks that come with running AI agents continuously, not the expensive initial phase of building and training AI models, but the day-to-day operation of those agents inside a business. Amazon says Graviton chips offer better cost-efficiency than the alternatives, and passes those savings along to customers. Graviton now powers the majority of new CPU capacity added by AWS, and 98% of AWS's top 1,000 customers are already using it.
This deal follows a pattern. Meta signed its own Graviton deal with AWS last month, deploying tens of millions of chip cores to run AI across its platforms. Anthropic, which makes the Claude AI assistant, committed $100 billion to AWS over 10 years, focused on Amazon's other custom chip line. Two large unnamed customers reportedly tried to reserve all of Amazon's Graviton capacity for 2026. Both were turned down.
Amazon's custom chip business, which includes Graviton and its other chip lines, now generates over $20 billion in annualized revenue, growing fast. That is a significant business line in its own right, separate from traditional cloud storage and computing services.
For business operators, the relevant takeaway is straightforward. If your organization already uses Snowflake, or any major cloud-based data platform, the AI tools that sit on top of that data are maturing fast. The ability to ask questions of your own business data in plain language, automatically generate reports, and run agents that monitor and act on data continuously, these are no longer future capabilities. They are production-ready today, used by companies across manufacturing, finance, insurance, and logistics. The infrastructure that makes it affordable is what this deal is about.