Snowflake puts Anthropic’s Claude inside enterprise data as $6 Billion AWS deal looms

Snowflake used its Summit 26 event to highlight growing enterprise adoption of Anthropic’s Claude inside its Cortex AI platform, and the timing carried more weight than a typical partner update.

Just days earlier, Snowflake disclosed a $6 billion multi-year infrastructure commitment to AWS, which reframed the Anthropic news considerably. One announcement was about model access. The other was about the cost base and cloud capacity needed to run AI agents at production scale.

Together, the two moves tell a fairly direct story about how enterprise AI is being packaged right now. Model access, data governance, cloud infrastructure, and marketplace procurement are being arranged into a more controlled buying path, convenient for customers who want fewer integration headaches, and deeply profitable for the platforms building that path.

The practical case for Claude inside Cortex AI is straightforward. Enterprises want AI systems that work directly on their existing business data without copying it into a separate unmanaged environment. Claude handles reasoning. Snowflake provides the data platform, governance controls, and security layer. AWS supplies the underlying infrastructure for most of Snowflake’s customer base. The result is less a product than a stack, and opting into it means opting into all three vendors simultaneously.

Companies including Block, eSentire, Indeed, Notion, and Deloitte appear across Snowflake’s customer examples, using the setup for compliance investigations, threat analysis, self-service analytics, and sales workflows. These are not experimental use cases. They involve compliance data, security incidents, and operational systems where governance is not optional.

That is precisely where the harder questions sit. Who owns an agent once it runs inside a production workflow? How do permissions carry across data boundaries? Can a business user indirectly surface restricted information through a well-worded prompt? Snowflake’s governance architecture gives it a more credible answer than most, but credible is not the same as complete.

On the infrastructure side, the AWS commitment reflects real demand. AI workloads are expensive to sustain, and once they embed into sales, compliance, and developer workflows, usage becomes recurring and the costs follow accordingly. Snowflake’s expansion into ten new AWS regions, including the AWS European Sovereign Cloud, also signals that data residency and jurisdictional compliance are shaping where these workloads run, not just how.

The trade-off for customers is familiar: fewer moving parts, but also fewer clean exits.