Enterprise AI rarely fails because a company lacks a good demo. It fails because the data, compute, procurement, and governance behind the demo live in four different systems never built to talk to each other.
Snowflake’s newest commitment to AWS is a direct bet against the failure mode just described. On May 27, 2026, Snowflake announced a multi-year strategic collaboration agreement with Amazon Web Services, including a $6 billion commitment over five years for Graviton compute and AI spend. CEO Sridhar Ramaswamy framed the deal as an accelerant for enterprise agentic AI adoption, not as a marketing exercise tied to a single product launch.
What the Deal Includes
The agreement covers more than raw compute spend. Snowflake and AWS described deeper product integrations across generative and agentic AI, expanded go-to-market through AWS Marketplace, joint investment in customer success programs, and coordinated workload migrations. Named customers Fetch and Hex appeared in the companies’ release as early examples of the collaboration in practice.
The dollar figure has grown steadily alongside Snowflake’s relationship with AWS. The company’s five-year AWS spending commitment moved from $1.2 billion at the time of its 2020 IPO, to $2.5 billion in 2023, to $6 billion now. Snowflake said it has surpassed $7 billion in lifetime AWS Marketplace sales and exceeded $2 billion in calendar-year Marketplace sales in 2025, doubling the prior year’s figure. Most Snowflake customers already run on AWS, by the company’s account, and the new agreement deepens an infrastructure relationship eleven years in the making rather than starting one from scratch.
Snowflake’s SEC filings show the company runs infrastructure across AWS, Azure, and Google Cloud. The filings note its cloud agreements generally can include minimum usage commitments, fixed and variable pricing provisions, and payment obligations triggered if a commitment goes unmet, but they describe the category in general terms and do not disclose the specific terms behind the new $6 billion AWS commitment. Snowflake has not said what share of total infrastructure spend the $6 billion represents, and the company’s most recently disclosed non-cancelable purchase commitments predate the new agreement, leaving no clean comparison point.
Why Agentic AI Needs the Data Layer
Agentic AI sounds like a model problem from the outside. In practice, it is a data-access problem first. An agent needs governed context before it can act on anything, the same customer records, financial tables, and operational data a human analyst would need, accessed under the same compliance rules.
Snowflake’s pitch centers on bringing AI to governed enterprise data, rather than moving sensitive data out to wherever a model happens to run. The company points to Cortex AI capabilities, including text-to-SQL, summarization, sentiment analysis, and entity extraction, as the connective tissue between raw enterprise data and the agents meant to act on it. AWS supplies the compute layer underneath: Graviton processors for general workloads and GPU-backed EC2 instances for model training and inference.
The framing matters more than the marketing copy suggests. The companies winning the next phase of enterprise AI adoption will likely be the ones solving access and governance first, not the ones with the largest model.
The Multi-Cloud Question
Snowflake has spent years marketing itself as cloud-neutral, a data platform running cleanly across AWS, Azure, and Google Cloud. A $6 billion, five-year AWS commitment does not reverse the positioning. Snowflake has made no public statement about scaling back support for other clouds, and nothing in the announcement claims exclusivity.
A deeper AWS relationship still raises a real question for customers standardized on Azure or Google Cloud, however. If the most advanced Cortex AI integrations, the fastest workload migrations, and the deepest joint engineering land on AWS first, multi-cloud optionality becomes more of a theoretical feature and less of a practical one for non-AWS customers. The tension is worth watching, not dismissing, over the next several product cycles.
Competitive Context
Place the deal against the rest of the field and the pattern sharpens. Databricks maintains cloud partnerships spanning multiple providers. Google pairs BigQuery with Vertex AI as a competing data-plus-model bundle. Microsoft offers Fabric and Azure AI as the default stack for enterprises already standardized on Microsoft tools. AWS counters with Redshift, Bedrock, and SageMaker as its native alternative to a Snowflake-on-AWS setup.
The enterprise AI stack is consolidating around data, compute, and model access sitting closer together, not further apart. Snowflake and AWS are positioning the consolidation to favor each company, rather than forcing customers to choose one over the other.
What to Watch Next
Snowflake’s Q1 fiscal 2027 results give the deal some financial backing: revenue of $1.39 billion and product revenue of $1.33 billion, up 34 percent year over year, with more than 13,900 customers now using Snowflake’s AI Data Cloud, according to the company. None of it confirms agentic AI has reached broad production maturity inside most enterprises. It confirms vendor investment and expected demand, which is a different claim and a smaller one.
The real test arrives over the next several quarters: whether AWS Marketplace sales keep climbing, whether customers migrate AI workloads onto Graviton-backed Snowflake infrastructure at scale, and whether Snowflake can keep its Azure and Google Cloud customers comfortable while AWS gets the deepest integration work first. The $6 billion number is a headline. Where the engineering hours go next is the story.
