Snowflake committed $6 billion to Amazon Web Services over five years in a partnership expansion announced this week, the largest cloud infrastructure agreement disclosed by either party since AWS began reporting select customer contracts in 2021. The deal runs through 2030 and centers on compute capacity for Snowflake's AI and machine learning workloads, not storage or basic data warehousing.
The announcement arrived without drama but carries weight. Snowflake already runs its multi-cloud platform on AWS, Azure, and Google Cloud. This agreement formalizes what was previously consumption-based spend into a fixed commitment at an undisclosed discount, locking Snowflake to AWS infrastructure as it scales Snowpark for generative AI and Cortex, its LLM inference layer. The $6 billion figure represents roughly 40% of Snowflake's trailing-twelve-month revenue, signaling the company expects AI workloads to drive the next phase of margin expansion rather than compete it away.
The move matters for two reasons. First, it clarifies who captures margin in the AI stack. Snowflake is betting it can sell AI-native data products at higher gross margins than raw compute, effectively arbitraging AWS's scale. If Cortex and Snowpark adoption tracks enterprise LLM usage forecasts—Gartner pegs 65% of enterprises deploying generative models by Q4 2025—Snowflake's $1.2 billion annual AWS run rate becomes a cost-of-goods lever, not an expense problem. Second, the deal signals AWS is willing to offer aggressive volume pricing to lock in consumption from infrastructure-heavy SaaS platforms. Snowflake's commitment creates a de facto exclusivity window for AWS in the data workload layer, even as Snowflake maintains technical multi-cloud posture.
The timing aligns with Snowflake's pivot from pure data warehousing into AI orchestration. The company reported $900 million in product revenue last quarter, with AI and ML workloads accounting for 18% of compute consumption, up from 11% a year prior. AWS gains predictable revenue and deeper embedding into Snowflake's architecture, which processes queries for 9,800 enterprise customers. For AWS, the deal is a hedge: if Snowflake's AI layer wins, AWS captures the compute spend regardless of which LLM customers deploy.
Watch Snowflake's gross margin trajectory over the next three quarters. If the $6 billion commitment comes with tiered discounts tied to volume thresholds, margin compression could surface in fiscal Q2 2026 earnings, expected late May. Also monitor whether Google Cloud or Azure respond with competing offers to Databricks or Palantir, the two closest analogs in the AI-native data platform category. Those deals would clarify whether hyperscalers are entering a subsidy cycle for platform partners or if this pricing is unique to Snowflake's scale.
The $6 billion commitment is less a partnership and more a re-rating of infrastructure as a fixed cost in the age of inference-heavy workloads. Snowflake is buying certainty. AWS is buying time.