Snowflake’s founders built the company on Amazon Web Services in 2015, and on Wednesday the data-cloud company deepened that bet by an order of magnitude. Snowflake said it signed a multi-year strategic collaboration agreement with AWS that includes a $6 billion infrastructure commitment — its largest ever — as enterprises move AI workloads from experiments to production.
The money is tied to specific silicon. Snowflake will spend on AWS’s Graviton processors, the Arm-based chips Amazon designs for general-purpose computing, and on cloud GPUs for AI training and inference, according to the companies. The five-year commitment implies average annual spending of $1.2 billion, roughly triple the pace of Snowflake’s previous arrangement with the cloud giant.
It is the third time Snowflake has expanded its AWS commitment. At its 2020 IPO, Snowflake disclosed a five-year, $1.2 billion deal with an unnamed cloud provider — which turned out to be AWS — with $350 million due in the final year. In 2023, the figure climbed to $2.5 billion. Wednesday’s agreement more than doubles it, and it arrives as Snowflake’s business re-accelerates.
The announcement landed alongside a blowout earnings report. Snowflake shares jumped 36% on May 27 after the company beat expectations and raised guidance, pushing its market value to just over $60 billion. The stock move and the AWS deal are two sides of the same story: customers are routing more data through Snowflake’s platform and paying for AI features like Cortex, its suite of machine-learning functions, and the agentic workloads that chain together models, tools, and data.
The distribution machine is AWS Marketplace. Snowflake said lifetime sales through the marketplace surpassed $7 billion, including more than $2 billion in 2025, more than double the prior year. The majority of Snowflake’s customers run on AWS today, the company said, and the new agreement expands go-to-market through the marketplace with simplified contracting and faster procurement.
Customers are beginning to use the AI capabilities in production. Snowflake cited Fetch, the rewards company, and Hexare among enterprises deploying AI applications on governed data with Snowflake on AWS. The pitch is simple: keep the data where it is governed and secure, and bring AI to it, rather than copying data into AI platforms. “We’re making it easier for enterprises to bring AI directly to governed data,” Snowflake Chief Executive Sridhar Ramaswamy said.
The deal fits a pattern across the industry. Cloud providers are converting AI ambitions into contractual commitments — OpenAI has agreed to spend tens of billions of dollars on Microsoft’s Azure, and Anthropic has similar arrangements with Google and Amazon. Snowflake’s agreement lacks the equity cross-investments those deals carry, but the commitment structure is the same: long-term, silicon-specific, and large. AWS said the Snowflake arrangement is part of a renewed flurry of such commitments as enterprises move from AI pilots to production-scale deployments.
Ramaswamy, a former Google executive who took over Snowflake in early 2024, has made agentic AI the company’s strategic focus. The AWS agreement accelerates that push with deeper product integrations across generative and agentic AI, joint investments in customer-success programs and workload migrations, and strategic industry solutions. The deal also extends Snowflake’s global footprint: launches completed or underway in ten new regions, including Auckland, Cape Town, Bangkok, and AWS’s European Sovereign Cloud, an offering designed for customers with data-residency requirements.
For AWS, the commitment locks in a customer that generates substantial compute consumption. Snowflake workloads — data warehouses, AI inference, and the agent workloads that are its growth bet — are heavy consumers of CPU and GPU cycles. Analysts said the arrangement is part of AWS’s broader push to ensure the AI build-out runs on its infrastructure rather than competitors’.
For Snowflake, the deal is a hedge against its biggest threat. Databricks, the rival data platform, has been aggressively courting the same customers with the same AI-on-your-data message, and it has deep relationships with Microsoft and Google. Tying itself more tightly to AWS, where most of its customers already live, makes switching harder and gives Snowflake a partner with its own AI ambitions, including Amazon’s Bedrock platform.
The arrangement is not without tension. Snowflake has long emphasized cloud neutrality, letting customers run on Azure and Google Cloud as well, and it still offers those clouds to clients. Analysts said the $6 billion commitment tilts the balance, even if Snowflake insists the door to other clouds remains open. The age of cloud agnosticism is getting expensive, one analyst said.
For investors, the deal answers the question the earnings report raised: where does the re-accelerating revenue growth come from? The answer, per the AWS agreement, is AI — and the willingness to pay for the compute to run it. Snowflake’s largest commitment to date is also a statement about where the company believes the next decade of enterprise software will be built: on someone else’s cloud, at least for now.
