Snowflake Commits $6 Billion to AWS Over Five Years, Its Largest Infrastructure Deal Yet, With Graviton and GPU Compute at the Center
Snowflake's $6 billion, five-year AWS commitment is 2.4 times its 2023 deal, leaning on Graviton CPUs and GPU-accelerated EC2 to power agentic AI workloads.
Overview
Snowflake has committed $6 billion to Amazon Web Services over five years in what the data-platform company calls its largest infrastructure commitment to AWS to date. According to Snowflake, the company is “making a $6 billion multi-year infrastructure commitment to AWS, its largest to date, reflecting the accelerating enterprise demand for AI and data workloads running on AWS.” The agreement, announced on May 27, 2026, leans heavily on AWS’s custom Graviton processors and GPU-accelerated compute to power what both companies describe as agentic AI workloads.
The deal is structured as a five-year agreement, Cloud Computing News reported, describing it as a “five-year agreement with AWS valued at US$6 billion.”
What We Know
A steadily escalating commitment
The $6 billion figure marks a sharp escalation in Snowflake’s spending on AWS. According to The Next Web, Snowflake’s AWS commitment was $1.2 billion at its 2020 IPO and $2.5 billion in its 2023 renewal. The outlet notes that “the new $6bn agreement is roughly five times the 2020 commitment and 2.4 times the 2023 one,” and frames it as “the largest expansion of their 11-year relationship to date.”
Graviton and GPU compute at the center
The technical core of the agreement is AWS’s in-house silicon. Per Snowflake, the company “leverages AWS Graviton processors, delivering significant price-performance improvements for customers, and utilizing high performance, GPU-accelerated Amazon EC2 instances for AI model training and inference.”
Graviton is AWS’s custom Arm-based CPU line, The Next Web reported, describing it as “Amazon’s in-house Arm-server processor line, designed to replace x86 chips from Intel and AMD” and noting it is now in its fourth generation. The same report adds that “neither company disclosed which specific Graviton generation Snowflake is committing to,” and that Snowflake’s CEO indicated the company “will publish more detail at the AWS re:Invent conference later this year.”
AWS CEO Matt Garman tied the commitment directly to that silicon. “Snowflake has built on AWS since day one, and their deepened commitment to run on Graviton delivers the world-class performance, flexibility, and cost savings customers need to run data warehousing and AI workloads at scale,” he said, according to Snowflake.
The agentic AI framing
Snowflake positioned the deal around enterprise AI that acts rather than merely answers. “AI has generated enormous excitement, but for enterprises, the real challenge and opportunity is turning intelligence into action,” said Sridhar Ramaswamy, CEO of Snowflake, per the company.
The agreement covers Snowflake’s Cortex AI tools and Snowpark for development and machine-learning workloads, Cloud Computing News reported. The same outlet reported that the collaboration spans 10 new regions, including Auckland, Cape Town, Bangkok, and the AWS European Sovereign Cloud.
Two Snowflake customers were cited in support of the AI push. Daniel Block, General Manager of Revenue and Partnerships at Fetch, said: “With Snowflake Cortex AI, we’ve deployed a semantic agent that allows our sales teams to query campaign data in natural language and get instant insights,” according to Snowflake. Caitlin Colgrove, Co-Founder and CTO of Hex, added: “For teams using Hex to explore, analyze, and build with AI, having that layer be secure, governed, and performant isn’t a nice-to-have — it’s what makes enterprise AI adoption real.”
Marketplace and customer scale
Snowflake’s relationship with AWS extends well beyond the new compute commitment. The company reported it has 13,900 customers globally and has surpassed $7 billion in lifetime AWS Marketplace sales, according to Snowflake. Marketplace sales exceeded $2 billion in calendar year 2025, more than doubling year over year, Cloud Computing News reported.
What We Don’t Know
The companies have not disclosed which generation of Graviton Snowflake is committing to, and Snowflake has said it will share more detail at AWS re:Invent later this year, per The Next Web. The split between Graviton CPU spend and GPU-accelerated EC2 spend within the $6 billion total has not been broken out publicly. The terms governing the GPU instances Snowflake will use for model training and inference — including which accelerator hardware — were not specified in the announcement.
Analysis
The headline number matters less than its trajectory. A commitment that ran $1.2 billion at IPO, $2.5 billion three years ago, and $6 billion now signals that Snowflake’s growth is increasingly tied to AI-heavy compute rather than the data-warehousing workloads that built the business. The emphasis on Graviton — an Arm-based line AWS has positioned against x86 incumbents — also reflects a broader hyperscaler pattern of steering large customers toward custom silicon to improve price-performance, a theme that has recurred across recent AWS infrastructure commitments. Whether the agentic AI workloads Snowflake is betting on materialize at the scale the $6 billion implies is the open question the deal’s structure leaves unanswered.