Snowflake Inc. signed a $6 billion five-year agreement with Amazon Web Services that includes expanded use of Amazon's custom Graviton processors and AI chip infrastructure, the companies said Wednesday, as enterprises accelerate spending on cloud-based AI workloads.
"This agreement reflects the accelerating demand we're seeing from customers deploying AI workloads on our platform," Sridhar Ramaswamy, chief executive officer of Snowflake, said in a statement. "AWS's Graviton chips offer the price-performance we need at scale."
The deal nearly matches the $7 billion Snowflake has generated in total AWS Marketplace sales since its founding in 2012, according to AWS. Snowflake's customer spending on AWS doubled in 2025 to $2 billion for that calendar year alone. The new arrangement implies an average annual spend of $1.2 billion, up from $500 million in the final year of the prior $2.5 billion agreement signed in 2023.
The commitment comes as AI shifts from training large models to inference and agentic applications, which require massive general-purpose compute power. While Nvidia's graphics processing units handle training and reasoning, central processing units like Amazon's Graviton manage the orchestration and data movement behind AI agents — a workload that demands high CPU throughput. Amazon introduced its first Arm-based Graviton chip in 2018, and the processor has become the company's most successful custom silicon to date.
Snowflake separately raised its annual product revenue outlook to $5.84 billion for the fiscal year ending January 2027, up from a prior forecast of $5.66 billion and above the $5.68 billion analysts had expected. The company reported fiscal first-quarter product revenue of $1.39 billion, up 33% from a year earlier, with adjusted earnings per share of 39 cents. Analysts polled by LSEG had estimated $1.32 billion and 32 cents. Snowflake also said it acquired AI startup Natoma for an undisclosed sum.
The deal marks another major cloud customer choosing custom Arm-based processors over traditional x86 architecture from Intel and Advanced Micro Devices. Meta in April said it would deploy hundreds of thousands of Graviton chips for its AI compute needs, following a $10 billion commitment to Google Cloud months earlier. Anthropic, the AI model maker backed by Amazon, has pledged to spend over $100 billion on AWS over a decade.
For AWS, the Snowflake agreement validates its strategy of developing in-house silicon as a competitive differentiator. Amazon CEO Andy Jassy said on the company's April earnings call that Graviton allows Meta "to run the CPU-intensive workloads behind agentic AI with the performance and efficiency they need at scale." Amazon has also developed its own AI training and inference chips, which Jassy has said offer "better price-performance" than Nvidia's offerings.
Snowflake shares, which have a market capitalization of about $60 billion, surged as much as 33% in extended trading. The stock had gained roughly 12% year-to-date through Wednesday's close. Snowflake trades at approximately 10 times forward sales, a premium to some enterprise software peers but below its historical multiple as the company navigates its transition to an AI-native platform.
This article is for informational purposes only and does not constitute investment advice.