Snowflake Commits $6 Billion to AWS in Major Deal Centered on Graviton and AI Chips

Snowflake has signed a sweeping $6 billion, five-year spending commitment with Amazon Web Services — a fivefold increase from its previous $1.2 billion arrangement struck at the time of its 2020 IPO. The deal covers a range of AWS services and technologies, with a particular emphasis on Amazon's custom Arm-based Graviton processors and cloud-based GPUs for AI workloads. The implied average annual spend of $1.2 billion signals a deep and durable reliance on AWS infrastructure.

The agreement is the latest in a string of massive AWS commitments from major tech players. Anthropic pledged over $100 billion in AWS spending over a decade in April, and Amazon also holds a deal with OpenAI. Unlike those AI model partnerships — which include equity investments — the Snowflake arrangement is purely commercial. Snowflake's relationship with AWS stretches back years: in 2023, it had already expanded an earlier deal to $2.5 billion, and it first signaled intent to use Graviton chips back in 2022.

The announcement coincided with a blowout earnings report that sent Snowflake shares surging as much as 36% in after-hours trading. The company posted $1.39 billion in revenue for its fiscal Q1 (ended April 30), up 33% year over year, with adjusted EPS of 39 cents — well ahead of analyst expectations of 32 cents and $1.32 billion in revenue. Forward guidance also topped estimates, with Snowflake projecting $1.415–$1.42 billion in Q2 product revenue at a 12.5% adjusted operating margin. Snowflake also announced the acquisition of AI startup Natoma for an undisclosed amount.

Broader context: Amazon's Graviton chip, first launched in 2018, is gaining fresh momentum as AI shifts from simple chatbot interactions toward agentic applications that require substantial general-purpose compute. Unlike Nvidia GPUs — which dominate AI model training through massively parallel processing — CPUs like Graviton are better suited to orchestrating complex, sequential AI workflows across multiple agents. Meta said in April it would deploy hundreds of thousands of Graviton chips for exactly this kind of workload. Google and Microsoft have since introduced their own Arm-based server chips, but Amazon's Graviton retains a first-mover advantage in the data center Arm ecosystem.