Snowflake signs $6B AWS deal for Graviton AI CPUs

Cloud data warehouse giant Snowflake has signed a $6 billion, five-year agreement with Amazon Web Services, marking one of the largest commitments to AWS's custom Graviton chips. The deal is almost equal to Snowflake's entire $7 billion in sales through the AWS Marketplace since 2012. Snowflake has always run on AWS, but this new contract reflects accelerating customer spending—doubling to $2 billion in 2025 alone—driven largely by AI adoption. Snowflake's Cortex AI tool, which enables natural language database queries and automated reporting, requires significant CPU resources for inference and agent-based workflows. As AI shifts from GPU-intensive training to CPU-heavy inference and daily automation, cloud providers are racing to deploy efficient, cost-effective processors.

The AWS Graviton chip, based on ARM architecture, offers better price-performance than Nvidia's offerings, according to AWS CEO Andy Jassy. This has lured other huge deals: last month, Meta signed for millions of Graviton chips after previously inking a $10B deal with Google Cloud. Meanwhile, Nvidia's CEO Jensen Huang recently announced its new Vera CPU, targeting a $200 billion market, and claimed $20 billion in pre-sales. While cloud giants like AWS, Google, and Microsoft are pushing their own chips to reduce dependence on Nvidia, the competition is heating up. This $6B Snowflake deal underscores how AI's infrastructure demands are reshaping the cloud chip landscape, with CPUs playing an increasingly pivotal role in production AI workloads.