Snowflake signs $6B AWS deal for AI CPU chips, boosting Amazon
▼ Summary
– Snowflake signed a new $6 billion, five-year agreement with Amazon Web Services, nearly matching the total $7 billion it has sold via AWS Marketplace since 2012.
– Customer spending on AWS through Snowflake doubled in 2025 to $2 billion, driven by demand for Snowflake’s AI tool, Cortex AI.
– The contract includes increased access to AWS’s homegrown ARM-based CPU chip, Graviton, which is used for AI tasks beyond training and reasoning.
– AWS’s Graviton chips offer a more affordable option than Nvidia’s, and Amazon passes those savings to customers, attracting multi-billion-dollar deals like one with Meta.
– The deal signals competition to Nvidia, though Nvidia’s CEO announced a new AI-specific CPU and claimed a $200 billion market, while cloud providers benefit from rising AI demand.
Snowflake has inked a $6 billion, five-year commitment with Amazon Web Services, the companies revealed Wednesday, marking a major expansion of their long-standing partnership.
Although Snowflake originally ran exclusively on AWS and now also operates on Microsoft Azure and Google Cloud, this deal’s scale is striking. AWS notes that since Snowflake’s founding in 2012, the data platform has generated a total of $7 billion in sales through AWS Marketplace. The new contract alone nearly matches that entire historical figure. Snowflake attributes this acceleration to a surge in customer spending on AWS, which doubled to $2 billion in 2025.
The driving force behind this growth is, of course, artificial intelligence. Snowflake has offered its AI-building tool, Cortex AI, for roughly two years. The tool makes sense because Snowflake already houses much of an enterprise’s data. Cortex AI enables features like a natural-language interface for database queries and automated summary reports.
A key detail in this agreement is Snowflake’s focus on gaining more access to AWS’s homegrown ARM-based Graviton CPU. As AI shifts from training to daily inference and eventually to autonomous agents, CPU demand skyrockets. While GPUs handle training and complex reasoning, CPUs manage most of the other tasks, particularly those supporting AI agents.
Amazon CEO Andy Jassy recently claimed that Amazon’s in-house AI chips offer better price-performance than Nvidia’s offerings, even though AWS continues to use Nvidia hardware. Demand for AI processing is so intense that cloud providers are deploying chips as quickly as possible. Many major AI models are still architected specifically for Nvidia, but Amazon’s chips provide a more cost-effective option, and Amazon says it passes those savings to customers.
These chips are now attracting multi-billion-dollar deals. Last month, AWS signed an agreement to supply millions of Graviton chips to Meta for its expanding AI compute needs. That was a significant win for AWS, especially after Meta had committed $10 billion to Google Cloud just a few months earlier.
These moves also send a clear signal to Nvidia that cloud giants are building competitive alternatives. Google has been developing its own AI chips for years, and Microsoft launched its Maia AI chip in January.
Unsurprisingly, Nvidia CEO Jensen Huang last week declared he is ready to defend and grow his territory. He described Nvidia’s new AI-specific CPU, Vera, as representing a “brand new” $200 billion market for the company. He also claimed Nvidia has already sold $20 billion worth of the chip, following another record-breaking quarter.
While Nvidia may not easily cede market share to any cloud provider, AWS’s multi-billion-dollar deals show how AI is lifting all boats in the cloud ecosystem. Regardless of which companies ultimately benefit most from AI’s integration into work and home life, cloud providers are securing their share.
(Source: TechCrunch)
